PDA

View Full Version : FG better than FG+GI for indoor scene


royterr
08-05-2009, 05:08 PM
here's a very simple example of a mia_physical_sun and sky setup (http://sor.typepad.com/.a/6a00d8350600cb53ef0120a4c9b02a970b-800wi) with one portal light.
the sun and the portal light are emitting photons. When i compare the pure FG solution (with multibounce) to the FG+GI solution, the second one seems "flat" and unconvincing. Why is that, isn't GI supposed to make indoor scenes more convincing together with FG? how come the FG solution alone turns out better?


http://sor.typepad.com/.a/6a00d8350600cb53ef0120a4c9b02a970b-800wi

Cheesestraws
08-05-2009, 05:54 PM
The FG only one is just noisy/splotchy, add some grain in post to the GI one if you like that look.

InfernalDarkness
08-05-2009, 08:13 PM
Also, the pure FG render has more contrast. You can fix this with your framebuffer for GI, or in post.

JohnPetrucci
08-05-2009, 10:21 PM
The Third one seems the best one IMO

Mrguy
08-05-2009, 10:38 PM
Well I can see what royterr is talking about.

I've up'd the contrast in the gi+fg image and the image is still relatively flat. There's bounce light in the FG scene that just isn't in the GI + FG. The FG scene is getting better bounce light and better cast shadows.

Did you change any FG setting before you did it with GI?

Im relatively new on lighting intersiors with a portal light, learned it and couldn't get it work. But after seeing this I'll give it another go, I'll post tonight.

Also can you have more then one protal light for a room with more then one window?

royterr
08-05-2009, 11:09 PM
Well I can see what royterr is talking about.

I've up'd the contrast in the gi+fg image and the image is still relatively flat. There's bounce light in the FG scene that just isn't in the GI + FG. The FG scene is getting better bounce light and better cast shadows.

Did you change any FG setting before you did it with GI?

Im relatively new on lighting intersiors with a portal light, learned it and couldn't get it work. But after seeing this I'll give it another go, I'll post tonight.

Also can you have more then one protal light for a room with more then one window?

no didn't change the Fg settings.

yes you could have as many portal light as you want.

Aikiman
08-06-2009, 03:46 AM
The last one might lack the intensity but it seems to have more atmosphere, more particles in space floating around. Id try upping the photon intensity in the GI solution to get the brightness back.

royterr
08-06-2009, 04:42 AM
The last one might lack the intensity but it seems to have more atmosphere, more particles in space floating around. Id try upping the photon intensity in the GI solution to get the brightness back.

you are right the 3rd picture shows an interesting "atmosphere":
-note that the GI is driven by the portal light, so increasing their power wont be "physically correct"
-besides the corners/edges looks dark, which is a bit weird

Mrguy
08-07-2009, 12:25 AM
Didn't get a chance last night... So work is going slow right now and Im just re-creating the scene. What color are your walls default grey or white?

Mrguy
08-07-2009, 01:09 AM
Dang man after checking out your http://sor.typepad.com seems like you know alot about lighting and/or are testing a lot; which is really cool.

I hope I can help cause I'm trying to light a room too.

Between FG and FG+GI.. FG+GI is just a little bit darker.. but for some reason GI comes out like crud.

I've tried the portal light before with not the greatest success so I searched for other ways to make it and came across this link. http://www.vfxtalk.com/forum/maya-2008-mental-ray-portal-t13992.html?amp;t=13992

Besides the two FG bounce rays, and about 20% off white for the bounce color, everything is default. The rest of the noise is just the amount of samples.

http://i6.photobucket.com/albums/y242/Mrguy45/FGandGITest.jpg?t=1249606992

http://i6.photobucket.com/albums/y242/Mrguy45/portal_shader_network.jpg?t=1249606993

Sybexmed
08-07-2009, 11:04 PM
Royterr, ive been looking at your demo reel, and i must say that is some nice stuff. do you have some info on the video game at the start? how was it created etc?

royterr
08-10-2009, 03:56 PM
Royterr, ive been looking at your demo reel, and i must say that is some nice stuff. do you have some info on the video game at the start? how was it created etc?

what you have seen was a fake game footage for presentation purposes.
a year ago we started building a functional prototype but the project has been put on hold for a while now as i am working on other games.


cheers

romeu
08-13-2009, 08:19 PM
im not sure, but it could also be that your GI solution is smoothing to much the result,

you could try more photons with a lower radius so you get a more acurate gi solution (also, unfornutatly maybe noisyer)

also, one question, on your fg settings, do you have photon lookup turned on?

spiralof5
08-13-2009, 09:16 PM
You know, I'm having the same exact issue and I was actually going to post about it. I still might but I'd hate to be redundant.

I'm actually getting crappier results using GI + FG. I find that my GI results are getting very noisey but the FG alone is a lot smoother, brighter, and FEELS more accurate.

