PDA

View Full Version : Radiosity, animated



Pages : [1] 2

DrStrik9
05-20-2009, 02:38 AM
I'm pretty sure this topic has been flogged many times. But I did a search, and didn't come up with much. It's what I'm thinking about today.

When you're doing stills, you can get away with relatively "trashy" radiosity settings, because you're only rendering one frame. But as soon as you animate radiosity, you suddenly see huge and terrible flaws which were irrelevant in gorgeous still imagery.

The objective then, with animated radiosity, is to create the most desired detail under the most believable lighting, with the least noise, in the lowest render times. Your life may depend on it. :-) Am I right?

Since upgrading from 8.5 to 9.6, I've found it easier to do DOF with radiosity in Lightwave's amazing new (new to me) renderer than radiosity *without* DOF. Seriously. Animated radiosity with DOF is easier than just animated radiosity. I guess the DOF helps hide low-radiosity's broken teeth. All this got me to thinking, in the same way DOF virtually "requires" certain settings to be believable under the watchful eye of efficiency, and the cruel brutality of animation, then so must radiosity. (I think he's on to something here.)

Of course, I can read the manuals, and try to interpret it all into my understanding of math and science, and I can do experiments and tests with radiosity under a wide variety of circumstances, and I can spend the next six months or year rendering every possible recipe, but it still takes a LOT more experience than that before you know exactly where to go the first time, with confidence, everytime, regarding radiosity settings.

I'm sure the proper settings can be affected by scene factors and many other variables. But a well-seasoned place to start (or four) would be oh so nice, say, as a gift, from the Master Ninjas to the new apprentices ... :-D lol

Or at least, let's talk about what has worked for us in rendering radiosity, animated, in 9.6, in various situations. But I would like this to be about radiosity animations only, even if they are crude in nature, and the radiosity settings we used, with render times.

I realize everything's an art, even looking at settings for any render type. Different systems, different purposes, and all that. It's not a race, or a d*ick size comparison, just getting a feel for it is all. And I bet it would help a lot of people to discuss this topic again right now. -- ?

Today I finally got a very acceptable result in a simple radiosity animation without shadows jumping all over the place, but at 960x540, it took over 20 minutes per frame. That would be over 80 minutes per frame at HD resolution. Is this acceptable? Fast? Slow?

If you're into it, let's just assume the quality is good and talk about render times, sizes, and settings. Let's not talk about systems though. We'll know by your times whether you have a fast system or not. It's the settings and lowest RELATIVE times on your system for acceptable quality that I'm interested in. :-) Good places to start.

I hope this isn't too forward or pushy. I just really want to know this stuff.

Help me. Thanks,
Michael

biliousfrog
05-20-2009, 03:17 AM
http://www.except.nl/lightwave/RadiosityGuide96/

SAHiN
05-20-2009, 05:16 AM
Today I finally got a very acceptable result in a simple radiosity animation without shadows jumping all over the place, but at 960x540, it took over 20 minutes per frame. That would be over 80 minutes per frame at HD resolution. Is this acceptable? Fast? Slow?



I understand your dillema, as I had to discover my own ways to get around animated radiosity.. 80 mins per frame sounds about right depending on your system.. I am not sure what kind of projects you are working on but, you have to be project-wise to use radiosity. I think u should check the tutorial posted on previous message.
If your lights are not moving in a scene, you may wanna cache radiosty to save render times.. My experience is every scene will require its own solutions when it comes to radiosity.. ( later on today I will try to post an animation with radiosity..maybe it will help to talk over that one)

biliousfrog
05-20-2009, 05:34 AM
SAHiN is correct, animated radiosity needs to be treated differently with each scene and only experience + trial and error will determine how well you manage with it. Caching can help a lot and testing the different radiosity solutions will yield different results but sometimes it's more effective to just bake to textures or even fake it.

Personally, I couldn't justify 80 minutes per frame even 20 minutes would be pushing it, I just can't afford to tie up the computers for that long. My showreel has a few scenes with radiosity using various methods, I'd say that the average render time would be 5-10 minutes at 1280x720 on a reasonable spec workstation. The scenes that definitely use GI are the container ship and arch-viz animations.

Bear in mind that GI has changed significantly since 8.5 as well as the cameras and AA. To achieve the best from 9.6 you need to learn how to use those areas effectively (see Except's guides) otherwise you'll be holding the application back by using out of date methods.

DrStrik9
05-20-2009, 09:08 AM
I've baked it before, a long time ago; what a hassle, depending on scene and contents (too many small items in that case). And I cached it for an earlier test in 9.6, which took so long to render that I did it in chunks at night, but lost track of my cache file. Oops. LOL

All the new lights, cameras, how AA is addressed, radiosity -- everything about the renderer, all of it is wonderful, and I sense very powerful, compared to earlier LW. Thanks much for the link to Except's radiosity guide. I'll start there. :-)

Michael

rezman
05-20-2009, 09:12 AM
Don't forget Fprime. While the native renderer has improved dramatically, you can still do the higher end rendering faster (in my experience) with Fprime.

And, I second baking as well.

DrStrik9
05-20-2009, 01:50 PM
Don't forget Fprime. While the native renderer has improved dramatically, you can still do the higher end rendering faster (in my experience) with Fprime.

And, I second baking as well.

Yes, but as of now, unless there's something I haven't heard about yet, FPrime is still not available for Mac Universal Binary, which means I'd have to go back to doing everything in CFM, which is slow and sometimes cranky.

The thing that got me started on this radiosity adventure is the Exit Portal light, with its animatable spherical map, for dynamically projecting colors and light into various parts of a room. This allows for some amazing looking stuff. I suppose it could be baked, but it is uber-cool when animated, even in subtle amounts. Which means no rad cache, no baking, and unless future FPrime can recognize and parse third-party lights, no FPrime either. (?)

Michael

LW_Will
05-20-2009, 02:34 PM
With no time limits, that 80min sounds reasonable, pro friend is up around 6-8 hours/frame. Beautiful stuff.

Anywho... "baking radiosity" is a very painless, if not instant thing. It requires that you set the render up and then run it. The computer notes the shifts in the radiosity and saves them to a file. Then it uses the file when it makes it computations. All handeled by the program. The old way was at least tedious and required checking each item, not the scene.

That Radiosity Guide is INVALUABLE. Read it. It will help.

DrStrik9
05-20-2009, 02:40 PM
I just read over Except's 9.6 Radiosity Guide, and I must say, this guy really knows his stuff. I will be doing some new tests, armed with more knowledge, and will undoubtedly be going back again and again to that resource.

Thanks again for pointing me in the right direction. :-)

Michael

GraphXs
05-20-2009, 08:18 PM
LWs Radiosity is great,but I hope they can improve the animated radiosity, I would love to see more presets for different situations. It would be great to have some settings for stills, animations, interiors/exteriors etc, that could get us started quicker and take some of the guess work out. Of course leave everthing optional/tweakable, but with so many options, rendering setup is becoming very complex. I just want to make pretty pictures as fast as I can.

DrStrik9
05-20-2009, 09:37 PM
With increased power comes increased complexity. The more choices you have, the more choices you have to MAKE (duh). :-) So I thankfully accept the complexity of GI. With presets would come the very real possibility of great looking radiosity renders at a huge cost in time. In a few hours, reading Except's Radiosity Guide and trying a few things, I've cut the render time down by almost 2/3 on my current simple test scene (from 20 minutes to 7!). The more frames in an animation, the more it pays to understand the controls and respect the power that comes from the complexity. When it comes to animated radiosity, any time savings is rewarded back to you geometrically. :-)

In another thread, I stated the desire for LW "smart zones," which could be something like an additional layer in LW, with various sets of settings and guides, plus links to online docs on everything, which could help the newcomer get acquainted more quickly. But the real issue would be knowing when to turn that baby-layer OFF, and step up to really learning WHY everything works as it does. LOL

I guess I'd rather be a scientist with skills than just a mechanic who knows what to do, but never learns why it works. (Having tools like Radiosity Flags makes it possible to really learn how GI works!)

Michael

clagman
05-21-2009, 11:44 AM
Sometimes it really helps to work with the "per object" GI settings.

GraphXs
05-21-2009, 11:53 AM
Yes, I do like all the complex GI stuff, and learning what each part does is important. I just wish it would work like surfaces, once you have something that gives you a start it would be nice to save GI presets. Exception has done a great job explanining all the aspects of LW GI. It just my feable brain forgets it. I've been spoiled with V-Ray (not that it can't be complex), but it has simple setting that get me going fast, i like that. In thorey LW GI also has simple GI setting, but its alot more trail and error.

joelaff
05-21-2009, 03:23 PM
I get good results in 5-20 min per frame (depending on the scene) using adaptive AA (0.17 threshold) Oversample 0.5-0.9 depending on how soft you want it. Higher will render faster as there will be less contrast to AA adaptively... And Monte Carlo Radiosity NOT interpolated, (directional rays only if you need them from reflective surfaces) Indirect bounces = 2, RPE = 9-14 depending.

This way you are letting the adaptive AA handle the noise removal instead of relying upon ridiculous RPE values. I don't like interpolated modes because I prefer the grain look over the blotchy "head cheese (http://en.wikipedia.org/wiki/Head_cheese)" look.

Note this is on dual quad core xeons or single Core i7s.

SAHiN
05-21-2009, 04:22 PM
GraphXs, I think VRAY is a totally different ball game.. It creates shaders of its own (similar to bakeing the radiosity).. No offense but All VRAY rendered stuff looks the same to me..its like a toast outta toaster, only the type of cheese used in it is different :)

I have to agree with joelaff, my settings are similar only I use 3 indirect bounces in most cases, so it would be wise to try the settings..

DrStrik9
05-21-2009, 11:16 PM
I get good results in 5-20 min per frame (depending on the scene) using adaptive AA (0.17 threshold) Oversample 0.5-0.9 depending on how soft you want it. Higher will render faster as there will be less contrast to AA adaptively... And Monte Carlo Radiosity NOT interpolated, (directional rays only if you need them from reflective surfaces) Indirect bounces = 2, RPE = 9-14 depending.

This way you are letting the adaptive AA handle the noise removal instead of relying upon ridiculous RPE values. I don't like interpolated modes because I prefer the grain look over the blotchy "head cheese (http://en.wikipedia.org/wiki/Head_cheese)" look.

Note this is on dual quad core xeons or single Core i7s.

LOL on the "head cheese (http://en.wikipedia.org/wiki/Head_cheese)" reference. Hee hee! :-)

I tried this on my little test scene with a moving light, and it turned out to be way too noisy. I went with 3 bounces, 14 RPE, Threshold 0.17, Oversample 0.5, and 5 AA passes. On a single quad core Xeon, it took 35:15. How many AA passes do you typically start with, or does it matter much?

Thanks. This is exactly what I was hoping for: recommended places to start. However, I think so far I prefer the interpolated methods, although the time spent testing various setting is going to be higher. Still, if a radiosity animation is very long, good setting pay back right away, with dividends.

Michael

DrStrik9
05-22-2009, 02:35 AM
I get good results in 5-20 min per frame (depending on the scene) using adaptive AA (0.17 threshold) Oversample 0.5-0.9 depending on how soft you want it. Higher will render faster as there will be less contrast to AA adaptively... And Monte Carlo Radiosity NOT interpolated, (directional rays only if you need them from reflective surfaces) Indirect bounces = 2, RPE = 9-14 depending.

This way you are letting the adaptive AA handle the noise removal instead of relying upon ridiculous RPE values. I don't like interpolated modes because I prefer the grain look over the blotchy "head cheese (http://en.wikipedia.org/wiki/Head_cheese)" look.

Note this is on dual quad core xeons or single Core i7s.

Ok, I'm gonna show and tell. (slower single quad core xeon) -- Please point out if I missed anything. Thanks.

First render is jolaff's recipe. Way too noisy (AA=5) Render Time here: 35:15.

Second render: same recipe with AA=18. Render Time: 2:00:00 (yes that's two hours) - and still too noisy, although the detail is definitely there. Still I don't think it would make a reality-believable animation.

Third render: Interpolated GI Monte Carlo 3 bounces RPE 200 SBR 100 AT 45-deg MPS 1.0/80 Multiplier 100 AA=5 Lanczos/Blue Noise PRMotionBlur 50% 1 Render Time = 10:05

Of course, I have some dancing stuff going on under the window, so will probably need to pay more somewhere. But even tripled, 10 minutes is a whole lot better than 2 hours. :-)

Michael

Oops noticed I uploaded psd files. Will add jpegs.

DrStrik9
05-22-2009, 02:39 AM
[QUOTE=DrStrik9;Oops noticed I uploaded psd files. Will add jpegs.[/QUOTE]

jpegs attached

toby
05-22-2009, 02:53 AM
I like the grainy look sometimes too, but *definitely* not for animations, it doesn't act like film grain at all. And it's still painfully slow. Do a crazy-fast interpolated render then add good film grain in post :)
The aliasing you have is probably the Lanczos, I get fewer artifacts than that with 1 pass gaussian, maybe even box.

Panikos
05-22-2009, 04:48 AM
I lost the count of how many hours/days/weeks/months I spent with GI.
Noiseless images are a result of compromise, that is : you may have a clean still image to masturbate with, but when time comes to animate you get inconsistent results.

I spent enough time. FPrime never disappointed me.
If you utilize all your rendering time with it, you get the best results.
It calculates everything, all changes, motions, displacements, etc

pooby
05-22-2009, 05:03 AM
Ive never managed to get anything near Fprime either, not if anything is deforming. Interpolated radiosity isn't designed for deformation animation and trying to get decent results out of it turned out to be a waste of time.
The best you'll get in any renderer of the sort (that I know) is some kind of frame/time interpolation.
Ie. a door slams shut but the light fades down over a few frames.
LW doesn't even do this. it'll splodge and boil.

joelaff
05-22-2009, 08:13 AM
Wow...

First off... always use Classic reconstruction filter with the perspective or advanced cameras. As confirmed by Matt Grainger, all the others are just post filters (though applied after each AA pass). If they are post filters then apply them in the post.

With Lanczos you are really shooting yourself in the foot. That is like running an unsharp mask after each AA pass and then expecting it to remove the noise. Ever try an unsharp mask on a noisy image with no threshold? It looks like garbage.

Use Classic "filter." If things are too soft in your end render then unsharp mask the whole thing in the post. This is what happens when you shoot film. Film is softer than the result with Classic. Though scanners, and people, apply USM to sharpen it up later.

If you still get noise then lower the threshold to 0.01. This will add another pass of AA. Though you may experiment with raising the RPE first (1 or 2 at a time), depending on whether or not you think the rest of the scene benefits from the AA, and how long each method takes.

BTW, always start with AA of 1. Only go higher than that if you are loosing small details (like really narrow polygons, like a car antenna). Even it that case you should rarely have to use an AA setting of greater than 4.

I always end up doing a few things to every render in post: Curves to add a slight filmic S curve, color correct, to pull out saturation usually, esp in the highlights, blur, and grain. You can often get away with more noise in the renders if you are adding more film grain in the post. For me, doing VFX, this is always to match either a background or foreground plate shot on film. Too much radiosity noise will not work, as it does not match the film grain at all. But if you get it down to a subtle level and then add a little grain on top you can get excellent results.

joelaff
05-22-2009, 08:15 AM
We used FPrime for years, and swore by it. Now with the latest 9.6 and still using my settings above I get comparable results that are abnout the same noise levels as FPrime in about the same amount of time or less sometimes (sometimes more, as FPrime really shines with radiosity, soft shadows from area lights, and soft reflections combined.)


I agree interpolated radiosity is a total waste of time, in LW and in Mental Ray, BTW. It is blotchy as all hell, flickers, and generally just sucks... Looks like a Quake map.

My biggest issue with FPrime is the lack of support for APS. We have been going displacement map crazy lately, adding a lot of realism. I also wish it could net render better. We need to be able to render things on the farm. If you only have one or two machines FPrime may ALWAYS be the way to go.

biliousfrog
05-22-2009, 08:58 AM
I agree, FPrime is pretty slow compared to a well set up scene in 9.6.

It's worth experimenting with cameras too. Often the perspective cam will be the one to use but sometimes the classic cam will be much faster...I had that with the last project I worked on, classic cam was almost 3x faster!

Material nodes can kill render times. Work out whether lighting quality is more important than materials and adjust to suit. Sometimes the lighting will be the star of the show and standard layered surfaces will be fine, othertimes you'll need complex materials but can get away with lower radiosity settings, it depends on what you're working with.

SAHiN
05-22-2009, 09:05 AM
Fprime does get good results fast..
And I have discovered, if you are using deformations on your objects (i.e bones, expressions, displacements etc.) dont change your subdivision order to "last".. Just set to "after motion" for fprime to reckognise it.. ( some of you may already know this, but I just discovered it) :)

pooby
05-22-2009, 10:03 AM
I agree, FPrime is pretty slow compared to a well set up scene in 9.6.


Are you talking about full un-interpolated radiosity? as I would love to see an example where LW beats Fprime speed wise.

Of course LW will beat Fprime if its using interpolated, because its simply a rough representation of radiosity that wont animate. Thats not a relevant comparison.

