PDA

View Full Version : What render settings does everyone use?



tonyhall007
05-22-2013, 01:03 AM
When rendering a final image in Lightwave, I usually set my minimum sample to 10 and the maximum to around 30. I then leave my threshold at 0.01 and set my lighting and shadow sample to 12. I do this to try and get a nice looking render, but also to keep the render time down. What settings does everyone else use and why?

MSherak
05-22-2013, 10:53 AM
Depends on the scene. On the average I would say I keep everything at 1 and use AS 32-1024 to clean up.

tcoursey
05-22-2013, 03:23 PM
yep I almost always keep my light and shading samples to 1. AA min 1 and max somehwere between 20-60 depending on scene. Threshhold of .01 It keeps things moving fast and is pretty clean. We use alot of radiosity and baking of the cache too.

Greenlaw
05-22-2013, 04:01 PM
All of them. :)

Seriously, though, it depends on what you're rendering. I use completely different settings between high contrast graphics, toon shaded, photoreal GI, and hair/fur, etc.

This document is definitely worth the read: LightWave 9.x Anti Aliasing Guide (http://www.except.nl/en/articles/94-lightwave-9x-anti-aliasing-guide)

While there, you may also want to look at the LightWave Radiosity guide too. Both of these guides are bookmark-worthy.

G.

Thomas Helzle
05-26-2013, 09:10 AM
I agree with Greenlaw - it's more helpful to know what the settings do so you can adapt them to your current situation.

I wouldn't recommend that AA guide for LW 9.x anymore though since this has changed a lot in 11.x - the doc addendum has it nailed down very well IMO.

I personally find your "default values" very high.

Having the min samples at 10 does multisample every pixel in your image, even if it is just a black background colour. This is usually only needed in complicates situations with lots of fine details where a setting of 1 leaves too much to the AAA to find out. In low-variance scenes, it's a total waste of time.

I usually end up with min 1-4 and max 8 to "whatever is needed".

I also find that the Light and Shading samples vary very much depending on the scene.
With only hard shadows, multiple Light samples won't do you much good and can be left at 1, but with area or dome lights, you may find that raising this will leave a lot less to do to the AAA and can drastically improve render times.
The same with Shading samples: if shading creates noise, it's usually better to increase that value than the global AAA in my experience, although I found the most drastic results with Lighting samples. I often leave shading at 1-4.

BTW.
I am not really convinced by the new adaptive sampling approach in LW 11.x.
While LW 9/10 introduced a form of adaptive rendering that was very hard to understand and control, it had one point going for it that the new one is missing IMO: It did multiple passes at once after each contrast test AFAIK.
The new AAA does test the contrast after each and every pass, which is very wasteful in my opinion or at least could use another optional setting.

Why?
Contrasty regions in most situations don't change that fast, what needs refinement after 5 passes usually needs still refinement after 6, so it is my impression that the new AAA is actually way too fine grained, while the older version did more sub-passes before testing for contrast again.
What I refer to is especially the DOF scenario: you can easily end up needing 1024 or more passes for the max AA setting and it is pretty much useless to check for contrast 1024 times. Instead, I think it would be great to have a kind of adaptive adaptiveness (AAAA TM ;-) ).

This is how AAAA (TM) could work: After the first pass (or min passes), contrast is tested. Then (at least) twice as much samples are rendered, then contrast test again, then twice as much samples again and so on. The reason behind this is, that it is my impression that a lot of time is wasted on those 1024 contrast tests where usually render engines can be optimized much better when more samples are shot at the same time, since you don't have to go through all pixels every time, and more data will be still in the cache (being more "local").
A setting could be introduced for the definition of the steps. It would be interesting to see if it would make sense to make those a multiple of the min setting or leave it independent of it.
A setting of 1 could mean test after each pass, a setting of 2 test after 2, 4, 8, 16 etc.

The reason for this idea is, that AA is something that is changing less drastic the more samples are involved:
The first sample (if you only take one) defines a pixel 100%
The second sample makes that 50:50, the third 33:33:33, the fourth 25:25:25:25
If you take 1024 samples, one individual sample is down to adding 0.09765625% to the result or if you think in 0 to 1 terms, 0,0009765625 of the final pixel...

