PDA

View Full Version : Rendering issue....



the3dredege
03-02-2011, 09:54 AM
Hello,
i'm trying to render a scene which is pretty high reflective but each time i get these ugly black splotches all over the place, mainly when radiosity is switched on.
Anybody have any ideas?

Thanks for your help



G

3DGFXStudios
03-02-2011, 11:37 AM
which version of lw are you using? Is cache turned on?

the3dredege
03-02-2011, 11:40 AM
thanks for reply. I'm using LW10. Tried with cache on and off, still wasn't happening. I used a nodal surface for the high reflective black, but went back to old LW surfacing which cleared it, but it's not as nice as the nodes. I've searched everywhere but no one seems to have any answers?

3DGFXStudios
03-02-2011, 11:42 AM
have you tried rendering it in 9.6? 10 has currently a radiosity bug in it.

the3dredege
03-02-2011, 11:54 AM
Ahh that i didn't try! I was running our of time.
Have you heard of or seen this before?

3DGFXStudios
03-02-2011, 01:56 PM
yup I've seen this before. Several times. Are all your radiosity flags turned off? (don't think that's the problem but it could help you find it) What do your radiosity settings look like?

the3dredege
03-02-2011, 02:07 PM
Radiosity is pretty minimal. Final Gather - interpolated with volumetric Radiosity. All other points default. I worked on the surface texture some more and with some decent lighting and fast fresnal managed to pull it back round enough to show the client.
Still odd though. I've had it once before and it just seems to be with nodes with high reflection?

Thanks for having a look. Really appreciate your help.

3DGFXStudios
03-02-2011, 02:27 PM
What happens when you turn on use transparency and directional rays?

Truecoz
03-04-2011, 01:44 AM
Don't use 'Interpolated' it's NO GOOD right now. Set 'Rays Per Evaluation' to 4. May want to up your 'Indirect Bounces'. I'd try these settings first.

3DGFXStudios
03-04-2011, 03:18 AM
Rays per Evaluation to 4? It better to set them to 1500. 4 Way to low even for testing.

Hieron
03-04-2011, 06:37 AM
Ehm.. what radiosity/interpolation bug??

(just started using 10 in production...)

3DGFXStudios
03-04-2011, 06:46 AM
Check the forums. There are a lot of threads about it.... ;)

Exception
03-04-2011, 07:53 AM
Probably a bug in one of the nodes. This happens when a ray encounters a state in which it errors out. Everything it touches becomes black. This is always a bug. I haven't seen any of these since LW 10's release, but you found one. Send the scene to [email protected] with a description, NT will be very happy with it. Meanwhile, I'd suggest looking for nodes which may cause this. If you use funky surface nodes anywhere, turn them off and see if it disappears.

Hieron
03-04-2011, 09:23 AM
Check the forums. There are a lot of threads about it.... ;)

You could have given a hint, just naming it 3x is a bit.... "GI" + "Issue" is not ideal to search. I'm not on the forums much these days.

You probably mean the "hide object from camera" bug? Hardly a showstopper for me.. I thought it was something worse.

Truecoz
03-04-2011, 11:53 PM
Rays per Evaluation to 4? It better to set them to 1500. 4 Way to low even for testing.

I'm sorry but you are wrong. With LW 10/9.6 the renderer has been reconfigured in such a way that that a higher setting makes no sense when a lower setting yields better results just try it for yourself and compare the difference. If you don't know about LW render-er current state from real experience then I suggest not posting, because now you've posted a misleading thread that's not true.

Hieron
03-05-2011, 10:24 AM
ehm lol?

Imho, 4 is way too low and 1500 is too high :P But just saying the other has no clue is not really nice.

Truecoz
03-05-2011, 11:21 AM
But just saying the other has no clue is not really nice. Agreed, in general it is rude to say someone doesn't have a clue. But that's is not what I said so please don't make anything more out of what was said. To tell someone that they are mistaken about a setting for radiosity, hopefully helpful, and actually if it saves time because if they render an animated sequence and it takes longer to render with higher settings, and that those settings even cause worse flickering animated, could save them in a crunch.

Hieron
03-05-2011, 04:26 PM
I'm sorry but you are wrong. With LW 10/9.6 the renderer has been reconfigured in such a way that that a higher setting makes no sense when a lower setting yields better results just try it for yourself and compare the difference. If you don't know about LW render-er current state from real experience then I suggest not posting, because now you've posted a misleading thread that's not true.

Ah I completely misread this post then. English is not my native language, I always thought that "if you don't know about x then I suggest not posting" and "you are wrong" and "is not true" to be sort of equal to "you have no clue'. Clearly I am wrong.

Since I am however eager to learn, I quickly set up a simple testscene to try out my new learned knowledge.

Attached are the results. I'm confused and need more guidance.

93345
Pre LW 9.6/10 way with high rpe

93342
"Reconfigured renderer" way


And going the non-interpolated way is just plain weird. Unless you have spare time I guess... alot.

lardbros
03-05-2011, 06:27 PM
I'm not sure which way the LW renderer has been re-configured for better results with lower settings. I think some of the anti-aliasing things were messed with in 9.6, or maybe 10, so you could lower your Area-light samples and get better results, or if you raised them very high it doesn't add massively like it used to before... but I could be just adding to the confusion here!? :S

Hieron
03-05-2011, 06:32 PM
While the new GI etc is great, I'm skeptical of the low RPE as suggested, but I guess I wasn't too subtle with my skepticism.. :) But if anyone can show how it's done, be my guest.

My only gripe atm is that sensible quality settings for Area Lights make animated GI cache go splotchy and the only fix is setting the quality overly high (AA comes seperate from GI and will make the Q of area lights overdone. Sadly.)

Truecoz
03-07-2011, 02:33 AM
My turn to post... I'm done with this thread after this one. But I attached scenes for anyone to check out my low RPE with high AA. Good Luck! :thumbsup:

3DGFXStudios
03-07-2011, 03:29 AM
While the new GI etc is great, I'm skeptical of the low RPE as suggested, but I guess I wasn't too subtle with my skepticism.. :) But if anyone can show how it's done, be my guest.

My only gripe atm is that sensible quality settings for Area Lights make animated GI cache go splotchy and the only fix is setting the quality overly high (AA comes seperate from GI and will make the Q of area lights overdone. Sadly.)

If you are skeptical about what I suggested you should read the radiosity guide because it does make sense what I was suggesting ;) :thumbsup:

funk
03-07-2011, 04:59 AM
My turn to post... I'm done with this thread after this one. But I attached scenes for anyone to check out my low RPE with high AA. Good Luck! :thumbsup:

That scene just renders black...:confused:

funk
03-07-2011, 05:09 AM
OK I just made the cube in the middle 200% luminous and compared your low rpe settings to interpolated higher rpe settings.

Low RPE is slower and granier. Am I missing something here?