Panikos
05-22-2009, 10:14 AM
if you have IK/bones or any other displacement that often fails to show in FPrime, bake as MDD. You simplify things and FPrime works fine.

If you have a displacement that still doesnt work with FPrime, add a blank texture displacement on the geometry. FPrime will then see it.

Panikos
05-22-2009, 10:19 AM
Since we are on topic, I am almost sure that Newtek wont spent any more resources for FPrime bridging.

My desires to see instancing in FPrime, or Volumetric Lights wont happen.
I challenge anybody brave enough to comment. Is there anybody ?

DrStrik9
05-22-2009, 10:43 AM
OK, so here's what I'm getting from the last several posts:

Forget Lanczos for GI; it adds noise. Use Classic, which is the only reconstruction filter that isn't really a post effect.

Forget AA above 1 unless you find the need for it. (makes sense)

Forget interpolated GI; it's always blotchy. (So far I have found this to be true, but I still have hope ...)

Use FPrime for GI once it's released for Universal Binary. (Great hints there, Panikos!) But maybe not, because NewTek might not find it in their hearts to spend the resources to make the necessary changes to LW for FPrime to have better access. (?)

How'm I doin'? :-) I'm lost is how I'm doing. LOL

Richard Hebert
05-22-2009, 11:09 AM
I agree with joelaff. My current project is in SD video 720 x480. The scene is a spaceship landing in the parking lot of the church for an advertisement for the youth. I'm using Monte Carlo (not interpolated), 25 rpe's, with 2 bounces with AA of 4 and no Adaptive Sampling. Unless the model fills the frame my render times are 3 and 5 min. for a radiosity pass on an older Single Core, 1.5 GHZ Mac Mini. In the still frame I had to blur the fighter quite a bit to help blend the shots (don't have grain matching software yet) and the composite was done with After Effects 5.5 (not production bundle). Also, I'm using a Real Camera with Mitchell Sharp reconstruction filter. Blur can always be added in post.

joelaff
05-22-2009, 11:13 AM
OK, so here's what I'm getting from the last several posts:

Forget Lanczos for GI; it adds noise. Use Classic, which is the only reconstruction filter that isn't really a post effect.


I would say with the advanced or Perspective cameras forget all but Classic reconstruction filter, period. They made a little bit of sense with the classic camera, but not with the newer cameras.

What we really need (and I have asked for) is filters for how oversampling is applied. It is always applied box right now.

jay3d
05-22-2009, 11:16 AM
I agree interpolated radiosity is a total waste of time, in LW and in Mental Ray, BTW. It is blotchy as all hell, flickers, and generally just sucks... Looks like a Quake map.



I STRONGLY disagree (on the LW part, cuz MR interpolated really sucks :D ),
When I did some tests for a same scene of LW 9.6 GI compared to Vray, mr, modo, guess what?, It beats the hell out of those regarding details, speed, and animated GI.

I was very impressed with the test results, the new team really got it right this time.

I will follow up with some tests

joelaff
05-22-2009, 11:21 AM
I'd love to find some settings that actually work. However, I have yet to find interpolated settings that meet my standards. Post some anims (with scenes) if you can....

Everything is going to vary by scene. I am sure there are some scenes with interpolated may work OK.

Honestly I would rather spend my time creating than testing out all the different setting per scene. This is why I like FPrime and non-interpolated Monte Carlo. They just work.

jay3d
05-22-2009, 11:34 AM
I'd love to find some settings that actually work. However, I have yet to find interpolated settings that meet my standards. Post some anims (with scenes) if you can....
.

You can start with those settings and tweak depending on the scene

Samples: 512
Secondary: 256
Bounces: 3
Min Pixel: 0.5
Max Pixel: 12
Use Gradients

AA: 1
Threshold: 0.04 or 0.03, and use 0.02 and 0.01 for final rendering
Over Sample: 0.7 and u can add to it some sharpening filters, but ONLY with the oversample setting as it will greatly enhance the AA combined

joelaff
05-22-2009, 12:32 PM
You can start with those settings and tweak depending on the scene

Samples: 512
Secondary: 256
Bounces: 3
Min Pixel: 0.5
Max Pixel: 12
Use Gradients

AA: 1
Threshold: 0.04 or 0.03, and use 0.02 and 0.01 for final rendering
Over Sample: 0.7 and u can add to it some sharpening filters, but ONLY with the oversample setting as it will greatly enhance the AA combined

I tried this on our current project (detailed running shoes with no person in them). It rendered in about 40% of the time of the non interpolated, but it still had blotchiness animated along most edges (concave areas). It flickered and was unacceptable. I am trying now with min 0.25 max 5 RPE 768 SBR 384.

joelaff
05-22-2009, 01:01 PM
The new settings still flickered. I guess I will try one more... This time RPE 1024, SBR 512, MinPS 0.15, MaxPS 2.0...



Note my angular tolerance is 45 deg. Should I be adjusting that instead/also ?

jay3d
05-22-2009, 01:02 PM
I tried this on our current project (detailed running shoes with no person in them). It rendered in about 40% of the time of the non interpolated, but it still had blotchiness animated along most edges (concave areas). It flickered and was unacceptable. I am trying now with min 0.25 max 5 RPE 768 SBR 384.

Did u baked the animated radiosity? or u calculate every frame without caching?

joelaff
05-22-2009, 01:17 PM
Did u baked the animated radiosity? or u calculate every frame without caching?

Every frame. No cache. Rendering with BNR.

It has been too long since I have messed with this.. Is there a way to bake it using screamernet? Otherwise baking is out of the question in most production facilities. Nobody renders with a single machine in production.

Will "automatic" bake using the network? Or would I have to do Bake Scene on a single machine?

Note that my shoes are animated with bones...

Thx

jay3d
05-22-2009, 01:30 PM
Every frame. No cache. Rendering with BNR.

It has been too long since I have messed with this.. Is there a way to bake it using screamernet? Otherwise baking is out of the question in most production facilities. Nobody renders with a single machine in production.

Will "automatic" bake using the network? Or would I have to do Bake Scene on a single machine?

Note that my shoes are animated with bones...

Thx

Baking animated radiosity should be done on one machine since each machine has its own distribution of noise pattern, and that's fine because baking animated GI will store only the sample records which will be faster than full evaluation, and final shading will be done on network with no problems.

U welcome :)

jayroth
05-22-2009, 01:30 PM
Every frame. No cache. Rendering with BNR.

It has been too long since I have messed with this.. Is there a way to bake it using screamernet? Otherwise baking is out of the question in most production facilities. Nobody renders with a single machine in production.

Will "automatic" bake using the network? Or would I have to do Bake Scene on a single machine?

Note that my shoes are animated with bones...

Thx

Joe, you need to cache the GI solution if you do not want it to flicker. That is the whole point of the cache...

pooby
05-22-2009, 02:03 PM
I don't know what you have in your shot, but if it has anything deforming -ie bones etc, then it is not supported by LW's interpolated radiosity. The animation cache will work only for animated 'solid' geometry

joelaff
05-22-2009, 02:05 PM
I follow how the cache works. It all makes sense. I just didn't know if you Newtek wizards had come up with a distributed solution yet...


Here is the issue... With this cached interpolated monte carlo the processing time is now in computing the GI solution. It takes up about 90% of the per frame render time. So now I have to do this on a single machine. The actual render takes very little time. So distributing the render doesn't help much, esp since both the radiosity calculation and the render both need to compute the APS and nodal displacement.

I understand that I can reuse the cache. But most changes would invalidate the cache (animation, major surface changes, lighting changes) requiring me to start all over and bake on a single machine.

This method may be great for small projects where you don't need or have a render farm. But it is not very useful to tie up a workstation to do the baking (can you bake from the command line without a dongle?). So LW is relatively cheap, and you could certainly buy more dongles to do the backing, but you are still limited by the speed of your baking machine. All this to save maybe 50% on the render time. So if your farm is two machines then the savings is erased, not to mention the headache of the baking, and keeping the cache in sync, storing the cache files, etc.

What we need is some way to do distributed baking. I understand this is difficult. We need to maintain the samples for the various points from frame to frame, right? (It's not that each machine generates random numbers differently; that is not how PRNGs work. They take a seed and always produce the same sequence from the same seed.) Perhaps this is something that could use a distributed database (or just a SQL server). This may simply turn into too many transactions for this approach, but perhaps not if all of the existing points were queried once at the start of the frame. Then the new points could be evaluated and when the are inserted they could be inserted just a single time (like always use the first sample received or whatnot).

For now the non interpolated Monte Carlo still seem like the best bet for animations for anyone with a render farm. Using a single machine to do the most time consuming part of the render just doesn't make much sense.

Note also that my scene uses bones. Does the caching even work properly with deformations? Looks like no... So the question may be moot for this scene. But even if I didn't have bones the above concerns would still be valid.

Panikos
05-22-2009, 03:11 PM
The time you spend to tune LW GI and to type all these, FPrime would have it ready.

GI in outdoor cases is easier since geometry is exposed to "light" with ease in accessing.
The problems begin in indoors cases where accessibility is restricted.

joelaff
05-22-2009, 03:41 PM
The time you spend to tune LW GI and to type all these, FPrime would have it ready.

As would non interpolated Monte Carlo. Which is my point...

toby
05-22-2009, 03:46 PM
always use Classic reconstruction filter with the perspective or advanced cameras. As confirmed by Matt Grainger, all the others are just post filters (though applied after each AA pass). If they are post filters then apply them in the post.
This is what I used to think, but while they are post filters, they use 3d data, which you can't do in post. These two images are 1 pass aa with no adaptive, classic vs. gaussian. Guassian only took one-tenth of a second longer, and it would take several passes for classic to match it. This only works with the perspective cam, with classic cam it is just a simple blur.

toby
05-22-2009, 03:47 PM
I don't know what you have in your shot, but if it has anything deforming -ie bones etc, then it is not supported by LW's interpolated radiosity. The animation cache will work only for animated 'solid' geometry
I think that's only one of the two types of radiosity - final gather, but monte carlo works?

toby
05-22-2009, 03:58 PM
If you still get noise then lower the threshold to 0.01. This will add another pass of AA. Though you may experiment with raising the RPE first (1 or 2 at a time), depending on whether or not you think the rest of the scene benefits from the AA, and how long each method takes.

Adaptive of .01 will double your render times, with no appreciable results. Adaptive is not 100% efficient, which it would have to be in order to replace multiple passes of aa. 6 pass with .05 would be faster and better, things like textures will shimmer with only one pass. It's subtle, but you'd notice it if you compare the renders side-by-side. I've gone up to 12 pass with a visible improvement.

toby
05-22-2009, 04:07 PM
With Lanczos you are really shooting yourself in the foot. That is like running an unsharp mask after each AA pass and then expecting it to remove the noise. Ever try an unsharp mask on a noisy image with no threshold? It looks like garbage.
Totally agree. Seems that professional print people use it, but I can't understand why - it puts a white halo around sharp areas of the image - blah!

pooby
05-22-2009, 04:22 PM
I think that's only one of the two types of radiosity - final gather, but monte carlo works?

Neither work, but Monte Carlo will work if you dont interpolate it. However its much slower than Fprime doing the same thing.

joelaff
05-22-2009, 04:29 PM
Adaptive of .01 will double your render times, with no appreciable results. Adaptive is not 100% efficient, which it would have to be in order to replace multiple passes of aa. 6 pass with .05 would be faster and better, things like textures will shimmer with only one pass. It's subtle, but you'd notice it if you compare the renders side-by-side. I've gone up to 12 pass with a visible improvement.

I don't find it doubles the render time, and it definitely has an appreciable result with radiosity. Try it out. You will see significantly less noise. More samples...

As for textures. I start with a base AA of 1. If I have texture issues I will increase it. Though I might try blurring my texture first...

Note, however, that I do NOT subscribe to the ridiculous notion that MIP mapping and pixel blending should be disabled. Some newer "schools" for 3d seem to teach this as some matter of fact. IMHO, they have never done any real production and don't realize that MIP mapping is there for a reason (to prevent texture shimmer, etc.). I have gotten objects from freelancers where they turn off MIP mapping automatically (they told me to in school). These cause all sorts of problems that they then waste time and processor power fighting with AA. Most of the time texture aliasing issues are a result of disabling this feature...

joelaff
05-22-2009, 04:31 PM
This is what I used to think, but while they are post filters, they use 3d data, which you can't do in post. These two images are 1 pass aa with no adaptive, classic vs. gaussian. Guassian only took one-tenth of a second longer, and it would take several passes for classic to match it. This only works with the perspective cam, with classic cam it is just a simple blur.

Note I would never use Classic without Oversample. (0.5-1.5 depending).

Fine textures need more AA passes than a simple cube. The oversampled version looks comparable to the gaussian.

Mike_RB
05-22-2009, 05:39 PM
mitchell is my favorite AA filter, in all apps.

toby
05-22-2009, 09:51 PM
Note I would never use Classic without Oversample. (0.5-1.5 depending).

Fine textures need more AA passes than a simple cube. The oversampled version looks comparable to the gaussian.
But you see that the reconstruction filters are not just a post blur right?

Oversample (1.5 here) is more similar to a post blur than the reconstruction filters, and took a full second longer. This image looks exactly like one with jaggies that was then blurred in photoshop. Gaussian also softens textures, and it's what they use on big-budget movies so that's good enough for me :)

toby
05-22-2009, 09:55 PM
here's Lanczos with it's horrific white halo

toby
05-22-2009, 10:17 PM
Neither work, but Monte Carlo will work if you dont interpolate it. However its much slower than Fprime doing the same thing.
You're right, it's phorked. Took a half hour to render a deforming box at 480 res.

Anybody try using motion blur passes to reduce flicker? I'm having some good results with just 4 passes

DrStrik9
05-22-2009, 10:18 PM
Hey toby, thanks for the poop on Gaussian.

That recon filter could help defeat some noise, yes? It seems to complete the AA job nicely as well.

toby
05-22-2009, 10:23 PM
Hey toby, thanks for the poop on Gaussian.

That recon filter could help defeat some noise, yes? It seems to complete the AA job nicely as well.
Actually I'm not sure it'll fight noise anymore than other types of aa. It uses geometry data to work it's magic, so it only works on edges. Noise and textures still need more aa.

joelaff
05-22-2009, 10:29 PM
But you see that the reconstruction filters are not just a post blur right?

Oversample (1.5 here) is more similar to a post blur than the reconstruction filters, and took a full second longer. This image looks exactly like one with jaggies that was then blurred in photoshop. Gaussian also softens textures, and it's what they use on big-budget movies so that's good enough for me :)


I would appear they are not purely post blur, yes. I would like to get a better clarification from Mark Granger. In another beta thread he said they were "basically" post filters. I would be interested in knowing precisely how they tie in technically.

Oversample is not a post blur, though. According to Granger it blends the existing sample with surrounding pixels based on the current sample's position within the pixel (samples are subpixel).

I assure you that not all movies use Gaussian... It tend to use Gaussian in mental Ray, but MR works differently, and you can control the radius of the gaussian.

No doubt that Lanzcos and Mitchel, which both provide sharpening (i.e. white and black halos) are bad news, esp for animation. Sharpen after the render if you need to...

Obviously different subject matter and lighting techniques require different rendering techniques.

DrStrik9
05-23-2009, 07:30 PM
I never expected such participation in this thread. Thank you. I have learned HUGE amounts of stuff (not that I get it completely).

I think I feel just a little bit cynical at this point, about GI in general, at least for animation. I still dig Except's perspective though on interpolation, and he gets great imagery, animated. So it's possible. :-) What I need to do is begin to develop some content, and learn about all the new lights (wow!), and the still-elusive node editor. Then I'll add GI in appropriate situations, when needed, and only when it works. This has been a great priority lesson for me. :-)

You guys are awesome.

Michael

toby
05-29-2009, 06:48 PM
You're right, it's phorked. Took a half hour to render a deforming box at 480 res.

Anybody try using motion blur passes to reduce flicker? I'm having some good results with just 4 passes
Done more tests, looks like motion blur only minimizes flicker, can't seem to get rid of it, unless maybe if it were set high enough to take much longer than other methods.

Let's set the wayback machine for 2006, when Kray rendered this anim, with something like 8 bounces, no cg lights, no cache, in about two minutes a frame, and with deforming geo.
http://rapidshare.com/files/238721850/DrunkenZombie.avi

LW hasn't even caught up with that yet.

pooby
05-30-2009, 06:35 AM
and he gets great imagery, animated. So it's possible.

It is possible to animate rigid objects with the animation cache. Objects such as a chair rotating would work but not humans walking etc. Anything that means that the mesh is changing shape over time (deforming).
Its not a matter of opinion. Its officially documented that deformations are currently unsurported by interpolated GI.

Exception
06-01-2009, 05:06 AM
Drstrik: if you liked the radiosity guide I wrote, perhaps also check out the AA guide I wrote: http://www.except.nl/lightwave/aa/index.htm

If you have any specific questions, shoot.