So the higher the max samples is set, the less change each pass will create.
And while I welcome the basic idea of the new AAA, which is way more understandable and controllable than the LW 9/10 one, I think it overshot it's goal by quite a margin.

Take for example Mental Ray (which I don't like but they have AAA nailed down pretty well IMO).
MR uses sampling in exponential steps, where 0 means one sample per pixel, 1 means 4, 2 means 16 and 3 means 64. Which is basically what I am talking about above. They know that the difference between 63 and 64 samples is completely meaningless in reality and their AAA is optimized for this kind of setup, so rays can be shot very efficiently and if I set max AA to 3, contrast is only tested 3 times.

For normal images I don't find this to be much of a problem, but when I tried some images with strong DOF and/or MB lately, I found it VERY hard to get smooth results, 1024 max samples wasn't enough but was very slow to render already, and I think this is related to all the contrast testing and it's drawbacks discussed above.
It is my impression that LW 9.6/10 did better in that case.

So I hope we will see some more development in this area.

Cheers,

Tom

Greenlaw
05-26-2013, 09:39 AM
I wouldn't recommend that AA guide for LW 9.x anymore though since this has changed a lot in 11.x - the doc addendum has it nailed down very well IMO.

I think you're right--force of habit I guess. :)

In fact, ever since 11.0 came out, I do find myself breaking a lot of my own long-standing guidelines for optimizing AA and speed. But, yes, it's best to have at least a basic understanding of what the settings do rather than trying to depend on a generic preset setting for all renders (which doesn't exist.)

I haven't watched these yet but here are two more AA guides for LightWave 11:

Quickstart Guide to Antialiasing in LightWave 11 (http://www.youtube.com/watch?v=_vHNd17aSiw) by DWBurman

Rendering, Sampling and Antialiasing (http://www.youtube.com/watch?v=UgNB9tZWUmM) by Rebel Hill


G.

Thomas Helzle
05-26-2013, 09:44 AM
hehehe - yeah, it was so hard to learn the LW 9/10 way of AAA that the new, much more understandable way is hard again - in the opposite direction ;-)

BTW, I made my rant above about the new AAA into a proper feature request:
http://forums.newtek.com/showthread.php?135709-Improvements-for-the-efficiency-of-Adaptive-AA-in-LW-11-x&p=1323538#post1323538

So if you share the sentiment, you may want to support it there.

Cheers,

Tom

OnlineRender
05-26-2013, 10:35 AM
went back to classic pass 9 just because in 11.5 area lights look hackit and yes I have read and yes I know its for the better but If I can get away with it ...

Mr Rid
05-26-2013, 03:25 PM
...

I personally find your "default values" very high.

Having the min samples at 10 does multisample every pixel in your image, even if it is just a black background colour. This is usually only needed in complicates situations with lots of fine details where a setting of 1 leaves too much to the AAA to find out. In low-variance scenes, it's a total waste of time.

I usually end up with min 1-4 and max 8 to "whatever is needed".

I also find that the Light and Shading samples vary very much depending on the scene.
With only hard shadows, multiple Light samples won't do you much good and can be left at 1, but with area or dome lights, you may find that raising this will leave a lot less to do to the AAA and can drastically improve render times.
The same with Shading samples: if shading creates noise, it's usually better to increase that value than the global AAA in my experience, although I found the most drastic results with Lighting samples. I often leave shading at 1-4....


What he said. And multi blur passes drastically change AA requirements. Am surprised what some people are generally defaulting to, and it indicates that AA settings remain too cryptic (after 2 tutorials, and reading docs and forums, I still dont quite get it). It shouldnt require a 45 minute tutorial to understand (which is not enough apparently). I really dont care what a sample is, or how the math works.

There's 'jaggies & noise' I want to get rid of. So increasing the value of an 'anti-jaggy/noise' value makes it 'better/render longer.' Lowering the value makes it 'worse/render faster' (although AS works backwards for whatever reason) Thats all I really want to deal with. I mash the gas pedal, it goes faster. I mash the brake, it goes slower. I dont care whats going on under the hood.



I also find that the Light and Shading samples vary very much depending on the scene.
...