You need to use AS with low tolerance (0.01) if you want to use low rpe with no interpolation but the render times sky rocket and you still have some grain

3DGFXStudios
03-07-2011, 05:43 AM
Sorry but your settings are wrong. Why does no one read the radiosity manual? RPE should never be that low. It's fast but it not accurate at all, and in animation it's never going to work. You need at least 100 rays. And your Secondary bounces are to high. you almost never need more than 50. A value between 10 and 100 is good for the secondary's. Also you mps is to high. Lower it to something less. I know your lighting looks smooth but that just because your rays are enormous.

funk
03-07-2011, 05:54 AM
@3DGFXStudios Agreed.

@Truecoz Read this guide by except who has done quite a bit of research:
http://www.except.nl/lightwave/RadiosityGuide96/index.htm

I start with mote carlo / interpolated / use gradients / use behind test / rpe 200 / sbr 50 / min ps 2 / max ps 100 / at 45, then tweak from there.

Min ps of 10 is way too big (when interpolated) and youll miss contact shadows and shadows in small details

3DGFXStudios
03-07-2011, 05:54 AM
Check these settings. 42second render and pretty accurate lighting.

3DGFXStudios
03-07-2011, 05:56 AM
@3DGFXStudios Agreed.

@Truecoz Read this guide by except who has done quite a bit of research:
http://www.except.nl/lightwave/RadiosityGuide96/index.htm

I start with mote carlo / interpolated / use gradients / use behind test / rpe 200 / sbr 50 / min ps 2 / max ps 100 / at 45, then tweak from there.

Min ps of 10 is way too big (when interpolated) and youll miss contact shadows and shadows in small details

Exactly! :D

funk
03-07-2011, 05:59 AM
Note though, that min ps doesnt have much effect in non interpolated FG mode (which he was using). It only seems to effect color bleed from secondary bounces

3DGFXStudios
03-07-2011, 06:07 AM
If you want to see what the radiosity does you can use the "Radiosity Flags" found in the Render tab on the right side. (studio productions style menu layout) ;)

Hieron
03-07-2011, 10:46 AM
If you are skeptical about what I suggested you should read the radiosity guide because it does make sense what I was suggesting ;) :thumbsup:


Eh dude. I wasn't referring to you, in fact I was siding with you that 4 would be too low. Pff, so Truecoz is going "if you have no clue don't post" to you, I defend you, and I get a rtfm back? Like I have no clue about GI in LW... as if that's rocket science.

pff :) Come on man, how could I ever be skeptical about my fellow dutchies anyway



My turn to post... I'm done with this thread after this one. But I attached scenes for anyone to check out my low RPE with high AA. Good Luck! :thumbsup:



Ow come on, you're just trolling now :) First that scene and object don't match, it's the wrong object. Second you use interpolation. Third your RPE is 16, not 4 (not that it matters). Fourth the scene lacks the 2nd layer... etc

Truecoz
03-07-2011, 07:39 PM
Eh dude. I wasn't referring to you, in fact I was siding with you that 4 would be too low. Pff, so Truecoz is going "if you have no clue don't post" to you, I defend you, and I get a rtfm back? Like I have no clue about GI in LW... as if that's rocket science.

pff :) Come on man, how could I ever be skeptical about my fellow dutchies anyway

Ow come on, you're just trolling now :) First that scene and object don't match, it's the wrong object. Second you use interpolation. Third your RPE is 16, not 4 (not that it matters). Fourth the scene lacks the 2nd layer... etc

Interesting that Funk got it right.

I don't want to post because the information being posted from others trying to peg what I'm posting as false almost solely on that's what it says in the manuals. And when a few of you constantly take what I'm saying out of context and try to justify your own position without really doing any homework. I don't know if I can help out because everything just becomes misconstrued, and probably the person who posted originally, gets all confused and doesn't know what to do and maybe they just give up on it altogether.

Most of you are basing your information on old guides, like what's posted below. I'll tell you guys what... How many of you go to the Los Angeles LW user group meetings? How many of you are have met with Rob Powers and the developers at Newtek directly? I'm in on meetings where the top LW artists in Los Angeles are telling Rob Powers and the LW developers what's wrong the LW render, and why they need to change the manual and the default settings. Anyway, I'm really just trying to help you guys out.

@Truecoz Read this guide by except who has done quite a bit of research:
http://www.except.nl/lightwave/Radio...de96/index.htm
This guide is based off the old render-er


OK I just made the cube in the middle 200% luminous and compared your low rpe settings to interpolated higher rpe settings.

Low RPE is slower and granier. Am I missing something here?

You need to use AS with low tolerance (0.01) if you want to use low rpe with no interpolation but the render times sky rocket and you still have some grain

High AA but with no AS, please turn off AS. Add even more AA to reduce grain more. Now animate the camera and render out a sequence. Try it for both scenes and compare. Please post your results. Thank you very much Funk for really trying this out.

geo_n
03-07-2011, 10:10 PM
AS should always be off anyway unless you're doing stills.
But this is interesting to set rpe to as low as 4 and sec bounces rays very high.
I always set it to rpe 16-24 and sec bounces rays to 4-16. Renders fast with those settings and less gi flickers.

Have to try those low rpe and high sec bounce on some scenes. Is this for lw 10 or applicable to lw 9.6?

geo_n
03-07-2011, 10:25 PM
just testing with rpe 4 and rpe 16.
rpe 4 has low brightness. Hmmm..will I have to adjust the lighting to account for the low rpe?

Truecoz
03-07-2011, 11:01 PM
I think it does go back to 9.6 but not entirely sure - I'm running 10 now, but I have to load some of those in 9.6 (10 is pretty backward compatible) because of an error I get with Fiber FX in 10. Seems to be the same settings work in 9.6 as in 10. So, I'd say yes to render radiosity is the same in 9.6 as 10.

Reducing the RPE and raising the AA has been faster from every render test that I've done.

One last thing... I work for a major VFX studio in Los Angeles. I'm sorry if I don't have the time to set-up a lot of examples and stuff. I'm just trying to do what I can do and help out the LW community with the time I'm allotted. I barely have any time to post solutions, or answer questions, but I'm trying to do as much as I can to help Lightwave grow more in the CG community.

3DGFXStudios
03-08-2011, 02:29 AM
nevermind

3DGFXStudios
03-08-2011, 02:51 AM
I'll test it, but I've to say that the settings I normally use create very nice renders with no flickering light. For example I made a commercial for Heineken with dielectric glass and volumetric radiosity and it rendered incredibly fast all 1080p (3000 rays). So if you are saying that I don't know what I'm talking about you are wrong. At our studio we use LW mainly for rendering and the reason for that is that LW renders fast and has good results.

so please dude don't say to people that they don't know what they are talking about if you don't know who you are talking to.

Hieron
03-08-2011, 04:15 AM
Ok, let's make this discussion efficient then, all this talk about user groups, big VFX studios and Rob add nothing to the amount of info or education, /care. Attached an example scene, I'd love to see it render faster. Please help a guy out and show how it's done. Or anyone else on the user group or so.


93423

I'm eager to learn.

93424
Old way: 30 sec render of which 22 GI and cacheable, 5 bounces. 1500 RPE. No splotching. Cacheable -> 8 seconds per frame time.