DrStrik9
06-02-2009, 01:21 AM
Thanks, Exception. All your tutorials are very helpful. :-)

toby
06-02-2009, 02:41 AM
Done more tests, looks like motion blur only minimizes flicker, can't seem to get rid of it, unless maybe if it were set high enough to take much longer than other methods.

Let's set the wayback machine for 2006, when Kray rendered this anim, with something like 8 bounces, no cg lights, no cache, in about two minutes a frame, and with deforming geo.
http://rapidshare.com/files/238721850/DrunkenZombie.avi

LW hasn't even caught up with that yet.
Done some more tests with motion blurred radiosity and I'm getting very decent results with higher settings. I can't see the artifacts in my animation scene that were in the test scene. Ended up with 96 rpe, 0.5 min spacing and 6 passes classic motion blur (and it's enough passes that I can turn off aa too). Worth a shot if you want gi on a deforming animation. Otherwise I can also recommend DP's Dome light, set to use the backdrop image, your HDR. LW's dome works, but with only one, single, solid light color over the whole scene, the lighting is really boring.

jay3d
06-02-2009, 03:17 AM
Done some more tests with motion blurred radiosity and I'm getting very decent results with higher settings. I can't see the artifacts in my animation scene that were in the test scene. Ended up with 96 rpe, 0.5 min spacing and 6 passes classic motion blur (and it's enough passes that I can turn off aa too). Worth a shot if you want gi on a deforming animation. Otherwise I can also recommend DP's Dome light, set to use the backdrop image, your HDR. LW's dome works, but with only one, single, solid light color over the whole scene, the lighting is really boring.

Now that's what i was talking about :thumbsup:
glad u made some tests.

and in addition to that it will make striking results when combined with PR MB that still other engines failed to do properly

toby
06-02-2009, 03:40 AM
Now that's what i was talking about :thumbsup:
glad u made some tests.

and in addition to that it will make striking results when combined with PR MB that still other engines failed to do properly
Erm... it doesn't work with pr motion blur, that's why I used classic.
http://www.newtek.com/forums/showthread.php?t=99069
Perhaps someone with a PC can confirm it too?

Panikos
06-02-2009, 04:05 AM
Any shortcuts or compromises in accuracy lead to flickering results.
In order to get more consistent results you need to dedicate more time in the calculations.
But again, flickering is there, you cannot eliminate it completely unless you consider all parameters in the GI calculation.

jay3d
06-02-2009, 06:01 AM
Some test using PR MB multipass with Interpolated GI (aka Irradiance Caching).

plz tell me if u notice some flickering :)

Note: change .doc to .exe self extracting arch

jwiede
06-02-2009, 01:00 PM
Any chance you could provide it in a format that doesn't require us to blindly run it (and hope no malware snuck into the executable)?

shrox
06-02-2009, 01:10 PM
I understand your dillema, as I had to discover my own ways to get around animated radiosity.. 80 mins per frame sounds about right depending on your system.. I am not sure what kind of projects you are working on but, you have to be project-wise to use radiosity. I think u should check the tutorial posted on previous message.
If your lights are not moving in a scene, you may wanna cache radiosty to save render times.. My experience is every scene will require its own solutions when it comes to radiosity.. ( later on today I will try to post an animation with radiosity..maybe it will help to talk over that one)

Yes, the pain of HDTV. Rendering takes about 4 times as long as NTSC. I do try to squeeze by with 720, but I stiil aspire to 1080.

joelaff
06-02-2009, 01:15 PM
Any chance you could provide it in a format that doesn't require us to blindly run it (and hope no malware snuck into the executable)?

No doubt. Nobody needs self extracting archives anymore. Everyone has Zip support, and Zip compression is good enough. It's not the best, but it is good enough. None of them save much on our type of content anyway.

jay3d
06-02-2009, 01:27 PM
Here is it in rar format:

toby
06-02-2009, 04:48 PM
Here is it in rar format:
Was this multi-threaded, and did you try my test scene?

Btw, VRay was doing this with no issues, cache or not, and the same sampled motion blur, back in '06, and it was crazy fast.

jay3d
06-02-2009, 11:38 PM
Was this multi-threaded, and did you try my test scene?

Btw, VRay was doing this with no issues, cache or not, and the same sampled motion blur, back in '06, and it was crazy fast.

Yup i tested your scene, just change the angular tolerance to 22.5 as it will enhance the quality of rotational gradients on curved surfaces.

this one was multithreaded and it was rendering 15s on each frame on my dual core.

EDIT: make sure to enable Directional rays too, to make it even more compatible with PR MB

geo_n
06-03-2009, 01:19 AM
Some test using PR MB multipass with Interpolated GI (aka Irradiance Caching).

plz tell me if u notice some flickering :)

Note: change .doc to .exe self extracting arch

no flickering. but there's no other light source and simple object.
vray has anim prepass function in sp build that can do deformed radiosity easier.
and well.. Finalrender is tops.
Then again these renderers cost almost as much as lw.

This guy showed us his demo at work not really an fx specialist but a lighting specialist. used vray anim prepass he said.
http://www.vimeo.com/2413135
http://www.vimeo.com/3902592
http://www.vimeo.com/4468888

toby
06-03-2009, 03:34 AM
Yup i tested your scene, just change the angular tolerance to 22.5 as it will enhance the quality of rotational gradients on curved surfaces.

this one was multithreaded and it was rendering 15s on each frame on my dual core.

EDIT: make sure to enable Directional rays too, to make it even more compatible with PR MB
Good to know about the tolerance, thanks! It's hard to pin down exactly what that's doing; it changes the render slightly, but usually doesn't seem to make it better or worse. I did try Directional rays but it didn't change anything. Were you getting the same error to begin with?

jwiede
06-03-2009, 03:54 AM
no flickering. but there's no other light source and simple object.Yeah, given the fairly uniform and simple lighting, I wouldn't really expect to see flicker.

Jay3d, I suspect you'd see flicker if you had less uniform lighting on the swingarm as it moved, such that precise sample locations resulted in greater deviation across more visible surface area. Add different light sources (with different chroma/luma) to left and right of camera and see what happens.

In your movie, the surface areas which significantly change luma values are rather small, with the rest rather uniformly lit, so taking samples from different points on the surface likely yields the same general luma value in most cases. Interpolation samples just won't vary much by location in that kind of environment.

Exception
06-05-2009, 10:51 AM
Then again these renderers cost almost as much as lw.

This guy showed us his demo at work not really an fx specialist but a lighting specialist. used vray anim prepass he said.
http://www.vimeo.com/2413135
http://www.vimeo.com/3902592
http://www.vimeo.com/4468888

I don't see why you would need Vray for that. Lw can handle this just fine as well. sure, you need to learn your tools well, but Vray is certainly a lot steeper of a learning curve than the relatively easy and comprehensible LW radiosity engine.
VRay has some nice stuff that would be nice in LW, but in 99% of the cases where I see it being exclaimed as superior to LW, the case in question concerned little investigation by a single user in LW versus a multi-user grand production, as the ones you've shown.

In our own demoreel on our site we only used LW, never more than 1 person on a project, and hardly more than a few days project time, if not done in a single day. We hardly even render on a network. All of the content in that movie was rendered on a single machine and hardly more than one or two days for the longer ones.

Also, the non-interpolated MC mode with directional rays off is underappreciated in LW. With a very low ray setting it's not all that slow and remember it refines with each AA pass. Fully PR blur compatible too, and guaranteed to have no flickering. Sometimes the days you spend on figuring out a way to do something crazy with the interpolated modes by a person cost more than just having a few nodes render. It's all a matter of balance.

joelaff
06-05-2009, 11:09 AM
Also, the non-interpolated MC mode with directional rays off is underappreciated in LW. With a very low ray setting it's not all that slow and remember it refines with each AA pass. Fully PR blur compatible too, and guaranteed to have no flickering. Sometimes the days you spend on figuring out a way to do something crazy with the interpolated modes by a person cost more than just having a few nodes render. It's all a matter of balance.

Sounds like something I have been saying throughout this thread ;-)

Note that there is a bug with PRMBlur where you can see polygon edges on deforming geometry. I discovered that you can work around this by turning directional rays ON. (See http://www.newtek.com/forums/showthread.php?t=98901 )

toby
06-05-2009, 01:23 PM
Sounds like something I have been saying throughout this thread ;-)

Note that there is a bug with PRMBlur where you can see polygon edges on deforming geometry. I discovered that you can work around this by turning directional rays ON. (See http://www.newtek.com/forums/showthread.php?t=98901 )
And a bug where multi-threading causes banding - no workaround.
http://www.newtek.com/forums/showthread.php?t=99069

toby
06-05-2009, 02:22 PM
I don't see why you would need Vray for that. Lw can handle this just fine as well. sure, you need to learn your tools well, but Vray is certainly a lot steeper of a learning curve than the relatively easy and comprehensible LW radiosity engine.
Absolutely not. I learned vray without any kind of guide like the one on your site, just one tutorial and a little experimentation and I got great results in a production, and it actually rendered *faster* with pre-baked radiosity... imagine that. As I've mentioned, that was in '06, and although I have a much better understanding of how radiosity works now, I'm still stuggling with LW's bugs.

VRay has some nice stuff that would be nice in LW, but in 99% of the cases where I see it being exclaimed as superior to LW, the case in question concerned little investigation by a single user in LW versus a multi-user grand production, as the ones you've shown.

In our own demoreel on our site we only used LW, never more than 1 person on a project, and hardly more than a few days project time, if not done in a single day. We hardly even render on a network. All of the content in that movie was rendered on a single machine and hardly more than one or two days for the longer ones.
I don't see how this is an argument for which renderer is better -



Also, the non-interpolated MC mode with directional rays off is underappreciated in LW. With a very low ray setting it's not all that slow and remember it refines with each AA pass. Fully PR blur compatible too, and guaranteed to have no flickering. Sometimes the days you spend on figuring out a way to do something crazy with the interpolated modes by a person cost more than just having a few nodes render. It's all a matter of balance.
I've tried this, you are paying for the extra speed with quality. Here is 48 rpe (1 min.30 sec.) vs. 6 rpe (1 min.); low rays equals less accuracy even with more adaptive passes. Aside from the loss of saturation this has lost highlights and directionality - the lighting is much more even. Not to say this is useless, but it should be known that it is not as good as using it normally, there is a price to pay, just like interpolated.

Chuck
06-05-2009, 03:50 PM
Sounds like something I have been saying throughout this thread ;-)

Note that there is a bug with PRMBlur where you can see polygon edges on deforming geometry. I discovered that you can work around this by turning directional rays ON. (See http://www.newtek.com/forums/showthread.php?t=98901 )

Has this bug been reported via our FogBugz database yet? If so, let me know the number (no need to post the link, just the Case number itself), and I'll make sure we get a look at it for the hotfix.

Chuck
06-05-2009, 04:06 PM
And a bug where multi-threading causes banding - no workaround.
http://www.newtek.com/forums/showthread.php?t=99069

As noted in the thread, not a bug, but rather one of the classic use cases that led to developing radiosity caching. Formerly, to have multiple processors you would have had to have multiple systems and the problem would have exhibited between frames, resulting in flashing. With multi-processor CPUs you can see it in strips of a single frame, unless you turn on cache radiosity, or turn off interpolation. This is simply how these features work to give you a clean and consistent render within frame and between frames when using radiosity.

In the thread you mentioned that since anim cache didn't handle deforming geometry (a documented limitation), you were trying to use the photoreal motion blur in place of the cache. Unfortunately, the difference in strips can't be compensated that way - the Photoreal motion blur feature is simply not going to be able substitute for anim cache for you, since it's just going to apply its blurring algorithm to strips that are intrinsically different to begin with, leaving them still different but each appropriately motion blurred. Until we can improve the animated radiosity caching to accommodate deforming geometry, the user will not be able to use interpolated radiosity in this situation.

Exception
06-05-2009, 05:50 PM
...and it actually rendered *faster* with pre-baked radiosity... imagine that. As I've mentioned, that was in '06, and although I have a much better understanding of how radiosity works now, I'm still stuggling with LW's bugs.

Lw's pre-baked cache saves render time as well.
As for Vray's 'ease of use', if you try something a bit more difficult, I find one easily gets lost in any of the 5 billion settings. But that's my personal experience.



I don't see how this is an argument for which renderer is better -


It isn't because I'm not making that comparison. Vray and Lw are both just tools, and very good at various but different things. All I said that Lw's render engine is capable of some outstanding results, if only one would take a bit of effort learning to understand it.



I've tried this, you are paying for the extra speed with quality. Here is 48 rpe (1 min.30 sec.) vs. 6 rpe (1 min.); low rays equals less accuracy even with more adaptive passes. Aside from the loss of saturation this has lost highlights and directionality - the lighting is much more even. Not to say this is useless, but it should be known that it is not as good as using it normally, there is a price to pay, just like interpolated.

I think I would not define an RPE of 6 as a reasonable number. Something like 20~30 would constitute a 'low' number for me.

toby
06-05-2009, 07:24 PM
Lw's pre-baked cache saves render time as well.Sure it does, as long as you only need a still image.


As for Vray's 'ease of use', if you try something a bit more difficult, I find one easily gets lost in any of the 5 billion settings. But that's my personal experience.
I think you're comparing all of vray's panels to lw's gi alone. If you added anti-aliasing, environment and buffer saving into lw's gi panel, it would be about the same # of settings as vray's. Each of those panels is about the same size/complexity.


It isn't because I'm not making that comparison. Vray and Lw are both just tools, and very good at various but different things. All I said that Lw's render engine is capable of some outstanding results, if only one would take a bit of effort learning to understand it.
It is capable of outstanding results. But it's behind Vray, that's all I'm saying. Vray's ability to deal with deforming geo and animated cache that's faster than uncached, should make that clear.


I think I would not define an RPE of 6 as a reasonable number. Something like 20~30 would constitute a 'low' number for me.
Then your speed advantage is minimized, and it's still less accurate, so it's relative; save a lot of rendertime, quality suffers a lot - save a little rendetime, quality suffers a little. I just wanted people to be aware that it's not full-quality radiosity if you're letting the adaptive add a bunch of samples.

geo_n
06-05-2009, 08:11 PM
Yeah, given the fairly uniform and simple lighting, I wouldn't really expect to see flicker.


I guess its simple but still impressive use of pure GI. I haven't seen anything as good in lw with pure GI animation render.
Maybe there are so lets post some links.:thumbsup:

geo_n
06-05-2009, 08:21 PM
I don't see why you would need Vray for that. Lw can handle this just fine as well. sure, you need to learn your tools well, but Vray is certainly a lot steeper of a learning curve than the relatively easy and comprehensible LW radiosity engine.
VRay has some nice stuff that would be nice in LW, but in 99% of the cases where I see it being exclaimed as superior to LW, the case in question concerned little investigation by a single user in LW versus a multi-user grand production, as the ones you've shown.

In our own demoreel on our site we only used LW, never more than 1 person on a project, and hardly more than a few days project time, if not done in a single day. We hardly even render on a network. All of the content in that movie was rendered on a single machine and hardly more than one or two days for the longer ones.

Also, the non-interpolated MC mode with directional rays off is underappreciated in LW. With a very low ray setting it's not all that slow and remember it refines with each AA pass. Fully PR blur compatible too, and guaranteed to have no flickering. Sometimes the days you spend on figuring out a way to do something crazy with the interpolated modes by a person cost more than just having a few nodes render. It's all a matter of balance.

Nothing personal but your archi demo is very static compared to his. Nothing much moving which looks very much like a walkthrough only.
vray or kray would eat that kind of render any day.
If you would post your animated scene file in your tutorial maybe that would make it easier to understand how fast AC is in lightwave. Fast is a relative term.

joelaff
06-05-2009, 09:13 PM
I've tried this, you are paying for the extra speed with quality. Here is 48 rpe (1 min.30 sec.) vs. 6 rpe (1 min.); low rays equals less accuracy even with more adaptive passes. Aside from the loss of saturation this has lost highlights and directionality - the lighting is much more even. Not to say this is useless, but it should be known that it is not as good as using it normally, there is a price to pay, just like interpolated.

An RPE of 6 is just too low. I would start at 10 and go up until the quality meets your requirements for the scene.

Exception
06-06-2009, 03:56 AM
Sure it does, as long as you only need a still image.

Works just fine on fly throughs. You can save hours of render time with it. What gives you the idea that it doesn't?


It is capable of outstanding results. But it's behind Vray, that's all I'm saying. Vray's ability to deal with deforming geo and animated cache that's faster than uncached, should make that clear.

I'm not getting into this discussion. It's not very helpful. Try to render a hybrid line/cartoon/GI animation in Vray. In return, Vray is better in some other areas than LW. Like I said: different tools, different strength/weaknesses. The desire to constantly point out the 'winner' or 'loser' is irrelevant.
Also, there's Kray in case you need more control and speed in a Vray-like kind of fashion. Comparing LW's engine to Vray's engine is a little bit off if you consider that yuo have access to Kray before you shelf out thousands of dollars for a Max<>Maya/Vray combo or similar. Kray then also has its strengths and weaknesses, but it's the only render engine I know that has a new version out almost every week and it beats Vray for speed most of the time. Then again, Kray isn't free of bugs either and it's hard to get a look at it's power if you don't own it. The demo version's not been updated for years. I keep asking for that, and it looks like that this summer a new one will appear. The demo version gives a completely wrong picture of Kray, as it's been developing really fast for years.