Likewise, it varies greatly depending on each light or shader. So it seems weirdly inefficient for LW11 to lump the quality of every light or the shading quality of every shader under a single sample setting. This would be fine as a global option, but not as the exclusive quality setting for all related items. I mean, why not stick all volumetric light qualities or HV qualities under a single Q value, or make all shadow map sizes the same, or make all meshes the same subpatch level? The answer should be obvious. When only one of four area lights needs less noise & to take longer to render, why would I want to make all four lights take longer to render? Am I missing something? Were per item Q settings THAT confusing?

If 'ease of use' over 'confusing optimization' is the logic, then why not reduce AA quality to one value?

Thomas Helzle
05-26-2013, 03:45 PM
hehehe ;-)

Yeah, I'm with you on that. "Making it easy" can go pretty wrong.
IMO, it would be great to have these defaults but be able to overwrite them on a per-item basis: best of both worlds.

I found modo pretty annoying in the other direction, since you have to virtually go through each and every light and shader to set their individual samples in more demanding scenes.
And it's the only renderer I ever used were you can get more noise from using too many samples.

Finding the right balance is quite a herculean task - I hope they will not overdo it in either direction.

Cheers,

Tom

Mr Rid
05-26-2013, 04:20 PM
h...
IMO, it would be great to have these defaults but be able to overwrite them on a per-item basis: best of both worlds.

Exactly. Thats what I meant by making the global sample values a handy 'option.'

pinkmouse
05-26-2013, 05:07 PM
Presets would be nice. That way beginners, (like me), would have something to start with and learn by tweaking, and people could share if they found a particularly good or fast setting for a particular setup, say, glass objects with volumetric lights or whatever.

COBRASoft
05-26-2013, 05:53 PM
My 'start' settings for 'Render' are:
Ray Recursion Limit: 4 (6 or higher with lot's of transparency)
Ray Precision: 4 (higher with heavy reflection/transparency)
Shading Samples: 1 (4 or 8 to get faster clean up with dome/dpdome lights)
Light Samples: 1 (same as shading samples)

If using Global Illum:
Type: Monte Carlo
Interpolated: Off
Use Transparency: Off
Ambient Occlusion: Off (in most cases)
Use Gradients: Off
Blur Backgound: Off (better to do a full precision blur in the image editor, tip from Matt Gorner if I remember right)
Volumetric Radiosity: Off (except when using volumetric lights)
Directional Rays: Off
Use Behind Test: Off
Use Bumps: Off
Intensity: 100% (higher for indoor scenes)
Indirect Bounces: 2~3 (higher for indoor scenes)
Rays Per Evaluation: 4~8 (higher for indoor scenes could be a must)
Angular Tolerance: 30 (25 for longer render times and better result)
Minimum Pixel Spacing: 1 (depends on resolution size)
Maximum Pixel Spacing: 50 (same as MinPS)
Multiplier: 100% (sometimes 50% is enough, improves render speeds too)

Cameras:
Try to use Real Lens, it gives a 'natural' lens distortion, Perspective is also good
Minimum Samples: 1 (almost never change this one, I prefer to change the Shading and Lighting samples for more control)
Maximum Samples: 32 (sometimes 16 will do, 64 will give much less noise with Motion Blur/DOF, I hardly go higher than 64)
Reconstruction Filter: Classic (I prefer to do this in Post)
Sampling Pattern: Classic (best of both worlds IMO)
Soft Filter: Off
Adaptive Sampling: On
Threshold: 0.03 (0.01 will give much longer rendertimes and is often not needed. The AS will lose time on almost flat surfaces and this time could be spend on higher Max Samples)
Oversample: 0

When using Motion Blur:
PhotoReal
Particle Blur: Off (unless you have particles in your scene :))
Blur Length: 50%
Motion blur Passes: 1 (Never went higher than 3)
Shutter Open: 0
Efficience: 100
Rolling Shutter: 0

jwiede
05-26-2013, 06:52 PM
BTW.
I am not really convinced by the new adaptive sampling approach in LW 11.x.
While LW 9/10 introduced a form of adaptive rendering that was very hard to understand and control, it had one point going for it that the new one is missing IMO: It did multiple passes at once after each contrast test AFAIK.
The new AAA does test the contrast after each and every pass, which is very wasteful in my opinion or at least could use another optional setting.