93425
New way: 2m30 sec render of which 1.2 sec GI, 5 bounces. 10 RPE. Still splotches a bit. (I'm sure this can be improved)

Scene is above, 30 seconds the time to beat and you got all my attention and gratitude if you can back it up.

ps: post considerably :) edited for niceness.
pps: I can see how it can be usefull for stills, if splotching is no issue and in some way the noise removed. It does have a lesser tendency for the really bad splotching with very low settings..

lardbros
03-08-2011, 05:42 AM
Most of you are basing your information on old guides, like what's posted below. I'll tell you guys what... How many of you go to the Los Angeles LW user group meetings? How many of you are have met with Rob Powers and the developers at Newtek directly? I'm in on meetings where the top LW artists in Los Angeles are telling Rob Powers and the LW developers what's wrong the LW render, and why they need to change the manual and the default settings. Anyway, I'm really just trying to help you guys out.



Sorry, but it would be great if you could have gone into more detail about why Interpolated has gone wonky in v10. I've not hit any problems so far...will give these settings a try later... so thanks for your words of wisdom...

Also, it might be worth being a little less angry, I'm not having a go, nor doing down the work that you do... but you sound angry in your wording... and although some of us don't go to LA for LW user meetings with Rob Powers and the devs, we do have similar meetings when they come up from time to time... just we haven't heard about this particular, and interesting radiosity info...

Thanks for informing!! :)

funk
03-08-2011, 07:40 AM
@Truecoz I'm not sure what I got "right" :) I just used your scene, edited the box surface and rendered it out. My point was, your settings were slower and grainier than using interpolated settings. I only showed the AS 0.01 example so we could see render times became huge. They become just as long or even longer with no AS and high AA (to remove the same amount of grain).

Am I missing something here? Can you please explain how this method is better?

Unless you can clarify, I'll stick to my current method which is based on the ideas in excepts radiosity guide. I get smoother radiosity and it renders faster

As I stated before, I normally start with monte carlo / interpolated / use gradients / use behind test / rpe 200 / sbr 50 / min ps 2 / max ps 100 / at 45 and tweak from there.

Please dont think I'm trying to argue. I would just like to understand where you are coming from and even attempted to use your settings. Thanks for your time! :)

jrandom
03-08-2011, 03:46 PM
Note though, that min ps doesnt have much effect in non interpolated FG mode (which he was using). It only seems to effect color bleed from secondary bounces

And today I learned that you can use Final Gather in non-interpolated mode. I never actually tried this before because I assumed that FG was always interpolated.

Heh.

How does non-interpolated FG compare to non-interpolated Monte Carlo?

Truecoz
03-08-2011, 10:24 PM
Don't use 'Interpolated' it's NO GOOD right now. Set 'Rays Per Evaluation' to 4. May want to up your 'Indirect Bounces'. I'd try these settings first.


Rays per Evaluation to 4? It better to set them to 1500. 4 Way to low even for testing.


I'm sorry but you are wrong. With LW 10/9.6 the renderer has been reconfigured in such a way that that a higher setting makes no sense when a lower setting yields better results just try it for yourself and compare the difference. If you don't know about LW render-er current state from real experience then I suggest not posting, because now you've posted a misleading thread that's not true.


ehm lol?
Imho, 4 is way too low and 1500 is too high :P But just saying the other has no clue is not really nice.

Here, I just re-posted what I posted and what you two have posted, in there entirety, just so everyone can see for themselves how ridiculous it is.

I never said 3DGFXStudios didn’t have a clue… Hieron did. In fact Hieron you keep taking small pieces of my post and re-posting it out of context. The only thing I posted about was only referring to the LW render-er and radiosity.



so please dude don't say to people that they don't know what they are talking about if you don't know who you are talking to.

Really… and consider that logical. Well you sound like an egomaniac. Apparently you think you shouldn’t tell people that they are mistaken, when the are mistaken, if they are considered (by what criteria I don’t know, but let’s say societies) more successful then you, or what, older than you, what’s the criteria. Please explain it, under what circumstances should you not tell someone they are wrong when they are wrong? But you’re right I don’t know who you are really; please enlighten me. Who are you?

I’m not angry just frustrated that Hieron and 3DGFXStudios are crapping on me just because I said 3DGFXStudios was mistaken about Lightwave’s render-er and radiosity. Then I tried to point out why I’d know more about it then you guys, and you took it totally the wrong way. I'm sorry I don't do crazy and you two are nuts! Good luck with that attitude!

Truecoz
03-08-2011, 10:56 PM
One last thing... When you go to view my profile here guys, you have my full name first, and last. You can look me up on IMDB.com, you can contact me directly both my email and website are there. On my website you can see all my work. Peace!

geo_n
03-08-2011, 11:18 PM
Truecoz - I think your setting might be for animation, right? I've also had good render time and quality ratio with lower numbers. Using with mblur to cheat the gi. But still not perfect yet in lightwave.
But for still render I do keep it in the 64 to 96 RPE range and secbounce to 16 to 24. I find the higher RPE gives sharper looking gi, good contact shadow and overall more depth in lighting.
1500 RPE seems very high. I have never tried it due to higher render times when I set over 300 RPE. Worth a look.

Truecoz
03-09-2011, 12:11 AM
Yes you should set RPE to 4. Attached is a screengrab of my all my settings plus the render time at 64 pass AA, no grain, 46 seconds. For animation you may want to use something like 8 - 24 pass AA for faster rendering times.

the3dredege
03-09-2011, 03:06 AM
Hi,
I'm the original poster of this thread and I thought I'd take a step back from all this for a couple of days and see how it runs. Although my post hasn't been answered in its entirety it has been really useful. I've been trying the various settings that everyone has been putting up into my own scene, and although i got round my original problem of the nasty black dots (it was a nodal issue) I've learned a whole lot more about radiosity. I too have always used the Except Radiosity guide to steer me which has been invaluably, but i think maybe an update to LW10 would be of real interest. Maybe they are working onit?
Anyway, although it seems a heated debate, I really appreciate it.
Thanks to everyone with the interest to post!

Hieron
03-09-2011, 04:47 AM
One last thing... When you go to view my profile here guys, you have my full name first, and last. You can look me up on IMDB.com, you can contact me directly both my email and website are there. On my website you can see all my work. Peace!

Yes, that work and website shown there makes me doubt your words on GI.


blabla, denial, crying

+

"Then I tried to point out why I’d know more about it then you guys, and you took it totally the wrong way. I'm sorry I don't do crazy and you two are nuts! Good luck with that attitude!"


Ok....



Yes you should set RPE to 4. Attached is a screengrab of my all my settings plus the render time at 64 pass AA, no grain, 46 seconds. For animation you may want to use something like 8 - 24 pass AA for faster rendering times.

46 seconds, and *it still splotches like mad*, has noise, and you don't use LWF but I guess you'll know better on that regard as well. It's 20.8 seconds here for later comparison. lwf or not ofc.

So I took your scene (really not ideal scene but hey it works this time) and changed it. Attached results, with LWF btw perhaps you learn a thing.. you know, the new feature probably discussed endlessly on the user group.