Then your speed advantage is minimized, and it's still less accurate, so it's relative; save a lot of rendertime, quality suffers a lot - save a little rendetime, quality suffers a little. I just wanted people to be aware that it's not full-quality radiosity if you're letting the adaptive add a bunch of samples.

You're generalizing. It works great in many situations, and you have a lot of control between quality and speed. Wether or not it's useful depends on the situation... like always. I have a lot of experience with this render engine, and many others, and I do concur that LW's deformed animated radiosity is a hiatus in its feature set. It's been long announced it's being worked on, but that's certainly no guarantee, I agree. I look forward to improvements in LW's engine, but your bashing of it certainly doesn't give a neutral perspective. There's things you can do in it you can't do in any other engine, and for many applications it's fast and great looking. Certainly if you want to achieve a personal look, do actual creative imaging in it, LW is hard to beat. It's hard to deviate from the 'Vray look' in Vray, and non-photoreal or hybrid work is virtually impossible. That also goes for Kray.

Also, it's not a 'big discovery' to say that lw's animated cache doesn't work with deformation. It's officially listed, it's not a secret, it's in my guide and everyone will tell you this. There's nothing new here.


Nothing personal but your archi demo is very static compared to his. Nothing much moving which looks very much like a walkthrough only.
vray or kray would eat that kind of render any day.

I'm not sure if you ever made a complete arch-viz scene with one person in one or two days where everything was animated? To me that seems like an inhuman effort. Next time I get a job with a budget that accounts for that and a week's worth of work, I'll let you know :)

Everybody knows that deforming geometry with lw's render engine is not one of its strong spots (lots of workarounds or slow rendering). Use Kray or Fprime instead. That's a documented and well known property of the software. If there's no deformation, it works pretty well in animation. Your suggestion that the static cache doesn't save render time seems very odd to me, as it clearly does. How often do you use it?

I do have movies with a lot of stuff moving, such as furniture that completely unfolds together with moving stuff in an interior going all kinds of places, but I can't show it as they're under NDA. Also, I found Kray to be a great help here at times, but the non-interpolated MC held its own, especially when trading off setup time versus render time.

Exception
06-06-2009, 06:59 AM
also... on further investigation:

I've tried this, you are paying for the extra speed with quality. Here is 48 rpe (1 min.30 sec.) vs. 6 rpe (1 min.); low rays equals less accuracy even with more adaptive passes.

Adaptive passes are one thing. AA passes are another. MC refines with AA passes as well as adaptive passes. This gives you a lot of control, and gets you your AA virtually for free. Not to mention PR motion blur, DOF, sampled light quality, blurred reflections etc.

toby
06-06-2009, 08:23 PM
Works just fine on fly throughs. You can save hours of render time with it. What gives you the idea that it doesn't?
You're telling me that you've never had it take longer with animated cache on, like everyone else? You've never seen the threads discussing this?


I'm not getting into this discussion. It's not very helpful. Try to render a hybrid line/cartoon/GI animation in Vray. In return, Vray is better in some other areas than LW. Like I said: different tools, different strength/weaknesses. The desire to constantly point out the 'winner' or 'loser' is irrelevant.
Oh sorry, I forgot the cardinal rule on the forums; don't point out any problems with lightwave. You can talk about the strengths, but you can't talk about it's weaknesses. Fortunately there is no such rule, maybe there should be a rule about taking things personally when some simple facts are laid out.
I am not "constantly pointing out the 'winner' or 'loser". You seem to think that lw's gi is just as good as vray's, and that's flat-out wrong. Your inaccurate claims are the only reason I brought anything else up.


You're generalizing. It works great in many situations, and you have a lot of control between quality and speed. Wether or not it's useful depends on the situation... like always.
Yea that's exactly why I said "Not to say this is useless". You're reading WAY more between the lines of what I'm saying and completely missing what I'm actually saying. Here it is again. "I just wanted people to be aware that it's not full-quality radiosity". So don't accuse me of saying "lw sucks, vray rules" unless I actually say it, and I won't accuse you of being a fan-boy.

geo_n
06-06-2009, 09:29 PM
I'm not sure if you ever made a complete arch-viz scene with one person in one or two days where everything was animated? To me that seems like an inhuman effort. Next time I get a job with a budget that accounts for that and a week's worth of work, I'll let you know :)


yes. we do film set extension and use vray for that. Maybe 3-4 scenes by one person in day when we have those kinds of projects. But like everyone we already have some preset models,etc, who doesn't use them?:D
Its the actual rendering in vray that is fast though. The AA in lw is slow and the lighting quality for interiors is blotchy. That's why I mentioned kray quality but we're talking about AC in lw which is not fast and I think faking gi for animation in lw is still the better way to go or make a gi background walkthrough then fake lighting for animated elements. Front project it if possible.
But being objective and speaking of pure gi render, finalrender, vray or kray is just faster, otherwise you and I wouldn't be using kray for lw. :D

jay3d
06-06-2009, 11:28 PM
Well guys, I wanted to say that VRay is _not_ faster when it comes to accuracy.

If u want a fair comparison, start with a simple interior, one colored and with many details like grooves or something, and render a reference image in FPrime or LW brute force MC GI. then render the same model in LW Interpolated with the proper edges details settings, and in Vray preserving the details too.

My colleague tried to tickle with vray settings for whole day and he did not come close to what i have with an unbeatable time frame. in the end he told me that he enabled edge enhancement feature in Vray which is a variation of ambient occlusion rendering to overcome this limitation in vray.

The minimum pixel distance in LW works in sub-pixel, to a surprise no other renderer supports that except finalrender. this ability made LW GI engine a highly accurate for even tiny details in the model without even doing an ambient occlusion pass.

Exception
06-07-2009, 04:21 PM
You're telling me that you've never had it take longer with animated cache on, like everyone else? You've never seen the threads discussing this?

You never specified 'animated', you just said 'cache'. The static cache system does save time. The Animated cache system usually doesn't, but that's not a bug.

I've worked for very very long and very hard to beat out every possible bug from both the static and the animated cache system. I've filed hundreds of bugs and spent months on it. That effort did pay off, we got some options to control specific situations and all that, and many bugs were sent to bug-heaven. But I doubt it's bug free, but I have not seen one since. There are some situations where it goes out of control and creates bazillions sample points. These occurences are now marginalized with 9.6 over 9.5, and predictable (and documented). That said, AC is far from perfect, and I do hope to see improvements soon. I'm afraid it might only happen in CORE. (which would be too bad, really)


Oh sorry, I forgot the cardinal rule on the forums; don't point out any problems with lightwave. .... Your inaccurate claims are the only reason I brought anything else up.

I have no desire to argue with you. I never accused you of anything nor did I take anything personally. If you read my last response you'll see that I agree with you that AC is far from perfect. What do you want? Do you want an argument? I prefer to have a mature conversation.


we're talking about AC in lw which is not fast and I think faking gi for animation in lw is still the better way to go or make a gi background walkthrough then fake lighting for animated elements. Front project it if possible.
But being objective and speaking of pure gi render, finalrender, vray or kray is just faster, otherwise you and I wouldn't be using kray for lw.

Exactly. There's plenty of things you can "get done" in Lw's native engine, but that either take too long or are too fidgetty. Dedicated engines are great for these moments. AC is useful only to a degree, while it enables some things, it forces you to work in ways you rather wouldn't in most. I do look forward to AC being mature and being able to receive object deformations as well.

That said, I do still think many people give up before knowing the actual limits of LW's own render engine and start pointing at other engines as being superior. Vray is great, but has a very narrow focus, and the same goes for Kray. I tend to embrace the flexibility of LW's engine, and miss the creative control when using others.

We've done a lot of Vray projects, mostly when consulting with other firms. It has its share of issues, as you may know, and its sweet spots where it goes like the wind. I do hope Grzegorz will release the new Kray soon. It really rocks.

Exception
06-07-2009, 11:44 PM
So what is the "best" way (using the standard LW renderer) of rendering character animation with lots of deforming figures? :)


Any way you wish. But if you want GI, I'd say using non-interpolated MC is your safest bet. It works guaranteed and is dead easy to set up. Just takes time to render. Anything else has lots of buts, ifs and so on.

toby
06-08-2009, 03:47 AM
You never specified 'animated', you just said 'cache'.
Wtf? I said :

Sure it does, as long as you only need a still image.
which pretty much means animation.



I have no desire to argue with you. I never accused you of anything nor did I take anything personally.

You said :

your bashing of it

The desire to constantly point out the 'winner' or 'loser' is irrelevant.
Those are accusations.



If you read my last response you'll see that I agree with you that AC is far from perfect. What do you want? Do you want an argument? I prefer to have a mature conversation.

Oh please do point out whatever immature things I said.
LW gi has problems = immaturity?

allabulle
06-08-2009, 03:57 AM
And here we go again.

Not you two, please. Come on. We are all reading you and learning from all experiences. So please. :--)

toby
06-08-2009, 04:00 AM
Okay...

So what is the "best" way (using the standard LW renderer) of rendering character animation with lots of deforming figures? :)

Can it even be done?
That's what I've been pounding on for the last couple weeks (on-and-off). The closest I've gotten to a solution is with interpolated backdrop and classic motion blur w/6 passes. It's supposed to be stop-motion animation looking, so motion blur amount is at 1%, otherwise it would need a lot more passes. Render times are 15% faster than brute-force gi (non-animation interpolated is 400% faster). Other options I tried are DP's light dome, which allows you to assign an hdr for the light source, and OcclusionII which also supports an hdr. But both turned out a little slower for my scene.

geo_n
06-08-2009, 10:29 PM
Anyway in the end lw is fine. Its the cheapest alternative for now and a complete package not just a renderer. Not the best or fastest renderer though. Even mray raytracing is faster than lw raytracing.
I hope to see fast AC interior scene file with moving objects which is what the thread is all about.:thumbsup:

Panikos
06-09-2009, 12:53 AM
Its painful to have many tools but none of them provides everything

Megalodon
Plain LW course provides various buffers export.
As far as exrTrader output from FPrime, Lightwolf can illuminate you

Exception
06-09-2009, 06:11 AM
Wtf? I said : "Sure it does, as long as you only need a still image."

which pretty much means animation.

Since when does a still image mean animation? The static cache works great for animations when appropriate, and saves tons of render time.

Panikos
06-09-2009, 07:51 AM
Animation means to give life to things that do not have.
Walkthrough is not animation

Paul Brunson
06-09-2009, 11:57 AM
Honestly I stopped using Fprime for radiosity renders because in my experience Lightwave reaches a better quality image quicker. Especially when doing test renders and using the radiosity cache. So I don't even use fprime to preview radiosity.

I been using the animation cache, and network rendering with Butterfly. At first I had problems with renders that bogged down and went super slow, but then I read Exceptions excellent article on how the cache works and made some tweaks, its been working great.

Personally I would listen to Exceptions advise. It took several reads through his radiosity article before I fully understood Lightwave's radiosity. Its very powerful and can be very fast.

Now I use radiosity in just about everything, it adds a nice look and when you take the time to learn how all the settings work its not hard to narrow down on some fast, flicker free settings.

Chuck
06-09-2009, 01:20 PM
Animation means to give life to things that do not have.
Walkthrough is not animation

For CG work, animation also just means movement. The camera is animated, i.e. moves, in a fly-through and that animation (movement) requires multiple frames in sequence in order to be depicted (a still image, on the other hand, consists of a single frame), and also some considerable coding to properly support, or there would have been no need to develop the static cache to support radiosity rendering of camera-only animation.

"Nevertheless, it moves." ;)

geo_n
06-09-2009, 08:56 PM
Honestly I stopped using Fprime for radiosity renders because in my experience Lightwave reaches a better quality image quicker. Especially when doing test renders and using the radiosity cache. So I don't even use fprime to preview radiosity.

I been using the animation cache, and network rendering with Butterfly. At first I had problems with renders that bogged down and went super slow, but then I read Exceptions excellent article on how the cache works and made some tweaks, its been working great.

Personally I would listen to Exceptions advise. It took several reads through his radiosity article before I fully understood Lightwave's radiosity. Its very powerful and can be very fast.

Now I use radiosity in just about everything, it adds a nice look and when you take the time to learn how all the settings work its not hard to narrow down on some fast, flicker free settings.

Would you mind posting some animation you made with AC? How fast and what system and single pass render? I've yet to find it fast enough to render especially now that everyone wants hd size. For walkthrough yes, most cache system is fast and flicker free. So animated objects are better added later in post. And I've mentioned it before that its the AA in lw that is slow so the final output with the gi calculation is inevitively slow. kray for example has really good AA and very fast. The same time it does AA the quality is high compared to lw AA.

Panikos
06-10-2009, 12:17 AM
I understand that the current LW GI has some weaknesses and its good to hide them to keep everybody "happy".

jay3d
06-10-2009, 02:19 AM
All 3D apps have limitations, and no one is perfect, this is a well known fact.

In 1998 Pixar did an animation and rendered it using PRMan, and guess what, without even raytracing(and GI), yet they did an amazing lighting & character animation.

I don't mind if full bones deformations are not supported _yet_ properly in lW animated GI (yet some deformations work with animated cache), but one can still do amazing rendering & character animation using what is currently available.

check out A Bug's Life :)

Panikos
06-10-2009, 02:31 AM
I agree. However I cannot interpret Camera movement as animation

toby
06-10-2009, 04:15 AM
All 3D apps have limitations, and no one is perfect, this is a well known fact.

In 1998 Pixar did an animation and rendered it using PRMan, and guess what, without even raytracing(and GI), yet they did an amazing lighting & character animation.

I don't mind if full bones deformations are not supported _yet_ properly in lW animated GI (yet some deformations work with animated cache), but one can still do amazing rendering & character animation using what is currently available.

check out A Bug's Life :)
I don't know anyone who thinks that a bunch of spotlights with shadow maps, or Bug's Life, are amazing lighting. And they had more than we do now as far as shading is concerned. They didn't just use a bunch of plug & play shaders you know. And when was the last time you saw something 'amazing', or photoreal, with no raytracing? Or with no gi for that matter?

Exception
06-10-2009, 04:32 AM
I understand that the current LW GI has some weaknesses and its good to hide them to keep everybody "happy".

What's with all this?
My guide CLEARLY lists the shortcomings of the cache. We CLEARLY established there are areas of the Animated cache that could do with substantial improvement.
Just because you claim that the word "animation" consists of more than an animated camera and others disagreeing doesn't mean anyone is "hiding" anything. Actually, I think the issues was pretty clear... Toby referred to a Still Image, which is a single frame, and not animation in any sense of the word, no matter how you interpret it.

Also, you can do animation with the static cache, just the lighting won't change. This is useful in some cases. (floating things mostly, cartoons and special effects).

While our terms may differ, I can assure you that in my line of work, an 'animation' means a fly through where mostly the camera is animated. Then to move cars and such I still use the static cache (shadow becomes a negative spotlight pointing down, etc).

Panikos
06-10-2009, 05:25 AM
Exception, if you see the time, its 2009 now.

These things were WOW some years ago.

jay3d
06-10-2009, 07:24 AM
Exception, if you see the time, its 2009 now.

These things were WOW some years ago.

What things? :stumped:

Panikos
06-10-2009, 07:32 AM
Baked GI and walkthroughs. I remember when I saw Lightscape, it was like 10 years ago.

jasonwestmas
06-10-2009, 07:42 AM
Animation means to give life to things that do not have.
Walkthrough is not animation

I agree but lack of deformation doesn't mean an object can't be animated either. Animation can strictly mean rotation and translation. :D

Exception
06-10-2009, 07:59 AM
These things were WOW some years ago.

Panikos, who's saying they are WOW now?

And... nice change of subject.
It's like a conversation with your grocer:
A: "Hello sir, do you have apples?"
G: "I painted my roof yesterday."
A: "Really? What color did you paint it?"
G: "I don't like your shirt."
A: "Why... well why don't you like my shirt? There's nothing wrong with it, is it?"
G: "Tomorrow is my sister's birthday."
A: "....Eh... any apples?"
G: "Sometimes my knees itch"
A: <flees>

Dr. Sbaitso comes to mind. "Do you want to talk about your mother?"
:)

Pff this discussion... If you like LightScape so much, here's a free LightWave to LightScape and back tutorial (http://www.except.nl/lightwave/lslwtute/lightscapelightwave.htm) I wrote 5 years ago. Have fun tesselating.

http://www.except.nl/lightwave/lslwtute/flow%20lw-ls-big.jpg

CC Rider
06-10-2009, 08:49 AM
I am fairly new to LW and have a lot to benefit from these forums and from the more experienced users here.
It's disappointing to me to click on a link when I see a possibly valuable topic/conversation devolve into the childish back and forth that tends to so often overcome so many of these threads...
What a waste...:mad:

DrStrik9
06-10-2009, 09:56 AM
Yes, I was able to start this thread easily enough ... now where is the STOP button? :-)))

Still, I've learned a great deal in the process.