I agree, as you note MentalRay (and others) take a more exponential-progression approach to contrast checking which seems to work fine, while checking after every single pass seems really wasteful once you're above say 4 or 8 passes, the contrast is already converging rather slowly, and the cost of contrast checking begins to dominate the time cost.


What he said. And multi blur passes drastically change AA requirements. Am surprised what some people are generally defaulting to, and it indicates that AA settings remain too cryptic (after 2 tutorials, and reading docs and forums, I still dont quite get it). It shouldnt require a 45 minute tutorial to understand (which is not enough apparently). I really dont care what a sample is, or how the math works.

The posts in this thread (and other recent ones like it) are a very clear indication, IMO, of the need for a decent set of "starting points" render settings templates for different situations. I see the amount of variance seen across users' "basic settings" in this thread actually as a negative, because of the contradictory nature of some users' settings versus others, etc.

That even many of the tutorials are now obsoleted by other tutorials and manuals saying what amounts to the exact opposite advice shows just how fragmented the user community's knowledge has become -- as many are still relying on now-obsolete advice as there are relying on current advice (and yet more working from yet different angles). Just posting responses in forum threads of suggestions for specific circumstances, and the occasional addendum section is clearly not working. The lack of coherency across all the various "guides", official documentation, and dev postings only excerbates the problem.

More documentation isn't enough to solve this situation anymore, it just winds up competing with existing, often obsolete docs. The app itself needs to provide more coherent ways of approaching the "baseline settings" issue. Starting templates have become a must-have, only then Newtek can begin to describe how the devs expect users to alter those templates to adapt to environmental conditions and scene difference.

Such an approach gives them a clear way to set forth viable control axes for the current renderer without falling into the quagmire of render settings' "what settings mean now" versus "what they used to mean" versus "what they've been generally misinterpreted to mean" which arises from too many voices without clear delineations of when certain advice became obsolete. Comprehensive reference documentation is another must-have, but by itself won't solve the situation anymore, IMO.

jwiede
05-26-2013, 07:08 PM
Starting templates have become a must-have, only then Newtek can begin to describe how the devs expect users to alter those templates to adapt to environmental conditions and scene difference.
I just want to clarify one thing: This only works if Newtek works to ensure the templates remain current with renderer changes, otherwise it will quickly devolve into a similar mess. Still, at least that templates are distributed with the app gives a fighting chance to ensure currency, unlike the docs/guides/posts approach used to date.

Thomas Helzle
05-27-2013, 07:30 AM
Not so sure about presets to be honest, but that may just be me.
There is a lot of different scenarios and I think they would be better served by good and well optimized example scenes to dissect, good documentation and some official tutorials.
ATM presets seem hard to do with the structure of how settings are distributed throughout the scene, but the unification sure made it simpler.

And honestly, the current settings in LW 11.x are not that complicated IMO in most areas.
Create a scene with a plane and a sphere and test each setting and what it influences - most things are possible to find out in about 10 Minutes.
The most important thing still is a curious and lively user in front of the machine ;-)

The thing that threw this whole thing of IMO was the weird concept in LW 9 & 10, where it wasn't clear at all what the renderer was doing and how much samples it would take.
That threw even seasoned CG veterans out of the loop and now remains as a residue of confusion.
Before that (LW 5.x to 8 or something, can't remember anymore), you had your fixed levels of quality where higher meant better and it worked well enough in normal scenes.
Now you have that again, but more flexible since you can enter your own numbers.

@Cobrasoft: ->Docs Addendum: " Examples of noisy lights ( = Light Samples) would be Area, Dome, Spherical and noisy surfaces ( = Shading Samples) would be things like blurred reflections or CarPaint..."
Ambient Occlusion is also part of Shading Samples.

My main problem is, that DOF is part of neither, so you can only use crazy maximum Samples settings to clean up heavy DOF images (I'd be happy to learn otherwise), which doesn't really work well and is very slow. Therefore my complaint about the excessive contrast checking.

Cheers,

Tom

Thomas Helzle
05-27-2013, 10:02 AM
One more thing:
To get a "feel" for the adaptive threshold, you can memorize that 1/256 = 0,00390 = ~0.004 = the difference between each colour step with 8 bit colours.