Tadaaaa! No splotches, 20.8 seconds went to 7.0 seconds.
This is getting utterly ridiculous, but I guess it's *my* attitude. :)

93468

93467

93469

But sorry, I'm probably just not seeing your genius. Not here and not on your website you pointed to. So, drag your user group in here or Rob and beat that 7.0 second, unsplotched. Challenge still stands, beat 7 seconds (+- 15 on your pc) unsplotched.


ps: the only downside to the 7.0 seconds normal GI render is the less sharp transitions of the light at the bottom. If I'd allow 20 seconds like yours, I'd still have it unsplotched, you wouldn't, and match the sharp transitions as well. Such sharp transitions don't usually come up strong as this in real projects though but hey, it could.

pps: even if you could cache the GI so the splotching doesn't show up as much, the GI part is super small (0.1 seconds)! The 64 AA passes take up most of the 20 seconds and are not cached. So where my 7 seconds render goes to 1.8 seconds for cached animation, yours goes to 21! That is 10x + slower!.. err.. myeah.. awesome indeed.

ppps: 2 renders of the same frame of your scene, with LWF. See how they are different? Try 2 renders of same frame on my scene, they match. Stuff like this sticks out like a sore thumb in animation and is seen too often, even on TV on cheaper work.


93471
93472
Slow 22 seconds renders, splotching and noise visible.

pppps: ( :) ) I would never post it so specific as I expect this to be rather common knowledge here, mostly thanks to Tom/Except and his great guides for 9.6 for which I happily donated. While I am very eager to learn, I do absolutely hate it when someone comes barging in with statements he cannot back up, tells others to be silent, meanwhile naming usergroups, Rob, IMDB etc yet cannot show how it is done better. I do not doubt this is discussed at such a meeting, but I would absolutely like to hear how it is done well, nothing of the above posts shows this, at all, remotely. (yet I didn't know about this FG mode, so I did learn of another, not ideal, GI mode) If this "new" FG mode is something NT is investing in, I sure would like to see it improved, *ALOT*.

edit: I keep editing the post for niceness in the hope you will actually read the info, and try it out.

3DGFXStudios
03-09-2011, 05:04 AM
That's not a website of a genius like you are pretending to be Truecoz. Hahaha LOL you know more than we do? Yea Right. Your site isn't showing that. Like you know us? Dude come one who are you kidding.

Hieron! You are totally right! That's how you do it!

lardbros
03-09-2011, 05:37 AM
Come on guys... it's not fair to judge someone on their website really... I understand that it's a public display of people's work and could be seen that it's there to be judged, but my website hasn't been updated in 6 years... not because I haven't done any decent work, but because I've been working so hard that I just don't have time to touch it or update it.

I think TrueCoz has been a little rude, barging in here and telling us he works at a major studio in LA and he talks to Rob P, and making it sound like we really should take notice... but be courteous to him... he's still trying to help in his own way!


Edit:
I can't find Truecoz's website on here... maybe he's removed his name from the profile?

Hieron
03-09-2011, 05:50 AM
One last thing... When you go to view my profile here guys, you have my full name first, and last. You can look me up on IMDB.com, you can contact me directly both my email and website are there. On my website you can see all my work. Peace!

I agree Lardbros, but when someone uses it himself in a discussion I can only assume it is used to prove a point. Even then, one can not comment on it? (No clue about the name, sitelink working fine here, just click on the nickname. Site is relatively up to date)

funk
03-09-2011, 06:07 AM
Yes you should set RPE to 4. Attached is a screengrab of my all my settings plus the render time at 64 pass AA, no grain, 46 seconds. For animation you may want to use something like 8 - 24 pass AA for faster rendering times.

You state there is "no grain", when there clearly is. I'm starting to think that maybe your monitor isnt calibrated and you can't see it? You also dont use a linear workflow, which will make the grain/noise more apparent.

I rendered the exact same scene in my earlier post with the same settings you had, but enabled interpolation and lowered AA. It rendered in 9.6 seconds with much smoother GI.

I'm sorry man, but your settings simply dont look as good and take 5 times longer to render.

Update: Im attaching a section of your render and adjusted the levels in PS so you can see the noise/grain (if your monitor has decent calibration, you'd see it in your original anyway). Now Im not saying this is "bad" grain... some people add noise in post anyway. I'm just saying you can render without this noise much faster, then add noise in post if needed.

Truecoz
03-09-2011, 09:37 AM
FYI I haven't had time to update my site in a bit as I been working. Anyway, all good. My website still there, perhaps your internet. When are you 2 guys going to post your full names, let's see what shows you've worked on at IMDB.com Anyway, I'm not holding a gun to anyone head saying you have to use my settings. But some of you are straight up lying on these posts. Again, good luck doing any animation with no flicker, with your settings. Next time you 2 are in Los Angeles look call me up we'll get together, hit up the town.:D

Hieron
03-09-2011, 09:50 AM
I don't feel the need to post "what IMDB shows" I've worked on (none btw, omg!) nor the need to catch up if I am in LA again, thanks.
*deletes and edits* :) bah.. ok I'll play nice too then 3DGFXStudios :) (damn, tough nick to use as a name)

3DGFXStudios
03-09-2011, 10:17 AM
haha I'm on IMDB ;) But what does that say? I see you are a bit older than me Chris so you have obviously more credits on IMDB than I have. But again what do credits on IMDB say about your render skills. Also I noticed that you are a compositor most of the time. That's also not related to GI settings. Correct me if I'm wrong please ;)

Could it be a possibility that both our settings are good? I never have flickering renders so I must be doing something right? Right?

When I'm in LA I'll contact you and you may also contact me if you are in Amsterdam. :) We'll have a beer then. ;)

cheers

jrandom
03-09-2011, 10:39 AM
Using an RPE of 4 is interesting -- the GI pass is very fast but very noisy for indirectly-lit areas. The AA of 64 is awfully slow (especially if you happen to have blurry reflections) but it does result in a grain that can be handled by today's video noise removal filters (see NeatVideo (http://www.neatvideo.com/)). This might bolster Chris's argument a bit.

I can't support his combative demeanor, though. :P

The advantage to the low-RPE non-interpolated approach is you don't get the subtle blotchies that show up in animations. This approach is far less useful for stills. I tried the RPE of 1500 which gave me faster renders but horribly slow GI (darned this blurry-reflection-surfaced floor!). I still haven't found a good balance between the two for something that results in animation-quality GI -- and I'm running on a 12-core hyperthreaded Mac!

(Note: I am still a total noob in Lightwave and have a fondness for computationally-expensive materials.)

Hieron
03-09-2011, 10:50 AM
But you do get the subtle blotchies/splotches. Check the last scene posted by Chris and the 2 example renders of them I posted, it still has that issue, right? If I put the settings so it is gone it is superslow.. :/

I use Neatvideo alot, if it were for just the normal high frequency noise I could live with it probably. And the fast GI pass makes caching useless speedwise.