Peace

Panikos
06-10-2009, 10:03 AM
A short thread means a dead forum

Exception
06-10-2009, 11:10 AM
A short thread means a dead forum

I painted my roof yesterday.

CC Rider
06-10-2009, 11:21 AM
I painted my roof yesterday.

...don't forget to put the light out and turn off the cat before you go to bed...

GraphXs
06-16-2009, 07:46 PM
So what is the way to get GI to work with deformations? Dowe just not use the cache? Just use MC straight? Or do we do the spinning light trick? V-Ray supports deformation, not sure about the caching but we used it for attract character animations and worked great.

Anyone ever use GI for animated characters in LW, or just standard raytracing and tricks?

toby
06-16-2009, 10:52 PM
So what is the way to get GI to work with deformations? Do we just not use the cache? Just use MC straight? Or do we do the spinning light trick? V-Ray supports deformation, not sure about the caching but we used it for attract character animations and worked great.

That's what started the previous argument.


Anyone ever use GI for animated characters in LW, or just standard raytracing and tricks?
Backtrack to about post #60, we talked about this a bit, I was in the process of figuring out what method to use, and the different types (interp. w/motion blur, brute-force, 3 kinds of amb-occ, DP Dome light with hdr, area lights) are not too far apart in speed/quality ratio, so it will depend on the scene.

gerardstrada
06-17-2009, 04:05 AM
For deformed objects, I found two solutions depending on the scene (outdoors/indoors) and depending on the way that we are solving the surfacing phase (channels, shadings or materials). These two solutions are two extremes; what it means we can pick an intermediate solution by depending on each case:

I call the Split Scene Method to the first solution: The basic idea is to treat the scene as it was a VFX shot. That is to say, the integration of a CG element with a real footage. Difference here is that our footage is not a real one but the same CG environment as a background plate.
Idea is to make a master scene (we can manage the multipass setup here) and we separate, in two different scenes, what it gets moved from what not gets moved.

All what not gets moved/deformed (locations, interiors, surroundings) is solved and rendered as it was a flythrough (baking radiosity/caching radiosity - do check the excellent Exception's tutorial), so no flickering problems here or unexpected results. Then save these sequences as EXR.

In the second scene, we treat these rendered sequences as a background plate (we can compose it in post or even project them over the static geometry). Then we solve the elements that get moved in two ways:
In the old school way (IBL rigs, domes, spinning lights, bouncelights in post or gradients, etc), or with the Modular Illumination method.

I call the Modular Illumination method to the second solution: This method commonly works fine to solve a whole scene (though it works better for outdoors). It's basically a hybrid between the new LW radiosity and multipass rendering and/or the old lighting tricks. Difference with the previous method is that here we don't isolate objects in different render passes but rather we isolate the lighting components of the whole scene in different modules by solving each module with the best technique that suits our needs.

Thus, the distant scene can be solved with IBL rigs (just for people not familiarized with this term, the distant scene is not what is far away, it's the direct lighting component). It can be solved also with BGRadiosity but an IBL rig is better since it's faster and we get the proper definition of diffuse shading (at this moment, diffuse models are not respected by GI solutions in LW). Besides specular shading is faster than blurred reflections too. This, obviously implies to convert all materials to shadings (Split Material and Material Blender are very useful here). Also, IBL rigs commonly help to define better the RO pass (exported in the shadow buffer) where some times there's no necessity of even making an additional AO pass.

The local scene is solved with real GI (commonly Final Gather - yes for animation). Again, for people not familiarized with this term, the local scene is not what is near, it's the indirect lighting component. The indirect lighting component can be decomposed in two modules. The radiance module (color bleed provided by the bounce light effect) and the radiance occlusion module - RO - (the shading provided by the same bounce light effect). Since what we need here is the first module (color bleed), we can use low values in RPE without the Interpolated option (in some outdoor scenes we can diminish the value for Multiplier option too). In some cases an AO pass could be necessary, but depending if our IBL rig provides RO data, we could skip that AO pass. Later in post-processing, we isolate the color bleed module, softening and 'de-graining' it if is necessary, and mixing it with the RO pass provided by the IBL rig, or a new AO pass.

Maybe I should write a detailed tut about this :)



Gerardo

SAHiN
06-17-2009, 05:34 AM
I always use GI and I always use characters walk around, sleep, wash their faces, eat, drink and if you wish I can get them to have sex..

I have no problem whatsoever with LW rendering characters deforming with GI ...

Whats this chatter all about man..?

Here is a frame out of an animation..
Camear starts at dome ceiling, moves down catches a man walk in from main entrence, while a woman walks off screen and another man walks in and walks to counter.. While the woman that walks outside the hotel gets refracted by the rolling doors as the man enters the hotel..

Enough animation for you Panikos? Would you like them to do some tai-chi while they are at it ??

Exception you cracked me up with that lightscape tutorial.. ROFL :D

:lightwave

Panikos
06-17-2009, 06:01 AM
If I want everything, I turn on the camera and I have real time GI with all features that god provided. No need to compose or render :)

geo_n
06-17-2009, 06:06 AM
I always use GI and I always use characters walk around, sleep, wash their faces, eat, drink and if you wish I can get them to have sex..

I have no problem whatsoever with LW rendering characters deforming with GI ...

Whats this chatter all about man..?

Here is a frame out of an animation..
Camear starts at dome ceiling, moves down catches a man walk in from main entrence, while a woman walks off screen and another man walks in and walks to counter.. While the woman that walks outside the hotel gets refracted by the rolling doors as the man enters the hotel..

Enough animation for you Panikos? Would you like them to do some tai-chi while they are at it ??

Exception you cracked me up with that lightscape tutorial.. ROFL :D

:lightwave

Looks like radiosity with area lights. Shadows aren't deep enough to notice flicker though in this situation. Looks good nonetheless. :thumbsup:

SAHiN
06-17-2009, 06:20 AM
If I want everything, I turn on the camera and I have real time GI with all features that god provided. No need to compose or render :)

it seems to me like you dont know what you want dude..
Maybe I could interest you on a Star Trek Replicator technology ?
That way you can say "Computer 20 seconds of GI character animation. daytime"

..and you would get it huh ?

You must be a Vray jokey.. where all is ready for you to click away and you can turn around and say " I made this" :neener:

SAHiN
06-17-2009, 06:24 AM
Looks like radiosity with area lights. Shadows aren't deep enough to notice flicker though in this situation. Looks good nonetheless. :thumbsup:

Thank you..
The thing about radiosity is,
each scene has a different requirement when it comes to setting it up.
I worked on film sets for years, and I know when it comes to lighting;
lighting a real life scene isn't that easy either..

it just takes some thinking and will to create...
When it all fails, start thinking :)

Panikos
06-17-2009, 06:26 AM
it seems to me like you dont know what you want dude..

Its better for you not to mess with my desires :agree:

geo_n
06-17-2009, 06:37 AM
Thank you..
The thing about radiosity is,
each scene has a different requirement when it comes to setting it up.
I worked on film sets for years, and I know when it comes to lighting;
lighting a real life scene isn't that easy either..

it just takes some thinking and will to create...
When it all fails, start thinking :)

Do you have animation with more contrasting or direct lights and hard shadows that are not shadow maps? The shadows are too faint on the stills.

SAHiN
06-17-2009, 07:17 AM
I will check my work and see if I have..
I usually do what my clients need and clients dont like hard shadows and hi-contrast animations..

They wanna sell property after all.

Scazzino
06-17-2009, 10:27 AM
So what is the way to get GI to work with deformations? Dowe just not use the cache? Just use MC straight? Or do we do the spinning light trick? V-Ray supports deformation, not sure about the caching but we used it for attract character animations and worked great.

Anyone ever use GI for animated characters in LW, or just standard raytracing and tricks?

I still use spinning lights for animation. Especially for deforming objects that need motion blur since I need to use classic motion blur passes for that, and the motion blur passes are when spinning lights are blended, so the spinning lights are basically free...

SAHiN
06-17-2009, 01:35 PM
This is the part of that animation.
I am really bad at compressing these animations for web. Somehow I can never get the quality and small file size together..I dono how people do it.?

Anyways, I tried to get as much quality as I could. Since the subject was flickering, I didnt want bad codec to make my animation flicker hey ? :)

Keep in mind this animation has no post whatsoever on it yet..its directly outta LW renderer.

I will try to write some tutorial about it with some scene screen shots etc.

I hope I can help solve this mess somehow

:lightwave

SAHiN
06-17-2009, 01:36 PM
Why do you need to use classic motion blur for deforming objects? Does photoreal not work? Or does it take way too long?

Oh man I hate that photoreal mo-blur also..
it takes forever to render..

I also use classic mo-blur.

DrStrik9
06-17-2009, 02:27 PM
This is the part of that animation.
I am really bad at compressing these animations for web. Somehow I can never get the quality and small file size together..I dono how people do it.?

I use Quicktime Pro ($30 from Apple.com), and "export for web" using the "desktop" preset. I took a 74.5 mb movie this morning and exported it this way and it was 2.5 mb when exported. The finished format is .m4v (MPEG-4 video), a little smaller than the original, and really looks great considering the size reduction. But as with all quality considerations, your mileage may vary. :-)

Cheers

SAHiN
06-17-2009, 02:32 PM
One more thing guys...

What are you rendering at? HD resolution? 1920x1080? What fps? 24p, 30p, etc? And why THAT resolution and frame rate? Always interested to know! :)

I always render 1280x720 then blow it up to 1920x1080 in DF..
This way I get faster render times and blowing the image gives me less CG looking smooth result.

SAHiN
06-17-2009, 02:51 PM
I use Quicktime Pro ($30 from Apple.com), and "export for web" using the "desktop" preset. I took a 74.5 mb movie this morning and exported it this way and it was 2.5 mb when exported. The finished format is .m4v (MPEG-4 video), a little smaller than the original, and really looks great considering the size reduction. But as with all quality considerations, your mileage may vary. :-)

Cheers

Cool..Thanks for this info mate..

Cheers

gerardstrada
06-17-2009, 04:52 PM
I still use spinning lights for animation. Especially for deforming objects that need motion blur since I need to use classic motion blur passes for that, and the motion blur passes are when spinning lights are blended, so the spinning lights are basically free...

I also use IBL rigs based on the spinning light trick. L I G H T B I T C H works great and it provides a predictable result without mblur for previewing. If more speed is needed, a rig similar to TGTALH is faster. The new PR mblur works also with the spinning light trick, the issues came with the forward mblur, which messed up all my old light rigs.



Gerardo

geo_n
06-17-2009, 06:31 PM
This is the part of that animation.
I am really bad at compressing these animations for web. Somehow I can never get the quality and small file size together..I dono how people do it.?

Anyways, I tried to get as much quality as I could. Since the subject was flickering, I didnt want bad codec to make my animation flicker hey ? :)

Keep in mind this animation has no post whatsoever on it yet..its directly outta LW renderer.

I will try to write some tutorial about it with some scene screen shots etc.

I hope I can help solve this mess somehow

:lightwave

The animation has lots of flickering especially on the entrance, under the counter and mostly on horizontal lines. I don't think gi is causing that or video compression. Its ok for archiviz but this won't be acceptable for tv cm and film. Lighting looks good though but video is jittering or flickering.

SAHiN
06-17-2009, 06:44 PM
The animation has lots of flickering especially on the entrance, under the counter and mostly on horizontal lines. I don't think gi is causing that or video compression. Its ok for archiviz but this won't be acceptable for tv cm and film. Lighting looks good though but video is jittering or flickering.

Not on the actual renders mate..

toby
06-17-2009, 09:39 PM
That's *totally* GI flickering. Compression flickering would show up everywhere, not just in tight places, where gi flickering does happen. I've been compressing for years now and I've never seen artifacts like that. I can tell that your min. pixel spacing is well above 1, I'd guess 4. There is a little aliasing, but both compression and aliasing are more consistent, they don't pop in and out like these gi artifacts are doing.

If you're still sure, can you post 5 or 6 high quality .png images from the sequence? (5 - 6 in a row that is)

Scazzino
06-17-2009, 09:54 PM
Why do you need to use classic motion blur for deforming objects? Does photoreal not work? Or does it take way too long?

Right it's not recommended for deforming objects and I'm also using HD Instance which also needs classic motion blur rather than photoreal motion blur... So since I need classic motion blur passes anyway, spinning lights come for free...

SAHiN
06-18-2009, 07:24 AM
That's *totally* GI flickering. Compression flickering would show up everywhere, not just in tight places, where gi flickering does happen. I've been compressing for years now and I've never seen artifacts like that. I can tell that your min. pixel spacing is well above 1, I'd guess 4. There is a little aliasing, but both compression and aliasing are more consistent, they don't pop in and out like these gi artifacts are doing.

If you're still sure, can you post 5 or 6 high quality .png images from the sequence? (5 - 6 in a row that is)

You may be a compression specialist, but I am a specialist at what I do.
...and my clients don't pay for things that flicker..
I worked on tv & film industry very long time and I can tell u that those industries can't even come close to how fussy advertising clients can..

I did my say, I sent some work to show and help ppl it can be done.
I see no further point at trying to deal with egos..

cheers

geo_n
06-18-2009, 08:47 AM
http://www.vimeo.com/2413135
http://www.vimeo.com/3902592
http://www.vimeo.com/4468888

let me just bump this from the vray guy. This is pure gi and no post, no cheats, no light rigs, (which I think I wouldn't do before vray anim prepass was available). the video compression seems to affect straight lines in vimeo so I think that's what happened to SAHiN video. And this is highly compressed. I got the hq version of the demo somewhere.
But in SAHiN video, what I notice is that the polys itself look like they're warping especially where GI would sample the hardest. A lot of colors flickering. I'm not sure where the error is. It could be GI or textures dancing.

SAHiN
06-18-2009, 03:14 PM
This is very good to know since I may also be using HD Instance. Thanks! :thumbsup:



But... but... please deal with my severely depressed ego. :)

Really SAHiN.... can you show us how you accomplished your GI workflow? I'll be doing quite a bit of deforming animation in a few months and would REALLY love to know (details) how you set these scenes up. No flicker with deforming character animation and relatively quick radiosity rendering would be a BIG plus. I'm looking into the other workflows posted here as well, but I'm VERY curious as to what your workflow is.

Again, thanks!

I am sure many people here can tell you why some things can't be achieved mate.. It seems Toby already discovered and revealed the secrets of the trade above..Read that and try it :)
I am done in this topic.. I got animations to finish.

Cheers

toby
06-18-2009, 03:24 PM
You may be a compression specialist, but I am a specialist at what I do.
...and my clients don't pay for things that flicker..
I worked on tv & film industry very long time and I can tell u that those industries can't even come close to how fussy advertising clients can..

I did my say, I sent some work to show and help ppl it can be done.
I see no further point at trying to deal with egos..

cheers
That's a great way to demonstrate how great lightwave is; denial, and accuse people of having ego problems. Sorry if I bruised yours, but if something looks like gi flickering, I'm not going to act like a fan-boy and make excuses for it.


let me just bump this from the vray guy. This is pure gi and no post, no cheats, no light rigs, (which I think I wouldn't do before vray anim prepass was available). the video compression seems to affect straight lines in vimeo so I think that's what happened to SAHiN video. And this is highly compressed. I got the hq version of the demo somewhere.
But in SAHiN video, what I notice is that the polys itself look like they're warping especially where GI would sample the hardest. A lot of colors flickering. I'm not sure where the error is. It could be GI or textures dancing.
The straight line artifacts in Vimeo are from not enough anti-aliasing. His videos aren't compressed enough to show artifacts that visible anyway, and Sahin's video, at 11mb for 9 seconds, is compressed far less.

And again, neither aliasing nor compression will create spots that *pop* from bright to dark, much less are artifacts focused in tight corners. Especially compression, because it has *no* idea where tight corners are, it only sees color and contrast. Note that there's no flickering or any artifacts on the high-contrast floor.

I can give you even more examples if necessary, I just don't want to be accused of being a lw hater or vray fanatic. So, before people make assumptions again, I'm not saying LW can't do this without flickering, it can. I'm saying that this example has gi flickers.

toby
06-18-2009, 03:31 PM
I am sure many people here can tell you why some things can't be achieved mate.. It seems Toby already discovered and revealed the secrets of the trade above..Read that and try it :)
I am done in this topic.. I got animations to finish.

Cheers
On a related subject, this is the second or third time someone asked for some techniques from you, and I find it odd that you won't reveal a single setting of your renders, nor give any idea of what your render times (and machine speed) are like. Not even if you used interpolated gi or not (although it became obvious once the video was posted). Do you not like to share your settings?

SAHiN
06-18-2009, 03:34 PM
That's a great way to demonstrate how great lightwave is; denial, and accuse people of having ego problems. Sorry if I bruised yours, but if something looks like gi flickering, I'm not going to act like a fan-boy and make excuses for it.


The straight line artifacts in Vimeo are from not enough anti-aliasing. His videos aren't compressed enough to show artifacts that visible anyway, and Sahin's video, at 11mb for 9 seconds, is compressed far less.