So if you would use that as threshold, the renderer would continue sampling until the difference between adjoining samples is something like RGB 128, 128, 128 and RGB 129, 129 129 - which would be only needed in very clean images where almost no structures and textures are present. Lower values would only make sense in situations where you work with 16 or 32 Bit images and intend to drastically increase contrast etc.

Usually you will stay higher though, since otherwise every textured surface in your image (where you actually want differences between pixels ;-) ) will get sampled to death.

Cheers,

Tom

pinkmouse
05-27-2013, 10:22 AM
...And honestly, the current settings in LW 11.x are not that complicated IMO in most areas...

Sorry Tom, but they are. I'm a creative type, not a programmer. Ideally, I just want a big slider with quality at one end and speed at the other. By all means, have all the current settings available behind the scenes for those that want to play, or understand the maths, but for the average user it's not what they want to be spending their time on.

Thomas Helzle
05-27-2013, 10:52 AM
I recommend unbiased rendering then. Start the render and just wait until it looks good.

Thea Render, Octane, Maxwell, Cycles etc. will do that for you.

With classical biased "fake" renderers, there is a certain need for knowledge if you don't want to render longer than needed.
No magic slider can solve that for you - otherwise it would exist.

I worked with artists with the same attitude as you (which I personally think says more about your teachers than about you) and usually after me explaining to them how things work and what is what, they had not much problems sorting out what is relevant for them and what can be forgotten for good.
It IS simple :-)

Everything else is just a widespread misunderstanding... ;-)

Cheers,

Tom

geo_n
05-27-2013, 11:03 AM
No magic slider can solve that for you - otherwise it would exist.


http://solidrocks.subburb.com/ :D
I've had speed gains but only 10% since vray is easy to use already but the slider feature for optimizing is pretty good for people who don't know vray much and those who don't really want to or need to.
Kray will have something similar for v3.0.

Thomas Helzle
05-27-2013, 11:11 AM
LoL, yeah, Snake Oil 3D ;-)

I'm sure somebody can hack together such a script for LW in an afternoon...

Cheers,

Tom

geo_n
05-27-2013, 11:13 AM
I'll be waiting for your script then. :D
Its not snake oil. Its an optimization system actually and it works. Or maybe G who is also implementing something similar for kray is not a genius after all.

Thomas Helzle
05-27-2013, 11:18 AM
I'm not interested in this kind of stuff, really ;-)

But each to his own.

Peace,

Tom

geo_n
05-27-2013, 11:43 AM
Complicated could be relative and in this case it is complicated for the OP. Don't know how teachers and user attitude come into play here. A preset system, templates, starting point, to optimize renders exists for other renderers, unfortunately it doesnt in lightwave....yet. So the OP asks which settings users use. Imho the settings can be simplified and some people like G for kray and even devs I'm working with are going that route.

pinkmouse
05-27-2013, 01:21 PM
...I worked with artists with the same attitude as you (which I personally think says more about your teachers than about you)

I'm self-taught! :D

I understand some people want to know every single detail and setting, if you're rendering a 30 second clip to 4K you need to squeeze every single second you can off your render times. Fair enough. But most of us don't need that depth of knowledge. Unless of course, you're offering to produce a tutorial to show us all how simple it is? ;)

Thomas Helzle
05-27-2013, 01:45 PM
Sorry I let myself be dragged into this kind of discussion and seem to have derailed the thread - wasn't my intent - sorry for that.

Cheers,

Tom

bobakabob
05-27-2013, 02:22 PM
Sorry I let myself be dragged into this kind of discussion and seem to have derailed the thread - wasn't my intent - sorry for that.

Cheers,

Tom

It's been a very interesting thread to read. As a starting point with renders I tend to start with classic render settings, an area light for soft shadows, Lights and Shading set to 1; Min 5, Max 15; Adaptive sampling 0.01 and 0.5. ReblHill's video is great at demystifying the controls and as he does in the tutorial it's worth zooming in 400% or so to see more clearly what's going on with pixels especially regarding noise and 'jaggies'. The default Monte Carlo interpolation settings seem to deliver very good quality.