So while it is interesting as the renderer is clearly doing a different thing than "normal" and thus it could be that this is something that will be developed more, I can't see it being usefull yet.. but if someone could take any of the example scenes in the thread and show otherwise, I'm happy to learn. Perhaps it is more usefull in big outdoor scenes, but using 64 AA on any of our big arch viz scenes would explode rendertimes, can't see that happen soon.

3DGFXStudios
03-09-2011, 11:42 AM
If you use less rpe than your lightspot/splotches will be very very big, they get so big that you can't recognize them as splotches. But if you look closely you see big areas of the image flicker....In a still image you can use a very low rpe and if you cache it you also can do that but if you want some accurate stuff you need some more rays. Ooo yes and very high AA will bite you in the bottom on an other level. Like Hieron mentioned with blurry reflections.

Truecoz
03-09-2011, 11:42 AM
Hieron

Yes, you can render a still image the old way if you like (I don't see blotches in my renders but whatever). But that doesn't work with an animated sequence. If you start using settings for rendering a still image, then different setting for rendering an animation sequence, that's probably a road that most would not like to go down. So, to keep everything consistent don't you think it would be advantages to find a setting that works for both. So let's see an animation sequence with your settings.

lardbros
03-09-2011, 11:45 AM
Let's see an animation sequence with your settings too TrueCoz. I believe you, but don't have the time to test!

Truecoz
03-09-2011, 11:49 AM
Yes, I'll post something later when I get home from work. Also, when rendering a still image, typically the render time matters less especially when you're talking about a matter of seconds.

jrandom
03-09-2011, 11:50 AM
I'm seeing an even bigger issues with the low-RPE technique: really bad lighting fidelity.

FG RPE-4/128, AA-64, 1m 31s render time:
http://www.newtek.com/forums/attachment.php?attachmentid=93483&stc=1&d=1299696370


FG RPE-1500/200 Interpolated, AA8 , 40s render time:
http://www.newtek.com/forums/attachment.php?attachmentid=93484&stc=1&d=1299696379

(Ignore the banding in the second image -- I don't quite have this newfangled linear workflow figured out yet and turning on dither makes it look terrible.)

I've attached the scene I'm using if anyone else wants to monkey with it. I've also not tried any animations yet.

3DGFXStudios
03-09-2011, 11:56 AM
Good to mention that time matters. But I'm willing to sacrifice some time to get an accurate render. If you cache your gi it won't flicker whatever settings you use. If you cache a frame with RPE at 3000 the second frame will be cached in a second and you won't have any noise and your GI doesn't flicker and if you compare the quality and time with the lower settings you'll notice that you don't win a lot of time using less rays.

3DGFXStudios
03-09-2011, 11:57 AM
I'm seeing an even bigger issues with the low-RPE technique: really bad lighting fidelity.

FG RPE-4/128, AA-64, 1m 31s render time:
http://www.newtek.com/forums/attachment.php?attachmentid=93483&stc=1&d=1299696370


FG RPE-1500/200 Interpolated, AA8 , 40s render time:
http://www.newtek.com/forums/attachment.php?attachmentid=93484&stc=1&d=1299696379

(Ignore the banding in the second image -- I don't quite have this newfangled linear workflow figured out yet and turning on dither makes it look terrible.)

I've attached the scene I'm using if anyone else wants to monkey with it. I've also not tried any animations yet.

Hey thanks for the scene. Nice one to test GI in ;)

jrandom
03-09-2011, 12:05 PM
Good to mention that time matters. But I'm willing to sacrifice some time to get an accurate render. If you cache your gi it won't flicker whatever settings you use. If you cache a frame with RPE at 3000 the second frame will be cached in a second and you won't have any noise and your GI doesn't flicker and if you compare the quality and time with the lower settings you'll notice that you don't win a lot of time using less rays.

The trouble with caching is that if you have a lot of moving elements in indirect lighting, caching isn't going to work so well.

I just did some dual frame tests so I could flip back and forth between them. (I also upped the light amount to something a little more realistic.)

RPE 4/128, No Interpolation, AA 64: HUGE areas of blotch (masked by grain in single stills) that vary wildly between frames, bad lighting fidelity. Useless for animation and grainy.

RPE 1500/200, Interpolated, AA 8: Some subtle large-scale blotch that varies a little between frames, zero grain. Far, far better than the RPE-4 settings posted here. Still problematic for animation unless you're doing a flythrough or have minimal moving elements so you can get away with caching the GI solution.

jrandom
03-09-2011, 12:07 PM
Hey thanks for the scene. Nice one to test GI in ;)

I freakin' love LWCAD. It makes me look like I kinda know what I'm doing, even though my modeling skills are still somewhere between "dead amoeba" and "tranquilized monkey".

Truecoz
03-09-2011, 12:15 PM
You're going to need stuff textured with some sort of bump mapping applied, and animation, then the old settings will not work. As most stuff you render out of Lightwave, I would think you'll want it to be nicely textured.

jrandom
03-09-2011, 12:19 PM
You're going to need stuff textured with some sort of bump mapping applied, and animation, then the old settings will not work. As most stuff you render out of Lightwave, I would think you'll want it to be nicely textured.

Textures will fix the banding, yes, but that's not going to fix the frame-to-frame large-scale lighting blotch differences, and the bounce lighting accuracy (see the areas by the door in my above renders) takes a real hit.

I'm not as worried about small-scale grain since a post-process noise removal pass can help greatly with that. Large-scale lighting blotches are the bane of my existence, and I'd really like to find some faster render settings that get rid of those for scenes where GI caching isn't workable.

3DGFXStudios
03-09-2011, 12:58 PM
I freakin' love LWCAD. It makes me look like I kinda know what I'm doing, even though my modeling skills are still somewhere between "dead amoeba" and "tranquilized monkey".

Haha haha you are modest guy.:bday: It also takes skill to do it in LWCAD :D :thumbsup:

Caching works also pretty nice with moving objects but not deformation, but in this case there's no animation so it's very suited for caching, but I agree it's bit pointless in "GI splotch testing" to turn it on :bday:

agreed .. Textures don't change anything. Only hide errors.

3DGFXStudios
03-09-2011, 12:59 PM
look how happy my post looks with all those smiley's :D

Truecoz
03-09-2011, 12:59 PM
I was just trying to point out that, if you really want to see interpolated and high RPE mess up then you'll need the textures with bumps applied. And animation with a good amount of motion.

funk
03-09-2011, 01:03 PM
I have been doing more tests with some of my old scenes, with low RPE FG and I have found some major problems with luminous polys. I'm getting random blotches and light leaks from frame to frame. I tried setting smaller min/max ps and started getting even stranger artifacts near the light generated from luminous polys (lines etc)

Truecoz, I'm trying my best to make this method work, but it's just not happening.

jrandom
03-09-2011, 01:14 PM
I, too, am experimenting, since it was this thread that introduced me to non-interpolated FG -- I've never played around with it before. I'm trying my damnedest to get it to work for non-cached radiosity suitable for animation but I can't seem to find any settings combination that gets rid of large-scale blotches, light leaks, and good secondary lighting without absolutely killing my render times, and there's still bad grain even at an AA of 64 (which drags my machine to a halt).