And again, neither aliasing nor compression will create spots that *pop* from bright to dark, much less are artifacts focused in tight corners. Especially compression, because it has *no* idea where tight corners are, it only sees color and contrast. Note that there's no flickering or any artifacts on the high-contrast floor.

I can give you even more examples if necessary, I just don't want to be accused of being a lw hater or vray fanatic. So, before people make assumptions again, I'm not saying LW can't do this without flickering, it can. I'm saying that this example has gi flickers.


First I had no idea we were obligated to defend LW here.. You must have confused me to NT marketing department..

Secondly, you can't say u like VRAY cause Vray flickers like hell..I've been there and done that.
We showed you ours, how about you show as yours for a change so we can talk over some solid work..?

..and don't go talking **** untill ou show me a work you've done yourself on VRAY..
I am done playing devil's advocate with you.

SAHiN
06-18-2009, 03:39 PM
On a related subject, this is the second or third time someone asked for some techniques from you, and I find it odd that you won't reveal a single setting of your renders, nor give any idea of what your render times (and machine speed) are like. Not even if you used interpolated gi or not (although it became obvious once the video was posted). Do you not like to share your settings?

LoL
Why do you need my settings..? You seem to know it all..?
Many people send me a message and ask me, and I answer them..

Many times I have said the same thing..each scene is different, and requires different setting, and since I am not the head of KURV studios or Desktop images all I can do is talk over a scene someone having problem with..

But guys like you don't want to try things, or find solutions - you want a replicator technology that responds to your voice commands ..and Iam no replicator..
You first gotta learn what being animator is all about.

toby
06-18-2009, 04:06 PM
First I had no idea we were obligated to defend LW here.. You must have confused me to NT marketing department..

Secondly, you can't say u like VRAY cause Vray flickers like hell..I've been there and done that.
We showed you ours, how about you show as yours for a change so we can talk over some solid work..?

..and don't go talking **** untill ou show me a work you've done yourself on VRAY..
I am done playing devil's advocate with you.
WHERE was I talking S***??? You're the one that's been argumentative - instead of presenting any evidence or logic as to why those flickers in your video are not gi flickers.

No one here has said that vray flickers, so I saw no need to show vray renders - and you would just accuse me of being egotistical for showing my work anyway!! I have to uncompress all the frames, this will take a little while.

SAHiN
06-18-2009, 04:10 PM
WHERE was I talking S***??? You're the one that's been argumentative - instead of presenting any evidence or logic as to why those flickers in your video are not gi flickers.

No one here has said that vray flickers, so I saw no need to show vray renders - and you would just accuse me of being egotistical for showing my work anyway!! I have to uncompress all the frames, this will take a little while.

So to put it in plain English, you have NO WORK to show us ?
It takes one to know one, instead of proving your point you want to decompress my work ?
LoL
Pathetic

toby
06-18-2009, 04:21 PM
So to put it in plain English, you have NO WORK to show us ?
It takes one to know one, instead of proving your point you want to decompress my work ?
LoL
Pathetic
I'm uncompressing MY FRAMES of MY WORK. I don't keep 3 year old projects unarchived, do you???

JEEZ wtf is wrong with you???

SAHiN
06-18-2009, 04:26 PM
I'm uncompressing MY FRAMES of MY WORK. I don't keep 3 year old projects unarchived, do you???

JEEZ wtf is wrong with you???

Can't wait to see them :D

BTW last project you did was 3 years ago ? :hey:

SAHiN
06-18-2009, 05:12 PM
SAHiN, sent you a PM.:phone_cal

LoL I saw it and replied :phone_cal

toby
06-18-2009, 05:33 PM
Can't wait to see them :D

BTW last project you did was 3 years ago ? :hey:
Yes, my work is *so bad* that no one has hired me for 3 years.

Is that what you're implying? Or that Vray has actually gotten *worse* in the last 3 years??

SAHiN
06-18-2009, 05:58 PM
Toby, untill I see some of your work, we are done !!

Dave, you should clean up your mailbox :D

toby
06-18-2009, 06:43 PM
Toby, give it a frakkin' rest, PLEASE?

SAHiN told us the animation did NOT have flicker. You essentially called him a liar. If I were SAHiN, I would be pissed too.

Again, give it a rest and go do something else before this VERY HELPFUL thread is locked.
I pointed out something he did/does not see. That is hardly calling him a liar. In return he immediately called me egotistical, and he's been on my a-s-s ever since, despite the fact that I've restrained myself from adopting a similar attitude. In my last post I responded to his unprovoked comment by *asking* what he was implying, and said NO MORE. Please re-read our posts. If this thread gets locked, it's not because of me.

Having a really hard time understanding why saying something looks exactly like gi flickering is so inexcusable.

Matt
06-18-2009, 07:23 PM
I gotta say, it does look like GI flickering under the shelf on the desks to me, but I'll take your word for it that the full resolution version does not have that there.

Just saying in the example posted it does *appear* to have GI flickering.

What were your settings for that animation? And did you use animated caching?

toby
06-18-2009, 07:46 PM
I beg your pardon? You assume that he is not professional enough to know that his video has GI flicker or not?
It's quite possible to be professional but not see everything in an image that someone else does. That's why there's cg supervisors, and that doesn't mean the artists that get notes or corrections are unprofessional or inexperienced.


You seem to have a chip on your shoulder about LW not being up to snuff and a grand conspiracy that LW isn't capable. And then someone shows up and says that it DOES work well enough and you say it doesn't by calling him a liar and that it IS GI flicker.


My god man. My saying this flicker is from gi isn't what this is about, is it. It's about the LAST argument, where I was pissed that LW doesn't support interpolation with deforming geo. Just because people freaked out that I had the gaul to be upset with LW, which happens to ALL of us from time to time, that means I think it's not capable??

like I said,

So, before people make assumptions again, I'm not saying LW can't do this without flickering, it can. I'm saying that this example has gi flickers.
Only an idiot would think lw can't do that scene without flickering. It just needs some tweaks and maybe some more motion blur.

toby
06-18-2009, 07:54 PM
Oops, I'd already uploaded my reel here:
http://rapidshare.com/files/241744239/Reel09_tobygaines.mov.html
The black car and the Halo shots (just the main character and rifle) are vray. Only the background behind the car was cached.
The speed racer shots are also rendered with final gather, in mental ray, and we were fighting gi flicker for 6 months, because it was a (very) beta version.

geo_n
06-18-2009, 10:33 PM
Megalodon - If you want to make something similar to SAHiN video then imho the fastest way for something similar is too use static cache for background. The building is not moving. Then for the humans moving just use arealights with backdrop radiosity and render separately. gerardstrada explained it very well. It is common workflow for vfx.
Since the thread started no one has posted any render with AC cache yet that is acceptable quality. Imho its not possible or its too troublesome to use AC cache in one pass pure gi with no light trick.

geo_n
06-18-2009, 10:39 PM
Oops, I'd already uploaded my reel here:
http://rapidshare.com/files/241744239/Reel09_tobygaines.mov.html
The black car and the Halo shots (just the main character and rifle) are vray. Only the background behind the car was cached.
The speed racer shots are also rendered with final gather, in mental ray, and we were fighting gi flicker for 6 months, because it was a (very) beta version.

Excellent reel. Very professional works. Are you using vray prepass on the shots? Also I'm curious why you used mray for speedracer instead of vray or finalrender? I haven't had much success with mray gi and only use AO and lightrigs. Its fast though.

toby
06-19-2009, 12:06 AM
Excellent reel. Very professional works. Are you using vray prepass on the shots? Also I'm curious why you used mray for speedracer instead of vray or finalrender? I haven't had much success with mray gi and only use AO and lightrigs. Its fast though.
Thanks!
It's been a while, I don't remember much about the prepass; Light cache is one of them, right? And there was monte carlo, and photons? I never used the light cache one, it didn't look good on it's own, it was like plain ambient light. I think I used the qmc one instead.

Digital Domain only uses renderman or mental ray for movies, and they're mostly Linux anyway. Mental ray was chosen for racer because it's better at raytracing, which is a must for the best looking cars, and the vfx supervisor prefers mental ray. But this was stand-alone mr, it's a bit different, and results will be different from studio to studio because you customize so much of it. It is really fast, but motion blur... :P

geo_n
06-19-2009, 09:55 AM
Thanks!
It's been a while, I don't remember much about the prepass; Light cache is one of them, right? And there was monte carlo, and photons? I never used the light cache one, it didn't look good on it's own, it was like plain ambient light. I think I used the qmc one instead.

Digital Domain only uses renderman or mental ray for movies, and they're mostly Linux anyway. Mental ray was chosen for racer because it's better at raytracing, which is a must for the best looking cars, and the vfx supervisor prefers mental ray. But this was stand-alone mr, it's a bit different, and results will be different from studio to studio because you customize so much of it. It is really fast, but motion blur... :P

Prepass is new in irridiance. I think you used irr/qmc if it was 3 years ago. Don't know wasn't doing this stuff back then:D.
lightache is good for walkthroughs and getting unbiased renderer mode in vray lightcache/lightcache setting.
mray, it is so fast with raytracing even the max ver. It is getting simpler to use while lw is getting more complicated.
I'm trying to learn mray after seeing so many good resources
http://mentalraytips.blogspot.com/star wars from Blur studios :thumbsup:
, but you got to love lw for the price less 1000US before and you can get excellent render engine with modeller/layout.

Btw I really liked speedracer movie. I wanted to buy a collectors dvd with a die cast car included. Maybe the west hated it. Its a japanese manga after all.
This is not speedracer quality but one very talented lw veteran and myself are making cutscenes for game
http://rapidshare.com/files/246321068/cg_chaser.mov.html
raw render, no post effects/particles. Not bad for default lw renderer. Flashy and shiny like speedracer:D:D

toby
06-19-2009, 01:17 PM
Prepass is new in irridiance. I think you used irr/qmc if it was 3 years ago. Don't know wasn't doing this stuff back then:D.
lightache is good for walkthroughs and getting unbiased renderer mode in vray lightcache/lightcache setting.
mray, it is so fast with raytracing even the max ver. It is getting simpler to use while lw is getting more complicated.
I'm trying to learn mray after seeing so many good resources
http://mentalraytips.blogspot.com/star wars from Blur studios :thumbsup:
, but you got to love lw for the price less 1000US before and you can get excellent render engine with modeller/layout.

Btw I really liked speedracer movie. I wanted to buy a collectors dvd with a die cast car included. Maybe the west hated it. Its a japanese manga after all.
This is not speedracer quality but one very talented lw veteran and myself are making cutscenes for game
http://rapidshare.com/files/246321068/cg_chaser.mov.html
raw render, no post effects/particles. Not bad for default lw renderer. Flashy and shiny like speedracer:D:D
Sweeet!
Isn't it great to do stuff like this and get paid for it??

geo_n
06-19-2009, 08:43 PM
Sweeet!
Isn't it great to do stuff like this and get paid for it??

Yes but I would like to do more your stuff. :p
Ours is with small budget.:D

Here's one with post effects. Compression looks bad.
http://revver.com/video/1632462/cg-carchase01/
What codec did you use for your qt video? It looks better than h.264.
Animation codec?


Lw motion blur is good and fast. :D Have to try mray
http://rapidshare.com/files/246495872/cg_chaser2.mov.html

toby
06-19-2009, 09:41 PM
Here's one with post effects. Compression looks bad.
http://revver.com/video/1632462/cg-carchase01/
What codec did you use for your qt video? It looks better than h.264.
Animation codec?
I think it's just more compressed than mine. Is that your compression or did the website do it? A higher keyframe value could allow you to turn up the quality without creating a larger file.
I used mp4, h264 just looks a tiny bit better or makes a little bit smaller file, but takes a long time to encode. If I don't feel like waiting, I use mp4.

geo_n
06-19-2009, 09:55 PM
I think it's just more compressed than mine. Is that your compression or did the website do it? A higher keyframe value could allow you to turn up the quality without creating a larger file.
I used mp4, h264 just looks a tiny bit better or makes a little bit smaller file, but takes a long time to encode. If I don't feel like waiting, I use mp4.

The website did it. The rapidshare link below that post is h.264. Its not as clear as your video though:D. hmmm

Anyway maybe SAHiN can post h.264 video also. Both he and I probably don't know how to post video on net. Maybe compression added to his videos flickering because its not gi according to him. Just wierd about the polys looking like blinking polys. I dont see that often. I see it in modelling error also with overlapping geometry.

toby
06-20-2009, 12:29 AM
Your compression doesn't look bad but the file size is a bit big. Try using 240 keyframes, multipass and frame re-ordering on, and quality at 75%.

Other things that make a difference are
1. how sharp your renders are; straight cg is considered too sharp, so most the shots I do get a litle bit of blur, .25 to .5 of a pixel. Softer video shows less compression artifacts.

2. Highly saturated colors, especially red, show artifacting very easily. I've never seen a compression artifact in Pixar movies (they take it very seriously), except for the bright red suits in the Incredibles. It's an artifact created by a compression method used in almost all compression types, and it's what causes pixelation. That is showing up in the letters and at the lower edge of the window when the car is sideways. If you desaturate that texture a little bit that should get better, and it might look a little more realistic too. But don't worry too much about it, if Pixar can't get rid of the artifacts in red then we can't be expected to!

geo_n
06-20-2009, 02:44 AM
Try using 240 keyframes, multipass and frame re-ordering on, and quality at 75%.


Where is that? quicktime pro?

toby
06-20-2009, 03:01 AM
Where is that? quicktime pro?
Yea, I was hoping it would be similar in whatever you're using... if you're on a mac you can use quicktime amateur instead, it's free

erikals
08-03-2009, 05:10 PM
toby, i agree that LW GI produces flicker with certain settings, but i posted a non-flicker solution in that other thread. So, we agreed on that, we made that scene work.

but like i said earlier, if you can get hold of a scene that produces flicker i will try to fix it.
no promises, but i will give it a go.
so, gimme a scene http://forums.cgsociety.org/images/smilies/smile.gif

toby
08-03-2009, 05:21 PM
toby, i agree that LW GI produces flicker with certain settings, but i posted a non-flicker solution in that other thread. So, we agreed on that, we made that scene work.

but like i said earlier, if you can get hold of a scene that produces flicker i will try to fix it.
no promises, but i will give it a go.
so, gimme a scene http://forums.cgsociety.org/images/smilies/smile.gif
I don't have any scenes that still flicker, not sure what you're talking about - ?

erikals
08-03-2009, 06:08 PM
sorry, my mistake, was going back and forth reading threads, i mixed up in the end. heh, never a smart thing to do... http://forums.cgsociety.org/images/smilies/smile.gif

it was kind of hard getting to a conclusion to solutions in this thread, but my conclusion as of now is,
-LW GI produces no flicker with the correct settings, but avoid AC, as GI with the right settings renders faster.
(someone share a .mov render and scene if they disagree)
-LW GI cannot compare to Vray for realistic renders. esp. architectural.
(though must say i have seen some close matches at times, anyone have a Vray scene to share?)

allabulle
08-03-2009, 09:52 PM
sorry, my mistake, was going back and forth reading threads, i mixed up in the end. heh, never a smart thing to do... http://forums.cgsociety.org/images/smilies/smile.gif

it was kind of hard getting to a conclusion to solutions in this thread, but my conclusion as of now is,
-LW GI produces no flicker with the correct settings, but avoid AC, as GI with the right settings renders faster.
(someone share a .mov render and scene if they disagree)
-LW GI cannot compare to Vray for realistic renders. esp. architectural.
(though must say i have seen some close matches at times, anyone have a Vray scene to share?)

I guess the ability that Vray has to simulate "linear colour space output and tone mapping" makes it hard to match with plain LW native render output. A good use of colours and post tone mapping could say another thing about the comparisons face to face of a render "from" LightWave to another "from Vray". Sure you don't seem to think too much of that if you are using Vray, though. It has its price tag, that easyness (I think Vray alone is more expensive than LightWave).

I mean, it has its ups and downs, but if you are willing to compare sure you can. But with a little thought on it, that's it. Or, a lot, it depends. :--)

I hope no one takes my statements here badly.

geo_n
08-03-2009, 10:28 PM
toby, i agree that LW GI produces flicker with certain settings, but i posted a non-flicker solution in that other thread. So, we agreed on that, we made that scene work.

but like i said earlier, if you can get hold of a scene that produces flicker i will try to fix it.
no promises, but i will give it a go.
so, gimme a scene http://forums.cgsociety.org/images/smilies/smile.gif

I can post one :). An interior scene which flickers like hell :devil:. I'll just have to remove some models which I can't share.

erikals
08-03-2009, 10:32 PM
I can post one :). An interior scene which flickers like hell :devil:. I'll just have to remove some models which I can't share.

please do, ready for a challenge :)
include textures if possible...

jameswillmott
08-03-2009, 10:40 PM
I can post one :). An interior scene which flickers like hell :devil:. I'll just have to remove some models which I can't share.

If only the camera moves in your scene and it flickers, then your cache isn't set up right. That's about the only thing that could go wrong...

erikals
08-03-2009, 10:41 PM
that said, i'm also searching for some Vray scenes and renderes, to see if i can match Vray in LW.