I even read Except's guide again. :)

The only thing that seems to combat large-scale light differences between frames is a very high RPE (both interpolated and non-interpolated).

3DGFXStudios
03-09-2011, 01:17 PM
I have been doing more tests with some of my old scenes, with low RPE FG and I have found some major problems with luminous polys. I'm getting random blotches and light leaks from frame to frame. I tried setting smaller min/max ps and started getting even stranger artifacts near the light generated from luminous polys (lines etc)

Truecoz, I'm trying my best to make this method work, but it's just not happening.

That's because I THINK he's wrong ;) (sorry Truecoz)

ps (Truecoz, I still want to drink a beer with you when I'm in the area) :bday:

funk
03-09-2011, 01:39 PM
Just some quick renders of the problems Im seeing with luminous polys and the FG low RPE method. I'm only using aa32 to "save time" here

First off... there are lighting differences between frame 1 and 2 with the same settings. This will cause major flicker in animations

Now the third image, I set the min max ps lower to avoid the flicker, but got strange "line" artifacts

jrandom
03-09-2011, 01:40 PM
Just some quick renders of the problems Im seeing with luminous polys and the FG low RPE method

?

I don't see any attachments. OH NO I HAVE GI BLINDNESS! WHERE'S THE FLOOR? I CAN'T REMEMBER MY DAD!

Edit: Oh, sure, they show up after I post this...

3DGFXStudios
03-09-2011, 01:42 PM
Just some quick renders of the problems Im seeing with luminous polys and the FG low RPE method. I'm only using aa32 to "save time" here

First off... there are lighting differences between frame 1 and 2 with the same settings. This will cause major flicker in animations

Now the third image, I set the min max ps lower to avoid the flicker, but got strange "line" artifacts

You can also use the radiosity overwrite option in the object tab to let the luma polys emit more rays. Very powerfull function that is.... that way you can keep the global value pretty low.

funk
03-09-2011, 01:46 PM
yeah sorry I hit post before I finished writing the post up!

Hieron
03-09-2011, 01:46 PM
My god this thread goes fast! So many posts :)
So, where's that working example....

funk
03-09-2011, 01:47 PM
You can also use the radiosity overwrite option in the object tab to let the luma polys emit more rays. Very powerfull function that is.... that way you can keep the global value pretty low.

Oh thats a great tip! I need to try that out

Hieron
03-09-2011, 01:51 PM
You can also use the radiosity overwrite option in the object tab to let the luma polys emit more rays. Very powerfull function that is.... that way you can keep the global value pretty low.

Can you overwrite the amount emitted, or the sampling on the object that receives them? I thought the latter.

funk
03-09-2011, 02:10 PM
Can you overwrite the amount emitted, or the sampling on the object that receives them? I thought the latter.

I'm just trying some quick tests and I think you are right. It seems to be how many rays the object receives.

3DGFXStudios
03-09-2011, 02:13 PM
Can you overwrite the amount emitted, or the sampling on the object that receives them? I thought the latter.

Good one.... Not sure. But I use it often on glas like semi transparent surfaces because you hardly see the radiosity effect. So that would be the receiving setting then.

Here's also a test with bump and interpolated radiosity. I don't see the problem here?

3DGFXStudios
03-09-2011, 02:25 PM
I'm just trying some quick tests and I think you are right. It seems to be how many rays the object receives.

Nice option for a feature request ;). "Photon emitting object overwrite", I call it ;)

Hieron
03-09-2011, 02:45 PM
I think the LW GI engine works a bit different, though.. not emitting photons I mean.
But I am all for more GI options... perhaps a good time to again push something I would *love* to have:

Per object extended GI settings:
-seperate cache file (optional ofc)
-GI type (MC, FG, interpolated or not, cached or not, animation cache etc)

This would allow great versatility in many scenes, where perhaps a few things move but not all needs to have animation cache on.

geo_n
03-09-2011, 11:31 PM
Using an RPE of 4 is interesting -- the GI pass is very fast but very noisy for indirectly-lit areas. The AA of 64 is awfully slow (especially if you happen to have blurry reflections) but it does result in a grain that can be handled by today's video noise removal filters (see NeatVideo (http://www.neatvideo.com/)). This might bolster Chris's argument a bit.

I can't support his combative demeanor, though. :P

The advantage to the low-RPE non-interpolated approach is you don't get the subtle blotchies that show up in animations. This approach is far less useful for stills. I tried the RPE of 1500 which gave me faster renders but horribly slow GI (darned this blurry-reflection-surfaced floor!). I still haven't found a good balance between the two for something that results in animation-quality GI -- and I'm running on a 12-core hyperthreaded Mac!

(Note: I am still a total noob in Lightwave and have a fondness for computationally-expensive materials.)


We have almost the same conclusion. I have used low rpe for animation too, not archiviz animation which uses static cache.
http://www.newtek.com/forums/showpost.php?p=1121102&postcount=44
But I've never gone below 16 RPE. So I got curious and will do some tests.

If you want animated gi for deforming characters, the most stable for lightwave is backdrop radiosity(some limitation) or montecarlo(noisy,slow).
fg works but not recommend for slow moving objects or high contrast lighting.
Animated gi cache according to lw programmer was never designed for animated deforming gi and only for rigid animated gi. I just read it this year after using lightwave for more than 3 years. That's why I never use it. Its better to use mblur trick with low RPE. The gi gets blurred and smooth in transition.
The best method which I hope kray will have someday is vray's irradiance animation prepass to cache non static scene then use brute force for sec bounce. Its simple to setup like fprime because its brute force and super freaking fast.

http://www.youtube.com/watch?v=maTvI0wqiFI

http://www.youtube.com/watch?v=0r2YRF91y24
:thumbsup:

Truecoz
03-10-2011, 01:24 AM
But I've never gone below 16 RPE. So I got curious and will do some tests.

If you want animated gi for deforming characters, the most stable for lightwave is backdrop radiosity(some limitation) or montecarlo(noisy,slow).
fg works but not recommend for slow moving objects or high contrast lighting.
Animated gi cache according to lw programmer was never designed for animated deforming gi and only for rigid animated gi. I just read it this year after using lightwave for more than 3 years. That's why I never use it. Its better to use mblur trick with low RPE. The gi gets blurred and smooth in transition.
:thumbsup:

Oh! Okay, this may explains everything. I'm working on full CG animated characters and scenes. So this is why I was told to use these settings. But I was also told this by a major Lightwave Guru when it comes to surfacing and rendering. He told me that the render-er has been reconfigured and that he's using these settings for just about everything.

Thanks for that geo_n.:thumbsup:

Okay, Nilis rendering the old way is better for everything except full animation with deformations. But I think both ways can still work. I apologize for saying you were wrong.:bowdown:

However, I do believe my render solution is better for character stuff. Attached is that scene with no blotches using some of my settings... and every lighting situation is different, so these settings will have to be augmented accordingly. ;)

3DGFXStudios
03-10-2011, 03:15 AM
Thanks Chris apology accepted ;)

Your last post makes perfectly sense. I've still just one tip. If you want clean renders with deforming characters you can cache everything that's static with higher settings and render the deformations with the lower settings like you suggested. And render everything in passes. Put everything together in comp and voila. :)

enough :argue: Now it's :beerchug: time!