...so if anyone has a Vray scene and render...

geo_n
08-03-2009, 10:45 PM
If only the camera moves in your scene and it flickers, then your cache isn't set up right. That's about the only thing that could go wrong...

Never had problem with "camera moving" only animation. My problem is always when everything moves :D.
The scene I have is fixed camera with moving objects. I'd like to try both camera and objects moving though. :D

JonW
08-04-2009, 05:55 AM
As Jameswillmott said,

If nothing in the scene moves or changes & its only the camera that is moving. You don’t use Animated Cache.

You can use as many computers as you want to render frames in any order you desire & they will join perfectly. Pixel peep the same frame from two different computers in Photoshop & do the top layer as Difference, everything will show up black.

Only move things if you really have to. your render times will multiply.

geo_n
08-04-2009, 06:12 AM
here is file. destroy it. :devil: :twak:
If single pass radiosity render works in lw I'm happy :D
I moved camera but original file has fix camera.

erikals
08-04-2009, 07:19 AM
hm, tricky... :/ ...

erikals
08-04-2009, 07:57 AM
this one is close, but not quite there, the solution is to adjust...
RPE (increase for higher quality)
SBR (increase for higher quality)
MPS (decrease for higher quality)

this scene is a bit slow as far as rendertime goes though, 6m30s on my Quad2.6GHz...

so, the trick here is to find a good balance between the 3 options above, RPE-SBR-MPS

oh, and always use limited region when rendering tests,
in this case i first found an area that had much flicker, then just selected that area and tested misc settings...

please feel free to test...

erikals
08-04-2009, 03:53 PM
just a note while testing, it is very important that the model is clean.
found two mistakes here in the area where the circle is,
-nonplanar polys
-overlapping polys

this often leads to rendererrors,...

...still testing though, found it hard to make this one work...
testing a new lighting setup, to see if that helps...

geo_n
08-04-2009, 06:46 PM
just a note while testing, it is very important that the model is clean.
found two mistakes here in the area where the circle is,
-nonplanar polys
-overlapping polys

this often leads to rendererrors,...

...still testing though, found it hard to make this one work...
testing a new lighting setup, to see if that helps...

Didn't notice that;D. I converted it from max vray. It looks very dim. No more flicker on the ceiling?:help:

erikals
08-04-2009, 09:46 PM
tested a bit,... a lot actually... :0

decided to start somewhat from scratch, using a simple light setup, to see just what was going on...

here's what i found,

templightwave - FG for animated objects,... conclusions so far... (http://www.youtube.com/watch?v=-ZbX-PZCNbM)

...so, i think one have to expect very long rendertimes for these types of scenes...

geo_n
08-04-2009, 10:55 PM
rendered your scene settings. 12min per frame on qc6600. lw interior scene with moving stuff really long rendertime I think.

erikals
08-05-2009, 12:08 AM
rendered your scene settings. 12min per frame on qc6600. lw interior scene with moving stuff really long rendertime I think.

unfortunatly interior renderings take lots of time, and if it is animation involved, it takes 4-ever....
however i don't think LW is bad when it comes to rendertimes,... that's what i've heard at least...

testing a scene now but the rendertime on it is huuge,...
however, it is almost flickerfree, i think.. i hope... :0

i opened the renders you did and made a .mov
as you can see it unfortunatly has a lot of flicker...
i find it somewhat strange though... it only flickers where the highlights are... hmm, anyone knows why?

attached .mov below...

geo_n
08-05-2009, 12:29 AM
yep lots of flicker. wanted to test out your setting on the actual scene not the proxy but it flickers and render time is long with low quality in the end.
Looking forward to your next test with the scene. :thumbsup:

erikals
08-05-2009, 12:51 AM
so, anyone have one of these to give away?....
http://www-03.ibm.com/press/img/22579.jpg

that said, what's the fastest/ cheapest processor out there now?

erikals
08-05-2009, 01:24 AM
ok, flickerfree,... i think...
very high rendertime though...

will try to decrease it...

erikals
08-05-2009, 01:48 AM
tip, in the scene above, decrease RPE to 3000.
this makes rendertime 6min instead of 25min (on my PC that is...)

also note the high value on the "Secondary Bounce Rays" setting, which looks to be quite unique for interior renders.

JonW
08-05-2009, 06:38 PM
you canít have cheap & fast........ but a 920 would be a good compromise. The i7s are very quick with radiosity. If you donít need 12gb of ram there are cheaper motherboards. Unless you are doing massive scenes (5 - 20+ million polys) any old graphics card is fine. Iíve got a GTX280 down to a 8600 & the 8600 is perfectly ok, but if you can go up a notch it will make handling larger scenes more pleasurable & add very little cost.

At the end of the day. Buy as much CPU grunt as you can afford. Then have enough ram to cover the work you typically do.

Graphics, HDs & everything else doesnít really count. (For rendering LW doesnít use graphic cards)

keep your old computer & set up Screamernet. Its a bugger to set up but if you take your time & get away from it for a few days here & there its well worth the effort, if you have a few old computers acting as door stops, use the extra rendering power, a bit here & a bit there adds up surprising quickly at no real cost other than your time to set up SN. (Once you have SN nodes/boxes set up & you use Remote Desktop, take the graphics cards out as you will want to cut down your electricity where ever you can, they are just not needed). Also just get the cheapest OS, XP64 is ideal, I have put this on my new W5580 box.



tmulti scene.....

Rendered in a bit over 2 min on my W5580 (So about 5 minutes on a 920) with those extremely high RPB & SBR figures.

You must use Monte Carlo for moving objects. I checked the objects with LWcad, there were 3 self intersecting polys. The right angle wall had sides missing & the robot had intersecting polys. AC hates all these things. One needs to build a scene as if it were real objects if using AC.

I tried a few set ups & I feel the most economical way for this scene is to Bake it. Then lock the cache.

Bake Radiosity (about 5 minutes on W5580, so about 10 on a 920)
Intensity 200% (I upped the % so I could see what was going on)
IB 3
RPE 1000
SBR 1000 (keep the same as RPE)
AT 1.0
MinPS 1.0 (this needs to be maximum of 1.0 sometimes around 0.75 helps with small details)
MaxPS 30
Multiplier 100% (This needs to be 100%)

Camera Properties
Only use AA, donít use AS. Using AA I find its often quicker any way. I also set it at 5 for this.

On W5580, E5450 & 5335 I am getting 29 frames an hour. So if you had say 3 x 920 CPUs (you can get a 3 pack of XP64 to save money) you would get around 21 frames per hour.

The Zip is the first 200 frames .mov, the artefacts are from compression.

JonW
08-05-2009, 07:46 PM
I have just tried with RPE & SRB, both at 512 & its pretty good. I’m spitting out 92 frames an hour.

RPE & SRB both at 216 there is a little bit of flicker. I’m spitting out 234 frames an hour.

I would say you probably need to use 512 when all the detail is in the scene, maybe you can get away with 343 on both.

toby
08-05-2009, 08:34 PM
I tried a few set ups & I feel the most economical way for this scene is to Bake it.
Agreed, I tried fg + motion blur settings that were taking longer than 11 min. for a small limited region, and still flickered like mad. I think it got worse when I turned down the light quality on all those area lights (my machine is very slow), and since the sample pattern is different for the lights every frame that shot the fg to hell. But now with AC I'm getting 2+ minutes for a full, anti-aliased frame and no flicker. I had the shadows turned for the spherical lights too.

geo_n
08-05-2009, 09:12 PM
I forgot where I rendered this. vray or kray :stumped:.

geo_n
08-05-2009, 09:51 PM
you canít have cheap & fast........ but a 920 would be a good compromise. The i7s are very quick with radiosity. If you donít need 12gb of ram there are cheaper motherboards. Unless you are doing massive scenes (5 - 20+ million polys) any old graphics card is fine. Iíve got a GTX280 down to a 8600 & the 8600 is perfectly ok, but if you can go up a notch it will make handling larger scenes more pleasurable & add very little cost.

At the end of the day. Buy as much CPU grunt as you can afford. Then have enough ram to cover the work you typically do.

Graphics, HDs & everything else doesnít really count. (For rendering LW doesnít use graphic cards)

keep your old computer & set up Screamernet. Its a bugger to set up but if you take your time & get away from it for a few days here & there its well worth the effort, if you have a few old computers acting as door stops, use the extra rendering power, a bit here & a bit there adds up surprising quickly at no real cost other than your time to set up SN. (Once you have SN nodes/boxes set up & you use Remote Desktop, take the graphics cards out as you will want to cut down your electricity where ever you can, they are just not needed). Also just get the cheapest OS, XP64 is ideal, I have put this on my new W5580 box.



tmulti scene.....

Rendered in a bit over 2 min on my W5580 (So about 5 minutes on a 920) with those extremely high RPB & SBR figures.

You must use Monte Carlo for moving objects. I checked the objects with LWcad, there were 3 self intersecting polys. The right angle wall had sides missing & the robot had intersecting polys. AC hates all these things. One needs to build a scene as if it were real objects if using AC.

I tried a few set ups & I feel the most economical way for this scene is to Bake it. Then lock the cache.

Bake Radiosity (about 5 minutes on W5580, so about 10 on a 920)
Intensity 200% (I upped the % so I could see what was going on)
IB 3
RPE 1000
SBR 1000 (keep the same as RPE)
AT 1.0
MinPS 1.0 (this needs to be maximum of 1.0 sometimes around 0.75 helps with small details)
MaxPS 30
Multiplier 100% (This needs to be 100%)

Camera Properties
Only use AA, donít use AS. Using AA I find its often quicker any way. I also set it at 5 for this.

On W5580, E5450 & 5335 I am getting 29 frames an hour. So if you had say 3 x 920 CPUs (you can get a 3 pack of XP64 to save money) you would get around 21 frames per hour.

The Zip is the first 200 frames .mov, the artefacts are from compression.

The scene looks very different than the original. Less lights have less flickers indeed. But my original scene had more lights so super flicker.
In the end I rendered this in vray baking the gi for bg and then using anim prepass on moving objects. I couldnt get good results in lw with this kind of enclosed scene. Outdoor is fine with background radiosity. :D

Heres the way to do it in vray. I wish its possible to do volumetric sphere in lw to isolate moving objects easier.

http://www.spot3d.com/vray/help/150SP1/tutorials_anim.htm
http://www.spot3d.com/vray/help/150SP1/tutorials_anim2.htm

toby
08-05-2009, 10:21 PM
I forgot where I rendered this. vray or kray :stumped:.
how fast was that render?
there's some flickering up by the ceiling light btw

geo_n
08-05-2009, 10:53 PM
I think 10-15 min. Its brute force I believe.

JonW
08-05-2009, 11:31 PM
I forgot to mention, I took the reflection of the wall & floor & forgot to put them back after fixing up the polys. But if its baked it shouldn’t add too much time to each frame.

I did most frames at 343 & it looks ok, but at 343 whether its ok will depend on the detail of the robot.

erikals
08-06-2009, 09:19 AM
JonW, nice settings! http://forums.cgsociety.org/images/smilies/smile.gif

one things though, would that scene work with animated characters or deformed objects?

JonW
08-06-2009, 09:43 AM
I don’t know, up till now I have only been putting clip maps of people in my architectural images & even then the type of renders I do most architect don’t want people ruining their temples.

I have literally just started with 3d figures. And then I need to work out how to animate them. Steep learning curve to say the least!

I need to get on top of this field so I am prepared, can’t leave it any longer.

erikals
08-06-2009, 10:32 AM
finding some interesting stuff here, in some cases, using interpolated is actually slower that a normal render... hmm,...

Danner
08-06-2009, 01:32 PM
"you canít have cheap & fast........ but a 920 would be a good compromise. The i7s are very quick with radiosity"

I bought a new intel box with Gigabyte Mobo, 6gb of ram, and a 920 i7 processor, Added a CPU cooler (coolmasterV8) and overclocked it to 3.1ghz. the rendering times are 3.6 times faster than my Core2 duo 2.66ghz

toby
08-06-2009, 02:28 PM
one things though, would that scene work with animated characters or deformed objects?
Yikes I don't even want to think about that...

finding some interesting stuff here, in some cases, using interpolated is actually slower that a normal render... hmm,...
You mean slower than brute force gi? That's exactly what I found before going back to AC :/

JonW
08-06-2009, 05:14 PM
Danner.

Yes, the i7s are quick. The 920 gives you the best Ghz for your dollar & an even better frame out put rate per dollar. Now if you haven’t done so try to get the 2.66 working as a node with Screamernet, no point allowing it to get lazy.

toby
08-06-2009, 10:26 PM
with AC I'm getting 2+ minutes for a full, anti-aliased frame and no flicker. I had the shadows turned for the spherical lights too.
My mistake, looks like a accidentally switched off *Anim* cache or forgot to save it, with it turned on it's taking 45 minutes per frame, and the bake is only 15 sec a frame. Looks like that's all the time you save by baking too. It's too slow to even experiment with on my old boat anchor!

Has anyone tried more bounces? It doesn't slow things down too much and could allow removing some of the lights. I can try one test render per day... :P

erikals
08-06-2009, 10:46 PM
i tried a bunch of bounces with interpolated, and i was very close to removing the flickering, however interpolated seems to have a pitch, beyond that pitch it looks to be impossible to remove any more flickering....

i hope i'm wrong of course.... http://forums.cgsociety.org/images/smilies/smile.gif

testrendering without interpolated now, however i'm bumping into another pitch, trying to remove grain,... for some reason the grain stays in some areas of the scene, it stays even when increasing rays, and for some weird reason, increasing the AA does nothing,...

really weird stuff going on...

(both tests above are based on using the FG method)

EDIT: important note, my aim here is to get deformed animations (character animation) to work inside this scene, as i haven't seen anyone in any thread managed to do that.

a character animation in an outdoor scene would be way easier, you'd just use the "Background only" method, or FG.
FG in an character animation outdoor scene produces way less flicker.

geo_n
08-06-2009, 10:55 PM
are you guys testing the dark new scene? Because adding more lights will give more flicker. So the test on the dark scene will not be valid anymore.

JonW
08-06-2009, 11:04 PM
If you post the scene you want tested I am quiet happy to render it. I am between jobs at the moment. I’ve got my i7s up & running after pinching the heatsinks.

geo_n
08-06-2009, 11:57 PM
If you post the scene you want tested I am quiet happy to render it. I am between jobs at the moment. Iíve got my i7s up & running after pinching the heatsinks.

http://www.newtek.com/forums/showpost.php?p=911276&postcount=198
that's the original file with the lighting. Incase the robot is giving trouble just delete some parts. Anyway some parts on the base is moving so there's still some movement as well as camera motion.

erikals
08-07-2009, 01:13 AM
ok, concluding rendering character animation and interior.
i came to the conclusion that JonW's method on the previews page is the way to go.

the idea is to first render the scene with the character, but make the character invisible for camera, only making the character's shadow renderable. (+interior)

the second step is to go the other way around, render the character only, but without baking, so the deformations on the character render correct.

import the two animations into e.g. After Effects, for then to merge the renders in AE.

erikals
08-07-2009, 01:25 AM
are you guys testing the dark new scene? Because adding more lights will give more flicker. So the test on the dark scene will not be valid anymore.

it will be very slow to render, but try this,...

-Intensity (leave as is)
-IB (leave as is)
-RPE (you will have to increase this a lot)
-SBR (you will have to increase this a lot)
-AT (should always be 1 for animation)
-MPS (should always be 1)
-MPS (start at value of 10, increase if render looks ok)
-Multiplier (should always be 100 for animation)

of course, all this above varies depending on the scene, and how the different settings interact/influence each other, but those are the basics rules i found.

unfortunatly i think the render is going to be very slow, as more lights equals more rendertime.

the first thing i think you have to do is increasing the RPE and the SBR. wouldn't surprise me if you have to set these to as high as 4000.

hopefully JonW can help, with his Lighningfast Xeon... http://forums.cgsociety.org/images/smilies/aarggggghhhh.gif

JonW
08-07-2009, 11:25 AM
Its only twice as quick as a mildly overclocked 920. Radiosity is where these CPUs (i7s & x55xx) are extra fast, my 940 @ 3.5Ghz does Radiosity about 30 - 50% quicker than my 5450 V8.

I was just about to do a straight Monte Carlo. I have a 19 million poly scene which give me less trouble than this.


Trail & tribulations! (notes of set ups, & basically all no good)

First did the file as is & changed nothing - lots of flicker.
2. MC + Cache + Animation, RPE & SBR 125 - no good
3. MC + Cache + Animation, RPE & SBR 216 - no good
4. Fixed Polys
5. MC + Cache + Animation, RPE & SBR 343 - no good
Baked 120, 2, 343, 343, 4.0, 1.0, 100, 100, every frame. (19.3 mb file about 40 minuets on 5580)
6. MC + Cache + Animation + Locked - still no good
7. 120%, 2, 343, 27, 1.0, 1, 10, 100 gi 4:06/ et4:37 on 5580
8. 120%, 2, 512, 216, 1.0, 1, 10, 100 gi 8:??/ et10:57 on 5580

8. Monte Carlo only 120%, 2, 125 - 32:35
Interpolated off
9. FG 120, 2, 64, 64, 1.0, 1, 10, 100 - 42.8/6:40 AA 2/AS 0.1
10. FG 120, 2, 64, 64, 1.0, 1, 10, 100 - 42.8/15:39 AA5 /AS off
11. FG 120, 2, 216, 16, 1.0, 1, 10, 100 - 17.8/-
12. FG 120, 2, 125, 16, 1.0, 1, 10, 100 - 15.7/22:09
13. FG 120, 2, 125, 16, 1.0, 1, 10, 100 - 14.9/

Getting better!