Hieron
03-10-2011, 04:53 AM
I'm really intrigued by that the renderer would be reconfigured... and it is going to be a "allround" setting.. I'm a bit worried :) Really am... I can't get it to perform as fast on the industrial/arch viz/scientific stuff we do. It is obvious though that the FG non interpolated thing is doing something different than usual, so it might be the case it is new. Perhaps this applies great to deforming characters etc, something I hardly ever do so can't test it out..

So, perhaps yet another good reason to request better "per object GI settings" ! :) :
Imagine static caching many objects in a scene to a file, hitting FG non-interpolated for the moving characters and FG-interpolated animation cached for the path they walk on?!

Wouldn't that be the perfect solution... Doesn't Kray have this? (without the per object GI cache file option)

geo_n
03-10-2011, 05:06 AM
There is a video by cageman about gi in lightwave which explains in detail and better than I could explain. But sadly its also using the old mblur trick. I was hoping that it was using animated gi cache because I thought this was the equivalent of vrays animation irridiance prepass but its not.

Hieron- Kray doesn't have it. What kray has is static cache similar to lw. When doing archiviz type animation with deforming objects you set it to shared gi which basically turn moving objects to be calculated in interpolated and the bg is static cache. There is flicker no matter what you do even on a separate pass. But this is ok for archiviz but not for broadcast show. Kray doesn't have the capability to cache moving objects yet like vray. So the only way is also bruteforce by using uncache gi mode in kray. My transformer z test has kray static cache bg and the robot rendered in separate pass added in post.

kray static cache with shared gi mode.
http://www.youtube.com/watch?v=LL_yqMUpz7s
kray with static bg cache and uncache gi for character
http://www.youtube.com/watch?v=-aibafDbMC0

Hieron
03-10-2011, 05:23 AM
Hey thanks man.. that's some nice examples! But Kray does have per (surface in this case) GI settings right? MC, (non) interpolated etc? It's been a while.....
The current GI engine in LW is nice, simple and a big needed step forward since 8 cycle (or 9.0/9.5 even? damn forgot). But I'd applaud any extra work in it for sure.. Especially for deforms etc.

Your transforming cars, wouldn't they have been doable with animation cache (native) set low, and using AO to get some details out? I guess passes works as well though

geo_n
03-10-2011, 05:25 AM
To differentiate vrays animation irridiance prepass from lw and kray engine, it is actually individual gi maps of the moving objects that vray calls during rendertime. You cant use nth frame to cache with it because its individual gi maps that changes over each frames. Sounds similar to lw animated gi cache but I don't know why the lw programmer didn't mention that lw version is onyl for rigid objects in the lw manual. And lw version is just so slow.
kray is so close to being vray if it can save out those individual gi maps of the moving objects. Maybe newtek should buy kray to speed up development.

Hieron
03-10-2011, 05:43 AM
Perhaps I need to spend a few weeks with Vray one time, Kray never stuck on me.. even though it was fun to work with. Now, if only I had some spare time :)

geo_n
03-10-2011, 07:34 AM
Your transforming cars, wouldn't they have been doable with animation cache (native) set low, and using AO to get some details out? I guess passes works as well though

I didn't try it because of past experience with animated gi cache. If it were just moving rigid objects like boxes or simple objects maybe there's advantage. But I still prefer motionblur trick in lw.
Anyone using animated gi cache btw?
The transformers have morphs and deform on some parts to cheat the transformation just not noticeable. I think animated gi cache would crawl during rendertime. Didn't use ao pass because its a test and lightwave ao is slow because of slow AA. I would use it only for final renders.

Hieron
03-10-2011, 08:44 AM
Ah kk, didn't notice the deforms. I use animated GI cache btw :) Sure has it's uses indeed, just requires a bit of twiddling and commonly I drop any fidelity for speed and use AO for that.

Truecoz
03-10-2011, 09:07 AM
Just one more thing... I was working on an big outdoor forest scene with radiosity, and I couldn't stop getting flickering on the trees, no matter what. High RPE, Interpolated, cached wasn't working. So in this instance setting the RPE to 4 worked better. I guess I would say if you're getting flickering because of using Interpolated and wished there was some other method to get it to work, well now there is. Peace.

I'm ready for that beer, but now it'll have to wait 'til after work. ;)

geo_n
03-10-2011, 09:15 AM
Ah kk, didn't notice the deforms. I use animated GI cache btw :) Sure has it's uses indeed, just requires a bit of twiddling and commonly I drop any fidelity for speed and use AO for that.

Do you have a sample where you used it? video not scene file. In which cases would you use it rather than the normal method like bg rad or mblur trick?

jrandom
03-10-2011, 12:57 PM
How many motion blur passes does it usually take? What sort of GI settings do you generally use for this sort of thing?

Hieron
03-10-2011, 05:10 PM
Just one more thing... I was working on an big outdoor forest scene with radiosity, and I couldn't stop getting flickering on the trees, no matter what. High RPE, Interpolated, cached wasn't working. So in this instance setting the RPE to 4 worked better. I guess I would say if you're getting flickering because of using Interpolated and wished there was some other method to get it to work, well now there is. Peace.

I'm ready for that beer, but now it'll have to wait 'til after work. ;)


Haha, actually, you can find quite some posts here from me, promoting low RPE for trees and grass :P We have a 3x3km scene here with I guess 20k trees, HDI instanced 200k poly's a piece. And I FG interpolation cache the shots actually. But use insane low RPE (and interpolation and caching btw)

http://www.happy-digital.com/instance_tutorials.php

Like that grass tutorial I did, even managed to get grass like that onto the entire 3x3km region, but that was more proof of concept. Didn't render too slow though... (added minutes to a 1280x720 shot)

I put Scene interpolated GI to low rpe like 4-16 and HDI uses those settings and allows caching. Then per normal object I set the Object GI settings to a more normal 200 RPE etc. (cached anyway so splotches don't matter)

Works amazingly well to have low rpe then. However, since you're effectively trying to squeeze millions of poly's into a single pixel sometime, the result may be very noisy but not due to GI, NeatVideo helps alot there.


Do you have a sample where you used it? video not scene file. In which cases would you use it rather than the normal method like bg rad or mblur trick?

I never use bg rad and never tried mblur trick tbh...

geo_n
03-10-2011, 10:52 PM
jrandom - depends on how fast objects are moving but minimum is 6 but set actual blurring to 0 so you don't get mblur at all just the extra passes.

Hieron - Your grass technique is similar to kray instance technique of putting surface override fg to min/max 50. 50 is a low value similar to low RPE. So low value has always worked in render engines it seems.
Do you have sample of use of animated gi cache?

Hieron
03-11-2011, 03:53 AM
Hi Geo, let me look something up later. I'm finishing a project and then will drive to Austria for a weekend holiday straight away :)

About the low rpe:
The reason it works on grass/leaves etc is ofcourse since there is no way you could notice splotches on those. And since for normal interpolated you can cache them well, it renders nice and fast. So not sure how you mean "it always worked" ? Sure, you get crappy *** GI solution that way, but on grass no one notices :) and it can be cached. I wouldn't dare use it on something "clean".

Then again, low GI RPE (100-200), cached, works fine on outdoor scenes as well. Do be wary of numbers like 4 though, it seems that the rays are not spread out too well then and it turns up lighter.

You could check my website, but it has some older animations of ours that do splotch, so that would be a bad example :P

Exception
03-11-2011, 06:36 AM
@Truecoz Read this guide by except who has done quite a bit of research:
http://www.except.nl/lightwave/Radio...de96/index.htm
This guide is based off the old render-er



I'm sorry to disagree with you, Truecoz, but LW10 does not have a 'new' renderer compared to v9.6. It's the same engine with some slight modifications that do not invalidate anything that is in the radiosity guide on our website. This is the main reason I have not updated it, because virtually nothing has changed for the end user, except that the quality is better than it used to be, and it's faster.

That said, there are a few 'tricks' that can work in particular circumstances. First, there's the old MB trick, where you use super-low GI settings but use the old motion blur method to develop passes that get merged. This works for many cases (including animation!), but not well for contact shadows.

The ultra-low RPE non-interpolated way is also a possibility. This also works in certain circumstances, but not in all. It's not that effective for interiors, really. Tesselator experimented with it for quite some time and got some remarkable results for motion graphics. In this case, MC is usually preferrable because it's less error-prone and doesn't have the secondary-ray interpolated pass which FG always has which may lead to flickering in animation.

However it's by no means a robust solution. In many cases it doesn't work: slow, ugly or other issues. It's mostly a nice one to try when doing animations, for stills it's not very appropriate in my experience. The Gi guide recommends a method that is robust. There's a few other freaky ways to make things work. If it works for you then it works for you. All lighting in GI is fake anyway, so the old spinning light trick also still stands as a possible avenue.

Truecoz
03-11-2011, 06:08 PM
I'm sorry to disagree with you, Truecoz, but LW10 does not have a 'new' renderer compared to v9.6.

Yes, I'm sorry I just meant it has reconfigured. And compared to 9.3. It's absolutely not a new render-er. My bad, thanks for clarifying.

Hieron
04-12-2011, 04:41 PM
Found a very good use for this non-interpolated mode while rendering some luminous TurbulenceFD "liquid". Which is I guess a truly worst case scenario for anything interpolated. It's currently rendering overnight but will put it up tomorrow. Thanks for bringing this mode to light for me Truecoz. It isn't a mode that works best for all (or even most?) standard situations. But it sure could use the exposure it got here.

Stuff like this should be a tad more easy to find.

Ow, and Newtek should allow the option to set this mode to objects in the per object GI settings panel. It would be really powerfull to make some objects interpolated/cached, and others not.

geo_n
04-13-2011, 12:23 AM
Hi Geo, let me look something up later. I'm finishing a project and then will drive to Austria for a weekend holiday straight away :)

About the low rpe:
The reason it works on grass/leaves etc is ofcourse since there is no way you could notice splotches on those. And since for normal interpolated you can cache them well, it renders nice and fast. So not sure how you mean "it always worked" ? Sure, you get crappy *** GI solution that way, but on grass no one notices :) and it can be cached. I wouldn't dare use it on something "clean".


It works in other renderers like vray which apparently vray users were using this same lw technique of low gi settings to smooth it out even before lightwave had the new lw 9.6 renderer.
I asked old 3dmax users and they said they used it before when vray didn't have the irridiance animation prepass. I used the same concept from vray by using low rpe in lw and it works.

Hieron
04-13-2011, 04:03 AM
Hey Geo, I read your "always" as "in all GI situations" instead of "in this situation for a long time". Sure, good to know it worked similarly well long before!

I hadn't read that "trick" of using really low rpe on grass before, just saw some mention about slow GI speed (not this thread) and started to mention how to combat that (and use NeatVideo to combat noise for similar situations, other causes). In the case of LW native and HDI, you need to set global settings to low rpe, which is a bit confusing at first I guess. Only reason I managed to get fast massive amounts of grass and trees so thought I should share.

What I meant to show was:
94346

No interpolated technique worked well for me on it, then again, "new" FG non-interpolated or MC non-interpolated were sort of similar. The "error" when using just 4 RPE was obvious though in both (everything comes out way too bright) so I used 12+ when it is more in line with what it should be.

Ideally I would have wanted to use anim cached MC without volume to cache sample points into the mesh, a few are enough here anyway. But whichever way I try, if then the volume is turned on the result is bizarre. Too bad, seemed like a good way to go about it here.

Tobian
04-13-2011, 09:07 AM
I got really excited about increasing the number of rays for luminous geometry there, till I read on, and it doesn't work. I find it quite annoying to use LW for small-ish luminous but bright polygons, it's virtually impossible to get stable radiosity out of it without very slow renders. LW seriously could use light-emitting geometry / photon cache / point cloud (yes please!) or some other solution to stop horrible blotching artefacts, of which quite a few techniques exist now, but aren't possible at all in LW.

Tricks which rely on high AA are a no-go if you have heavy-multisample materials, such as blurry refractions/reflections.

As a note though, I lower my angular tolerance a lot as it gives much better spread of light and less blotching, but in my last render I went as high as 2048 samples and it was still blotching round bright parts.

I think a lot of the problem with radiosity is the sampling pattern of the rays uses a voroni-cell like distribution across the screen. In a point cloud scenario it would gather evenly in corners, based on the mesh. With increased sampling round luminous geometry you'd get them behaving more like 'lights' and do shadowing better, and have less blotching, as well as (in theory) speeding up blurry reflections too.

One day we will get these advanced features in LW.. one day! :)

Well once people got past their gigantic ego's we got some good info going on in this thread :D

and if it used a more complex importance sampling schema

Hieron
04-13-2011, 11:24 AM
You are probably referring to me as well. But besides the gigantic ego I guess, anything related to "GI" and "better","faster", "new", "doing it wrong" is going to get me gigantically interested.

And sadly, nothing of that type of info was eventually presented in this thread. There was quite a bit of posting though.

I do agree with your point, that situation is very hard to tackle. non-interpolated with alot of noise -> NeatVideo is the best option I can think of. That tool can get rid of alot of noise.. Hopefully, improvements like that will make it into LW10.. 11?

Tobian
04-13-2011, 02:20 PM
Erm radiosity blotching is not 'noise' and you won't ever get rid of that with de-noisers...

Sorry I tailed off on the end of that sentence. Just had a busy day! :)

Hieron
04-13-2011, 02:27 PM
Hehe you may have been right! :)

Nah splotches will never go for sure, but I said "non-interpolated". That is fixed decently well with NeatVideo, right... Splotches are a worst nightmare indeed... and non-interpolated comes next so a better solution would be nice...