Final Gather
Interpolated OFF
Specular off for all lights (drops time by about a third)

14. FG 2, 125, 16, 10, 1.0, 2.0, 100 - 1:50/6:41
15. FG 2, 216, 16, 20, 1.0, 2.0, 100 - 2:06/5:22
16. FG 2, 343, 16, 20, 1.0, 2.0, 100 - 2:35/7:23
AA 3
17. FG 2, 512, 16, 30, 1.0, 2.0, 100 - 3:01/10:44
AA 5 AS off
18. FG 2, 512, 16, 40, 1.0, 2.0, 100 - 3:59/9:45 (13:30)


Conclusion.

Final Gather
Interpolated ďOFFĒ (its the Interpolation which is stuffing everything up)
Specular off for all lights (drops render time by about a third)

AA 5 & AS off, or
AA 3 & AS 0.1 + Soft Filter (for softer image & about 20% faster render)

Intensity 120%
IB 2
RPE - 512 (lower than this & there is too much noise)
SBR - 16
AT - 40 (High because there are so many area lights)
MPS - 1
MPS - 2 (Yes it is ďTwoĒ)
Multiplier - 100

gi 2:58/ et 13:32 on W5580 (920 about double)


Final notes.

When you lift the RPE render times get too long so if you raise AT it reduces the time. If PRE is too low there too much noise. The idea behind MPS being 2 is, the scene is doing a sort of an interpolation & it isnít while getting some of the render done in radiosity which brings back the time in the rendering stage.

I tried quite a few renders massaging the numbers, Quality & time falls apart very quickly.

Iíve got 10 frames so far & its looking ok, RPE could go up a bit more, & AT to manage render time. Iíll post a short .mov in the morning.

Frame 1 & 2 attached.

toby
08-07-2009, 01:21 PM
Why was AC no good?

JonW
08-07-2009, 06:42 PM
Everything I did with AC has flicker & was completely unacceptable, donít know why. The wall to the right still has flicker but its a lot better than before. I will do a few frames at a higher setting around frame 21 where there is a problem.

The polys in the scene really need to be made nice & clean with no overlapping objects either. There are some surfaces that are reflective that arenít really seen that much & donít warrant been reflective. The under side of the objects & some of the room fixtures etc.

If one is really careful & gets rid of every reflective poly that isnít absolutely necessary it will help a lot with the quality & as a reult the render time. At the very least keep it to 30 minuets on a 920, its not brilliant but at least you get 48 frames a day.


Zip 1.2 mb .mov first 94 frames

geo_n
08-08-2009, 11:36 AM
Can you post the scene file? I want to test how the rendertimes is like on my box. Lol!

toby
08-08-2009, 01:56 PM
Ok geo_n, we're ready, tell us how long the Vray and Kray renders take -

geo_n
08-08-2009, 08:38 PM
Ok geo_n, we're ready, tell us how long the Vray and Kray renders take -

I think it was 20-25 minutes but using different gi for both. kray was interpolated brute force then vray was rendered bg cache first then moving objects anim prepass semi bruteforce. The good thing with both renderers is that AA was good and lesser noise than lw renderer.

JonW
08-08-2009, 10:40 PM
The scene I have changed since posting the above. I fixed all the polys, aligned, welded etc, & changed reflective polys on the robot & the room where appropriate. The were a lot of dodge polys & various objects just overlapping, by only 100ths of a ďumĒ.

I did a test at RPE 1000 & it was introducing light leaks. 729 would be about the highest.

Its the internal wall that's the major problem. I would duplicate it & make one of them seen by camera but not seen by radiosiy. The second wall would be seen by radiosity & not seen by the camera, & for the one seen by radiosity make it matt, tun off all reflections etc. I do the same thing for grass textures around buildings & it fixes up a lot of problems.

Zip is scene with fixed polys but needs a duplicate matt wall for radiosity

toby
08-08-2009, 11:07 PM
I think it was 20-25 minutes but using different gi for both. kray was interpolated brute force then vray was rendered bg cache first then moving objects anim prepass semi bruteforce. The good thing with both renderers is that AA was good and lesser noise than lw renderer.
Wow so it wasn't so easy for the other renderers either. I thought they were going to embarrass lw on this one :p

geo_n
08-09-2009, 04:06 AM
Wow so it wasn't so easy for the other renderers either. I thought they were going to embarrass lw on this one :p

But I didn't have to do any major fixing or numerous settings to tweak like Jonw to get almost there with the render I posted with the ceiling slightly flickering as you noted. That was kray render I believe. :D
The vray workflow is easiest and most typical. http://www.spot3d.com/vray/help/150SP1/tutorials_anim.htm
Something in lw's AA and noise is too much jaggies in Jonw tests even with high setting? I will test how long it took to render Jonw's file.

Exception
08-09-2009, 05:30 AM
testrendering without interpolated now, however i'm bumping into another pitch, trying to remove grain,... for some reason the grain stays in some areas of the scene, it stays even when increasing rays, and for some weird reason, increasing the AA does nothing,...

really weird stuff going on...

(both tests above are based on using the FG method)


The FG method is your problem here. FG is ALWAYS interpolated, even when interpolation is off. When interpolation is off all secondary bounces are still interpolated, so you'll still have a chance of splotches and/or flicker/noise whatever.

If you still have AA issues even with high settings you should investigate wether or not they're HDr edges. Every render engine has that issue, and you solve it by rendering with 'Limit Dynamic Range'.

Exception
08-09-2009, 05:45 AM
I do have to say, this scene is just crazy and asking for trouble. Who'd put five huge spherical lights combined with six area lights in a non-textured box and uses 2 bounce radiosity, with objects moving through the spherical lights?
You might as well be off not using GI at all.

But, still a fun test :) Of course it renders way too slow for it to be a handy test.

Mind you, Noise Reduction is turned on in the LW scene file... that's also asking for trouble. It's a dirty 2D post process. Better avoid it.

geo_n
08-09-2009, 06:31 AM
That would be me. :P
Starting from scratch using AC would be welcome. :D

toby
08-09-2009, 02:45 PM
But I didn't have to do any major fixing or numerous settings to tweak like Jonw to get almost there with the render I posted with the ceiling slightly flickering as you noted. That was kray render I believe. :D The vray workflow is easiest and most typical.

Yea, just like the last scene we all hammered on. I wasn't thinking that lw did just as well here, just thought that the other renderers might have no problem at all.


Something in lw's AA and noise is too much jaggies in Jonw tests even with high setting? I will test how long it took to render Jonw's file.
The jaggies show up mostly because the white value is 400% on the ceiling luminosity. If you have the image in the Image Editor you can set the Exposure/White point to 4.0 and the jaggies will be gone. This is where the Limit Dynamic Range button comes in handy, but you lose any value above 100%.
The noise on the far left wall is similar; an area light is right next to the wall and is set to inverse squared falloff, that gives it a really intense value up close. In my scene I pulled it away from the wall.

toby
08-09-2009, 03:03 PM
I do have to say, this scene is just crazy and asking for trouble. Who'd put five huge spherical lights combined with six area lights in a non-textured box and uses 2 bounce radiosity, with objects moving through the spherical lights?

I was wondering about this geo_n, did you have a similar setup in the other renderers? Are all the lights there just to brighten it up? Attached is my render with 8 bounce and no lights (but I did expose it up, just like you would with a camera ), I think it looks more realistic, and not having to deal with noisy sampling would make it a lot easier to render. Using 'lumigons' instead of lights is another option.


Mind you, Noise Reduction is turned on in the LW scene file... that's also asking for trouble. It's a dirty 2D post process. Better avoid it.
Totally agreed. I switched that off as soon as I noticed it, it sucks power just to make your render suck.

Exception
08-10-2009, 04:04 AM
I think the flicker in Animated Cache and other methods used to get this thing to render flicker free is largely due to the large self-illuminated polygon. I'm running some tests to see if this can be fixed easily.
Kray would handle this much better because it has a little system built in that converts bright polygons to lights. Lw could do with such a system.

erikals
08-10-2009, 09:04 AM
um, ok, for some reason I got the AA to work, not sure what happened there,...
the AA quality isn't all that though,.. see picture number 3.
it basically took 40 minutes http://forums.cgsociety.org/images/smilies/aarggggghhhh.gif to render that small limited region, something seems wrong...
AS didn't really help, so turned it off, might help for motion blur though

http://www.newtek.com/forums/attachment.php?attachmentid=76137&stc=1&d=1250002930

geo_n
08-10-2009, 11:34 AM
I was wondering about this geo_n, did you have a similar setup in the other renderers? Are all the lights there just to brighten it up? Attached is my render with 8 bounce and no lights (but I did expose it up, just like you would with a camera ), I think it looks more realistic, and not having to deal with noisy sampling would make it a lot easier to render. Using 'lumigons' instead of lights is another option.

Totally agreed. I switched that off as soon as I noticed it, it sucks power just to make your render suck.

I dont think I did. With vray I just use exposure in camera and tone mapping. But the lights on the side were there because client wanted illumination there even if there's no light fixture. Even that ceiling light they wanted overbright. :D Anyway the lw scene is a quicktest if I can render the same look from vray with fast render time.

erikals
08-10-2009, 12:25 PM
noticed, there is a illuminated poly in the sealing that is missing in the LW render...

erikals
08-10-2009, 12:42 PM
ok, did some tests, here again with no baking/cache of any sort,...
cpu, Quad 2.6GHz

test 1, Monte Carlo, rendertime for limited region, 11min (yes, it's high)
good, but still has trouble with getting rid of noise
(increasing/decreasing values did not help, if you know of a method here, please share...)
http://www.newtek.com/forums/attachment.php?attachmentid=76147&stc=1&d=1250015548
http://www.newtek.com/forums/attachment.php?attachmentid=76148&stc=1&d=1250015548


test 2, Monte Carlo interpolated, rendertime for limited region, 11min (yes, it's high)
pretty good quality actually, note though, fine lines will flicker if MaxPixSpc is 1.5 or more, for other scenes with less fine lines you can increase this and save lots of rendertime.
http://www.newtek.com/forums/attachment.php?attachmentid=76149&stc=1&d=1250015548
http://www.newtek.com/forums/attachment.php?attachmentid=76150&stc=1&d=1250015548


test 3, not shown here, FG interpolated, same rendertime and quality as MC interpolated.
(strange, thought it would be faster, might be just a coincidence..)


.

erikals
08-10-2009, 12:50 PM
attached scenes, MC and MCi
so, these scenes should have no problem showing a character animation,
the only thing you might want to try here, if you are going for motion blur, is to turning on adaptive sampling and set it to something like 0.5-0.8 or so...

toby
08-10-2009, 01:21 PM
I think the flicker in Animated Cache and other methods used to get this thing to render flicker free is largely due to the large self-illuminated polygon. I'm running some tests to see if this can be fixed easily.
Kray would handle this much better because it has a little system built in that converts bright polygons to lights. Lw could do with such a system.
Actually, I've used much brighter lumigons and was able to eliminate flicker even without AC. This is the first time I've seen flicker with AC! Couldn't this be from the changing sampling pattern of the lights from frame to frame?

You actually can convert the ceiling object into a light using dp custom light, but the sampling pattern is worse than lw's :\

erikals
08-11-2009, 05:06 AM
ok, now for testing the setup using lights,...
first test, render only, flickers, but lighting is ok...

http://www.newtek.com/forums/attachment.php?attachmentid=76179&stc=1&d=1250075031

geo_n
08-11-2009, 05:13 AM
noticed, there is a illuminated poly in the sealing that is missing in the LW render...

Which poly in the two renders? They're both vray btw.

erikals
08-11-2009, 05:32 AM
Which poly in the two renders? They're both vray btw.

oh, my wrong, thought it was Vray vs Kray or something,...

...i'm testing flickerfree(?) now,... let's see,...

JonW
08-12-2009, 12:37 AM
I changed the materials so they represent closer to real surfaces (50 to 80%, 80 still a bit higher than I like it). for the the defuse setting, also the main walls removed the reflective surface. Then increased the lighting to about 80% for the area lights.

If most of the spherical lights are replaced with spot lights, spherical are just not needed in most cases, I think this will get rid of the remaining flicker. 1 also put the 45 degree corner polys on the internal wall.

The render times are more reasonable with the above changes.

I tend to go up with AA by 5, 9, 17, 33, 65 etc, I find in general its better to jump to the next double number +1

This render I did at AA9

Zip 82 frames

erikals
08-12-2009, 01:16 AM
looks very good http://forums.cgsociety.org/images/smilies/smile.gif
what do you think?

to decrease flicker further you can increase SBR,
but more effective decrease maxpixspace, however that will increase renderspeed alot. (tons) at least on the first frame. (not too familiar with baking)

the 'problem' of course is that when you decrease maxpixspace, you also have to increase SBR, and sometimes RPE.

i think that's pretty close though... http://www.spinquad.com/forums/images/smilies/cool.gif (but close enough...?)

erikals
08-12-2009, 01:31 AM
curious, anyone knows what light type that has a circular look to it that renders fast?
spot light?

JonW
08-12-2009, 03:08 AM
I think the main issue is, rather than fiddling around with Layout settings, is try to get a typical real would representation of the surfaces.

Diffuse..... If this is not right you will go around in circles pulling your hair out, & I don’t have a lot to spare!

Start at around 50% & move up or down from there (I don’t know who genius was who has the default in LW at 100%, it gets all the new learners going off in the wrong direction with disastrous results; It happened me!)(A white ceiling shouldn’t be more than about 70%) Until you work this one out you will always have serious problems with your lighting , thus final renders!

The sphere lights on the side don’t need to be spherical they can be spot lights pointed to the wall & IB will do the trick, & the others can be replaced with area lights. The one in the roof does not need to spherical under any circumstances.

If you want to get the renders down to a reasonable time, something has to give, reflective walls & other unnecessary reflective surfaces, spherical lights, & diffuse surfaces at 100%.

The whole surface & lighting percentages need to be revised from scratch.

toby
08-12-2009, 03:40 AM
Well Lambert isn't realistic to begin with, if you want to take it that far you should be using Oren Nayer on everything, with at least a little roughness. And you can't really duplicate the number of light bounces in real life, so your surface settings need to be higher, that was especially true before there was radiosity. Oh and also, the light falloff needs a linear workflow to be accurate.

But, just imho, if the value you're getting on the walls is the value you want, that's the important part. If you lower your diffuse, then increase your light intensity because it's too dark, you've just wasted your time. I had a supervisor that made us use a sun at 300% intensity, and lower diffuse to compensate; we ended up at about 33%. Complete waste of time.

I also thought of this as more of a comparison to the other renderers, so things like adjusting the model kind of defeat the purpose. So I do think the spherical lights should be removed if they were not used in the other renderers, they're probably causing problems, and the unmotivated light source looks weird anyway.

toby
08-12-2009, 04:05 PM
Here it is with all the area lights replaced with lumigons, spherical lights removed and not replaced, and 8 bounces gi, AC, no color correction. 5 minutes to bake; renders were between 8 and 10 minutes per frame on my 2x2 G5! Probably 2-3 on you guys' machines

Still a tiny bit of flickering on moving parts, and some banding due to 32bit displays, worsened by compression -

toby
08-14-2009, 03:28 PM
Here it is with all the area lights replaced with lumigons, spherical lights removed and not replaced, and 8 bounces gi, AC, no color correction. 5 minutes to bake; renders were between 8 and 10 minutes per frame on my 2x2 G5! Probably 2-3 on you guys' machines

Still a tiny bit of flickering on moving parts, and some banding due to 32bit displays, worsened by compression -
anybody trying this on a fast machine?
(forgot to add, I'd turned off reflection)

toby
08-15-2009, 03:21 PM
Figured I'd better try it with reflection back on, see if that caused flickering. Still looks ok. Rendertime jumped from 10 to 30 minutes, with the first frame taking 1 1/2 hours :screwy:

erikals
08-15-2009, 09:01 PM
yep. that was my conclusion as well, flickerfree = 1+ hours...

geo_n
08-15-2009, 09:34 PM
blurry reflection in lw=bad:D
blurry reflection in kray or vray = good :thumbsup:

toby
08-15-2009, 09:39 PM
I thought you had a much faster machine than mine? And the 1 1/2 hour was only for the first frame, the following 29 frames were 30min.

Actually my settings left some reflection noise so it's probably not that big a difference when you factor that.

erikals
08-16-2009, 08:10 AM
i think the conclusion is that LW is not very fast as for rendering interiors... though can be rocketfast for rendering outdoor scenes in conjunction with motion blur...

...might be that Kray is the way to go for interiors, and even more so if it is an animation, i wonder how it's motion blur is though...