I've actually went back and am planning on doing Floze's Final Gather lighting tutorial. Upon reading through, I noticed he does NOT use GI at all. His renders look real nice.

So I guess my question reiterates what royter's does: Do you even need GI for interior scenes? What is the point if you are relentlessly rendering and never acheiving a great result? Do I even need GI?

I get very confused with rendering interiors because everything I try, importons, GI, FG, and if I feel like waiting hours, irradiance particles. I've read a lot of tutorials and have gone through some Digital Tutors and Gnomon videos on rendering with FG. The problem is that none of it really relates specifically to interiors, which is where I find I have the most problems.

royterr
08-13-2009, 10:37 PM
im not sure, but it could also be that your GI solution is smoothing to much the result,

you could try more photons with a lower radius so you get a more acurate gi solution (also, unfornutatly maybe noisyer)

also, one question, on your fg settings, do you have photon lookup turned on?

yes


Do you even need GI for interior scenes? What is the point if you are relentlessly rendering and never acheiving a great result? Do I even need GI?

I get very confused with rendering interiors because everything I try, importons, GI, FG, and if I feel like waiting hours, irradiance particles. I've read a lot of tutorials and have gone through some Digital Tutors and Gnomon videos on rendering with FG. The problem is that none of it really relates specifically to interiors, which is where I find I have the most problems.

importons are the most accurate, very similar to FG and faster.But be aware that they are still buggy and not ready for a production use (as if MR4Maya was ready, anyways i will try to stay positive)

Now GI+FG is supposed to be more physicaly accurate then FG alone.
if you ask Jooja from HighEnd3D apparently the contrated FG solution is a bit artificial and GI+FG is the most accurate one, you just have to add some contrast in post.
I am just not convinced of that, and no "proGI" user has prooved me with some visuals the usefulness of GI for interiors.

*beside the light solution, notice how my beveled coners look weird/dark in the FG/GI solution and natural in the FG one....still no answers.

Arcon
08-13-2009, 11:17 PM
honestly when i hear people in CG talk of "physically accurate" i just have to laugh... what's physically accurate about the light's intensity or position outside the window...? the physically accurate one is always the one which looks more real & sells the image, doesn't matter how it got there ;)

spiralof5
08-14-2009, 02:48 AM
Well, sometimes I think that is correct and sometimes not. I guess when doing photo-real, you have to ask yourself, is photo-real actually photo-real?

Aside from that I realize that FG and GI are supposed to be used together. I just am finding more often that people are abandoning the GI solution all together when using Final Gather for interiors.

As far as the renders being darker in the GI solution, I thought of something today. My concept of brightly lit interiors are usually wrong. They are usually a lot brighter in my mind than they are in real life. I opened all my blinds today and let the light in. I have two big windows in a small apartment. I was actually pretty surprised at how dark the inside looked as opposed to how bright I thought it would be.

InfernalDarkness
08-14-2009, 07:27 PM
I am just not convinced of that, and no "proGI" user has prooved me with some visuals the usefulness of GI for interiors.

Ever seen caustics? Photorealistic glass? Liquids?


I was actually pretty surprised at how dark the inside looked as opposed to how bright I thought it would be.

Indeed, this statement kinda outlines the two schools regarding "photoreal" versus "real". In most "photoreal" renders, and indeed in photographs too, you'll see a blown-out white area where windows are. This is typical of Vray renders in general, and I suppose MR too where people approach that quality.

This is typical of photography too, but certainly NOT typical of "real life", or what one sees with ones eyes. Only when transitioning from darkness to light would you see this bloom effect, and so long as your eyes work properly, it disappears very fast. We don't even notice it.

In my experience, when rendering towards photorealism/realism, my clients react much better to the "photorealistic" renders with blown-out window lighting. Even though you lose a lot of detail, I think this is what their minds are trained to see on paper or print, or on the web for example. When I show people my "realistic" renders, they don't react nearly as positively. Kinda annoys me, because when I see "photorealistic" renders the first thing I look for is how crappy their windows look, and how unrealistic the environment outside looks.

But I have to agree :

the physically accurate one is always the one which looks more real & sells the image, doesn't matter how it got there

I don't agree with your "physically accurate" part, but maybe "monetarily accurate" would be a good paraphrase? (grins)


I use FG and GI both almost always, because the "physically accurate" materials in MR react much better to photons. Granite countertops, glassware, windows panes, etc. Shower glass doors and such demand GI/caustics. FG alone doesn't do them justice really.

But when time is a factor (when isn't it a factor?) using just FG works great, and of course it's much smoother to behold!

royterr
08-14-2009, 09:12 PM
Ever seen caustics? Photorealistic glass? Liquids?

i was not talking about caustics photons only normal photons (i do agree that caustics are indispensable in these kind of situations: liquiid/water/glasse...).
But i have not seen 1 interior render in witch the GI/FG solution is better than FG solution (again no caustics just normal photons), if you have an exemple plz share it with us.

InfernalDarkness
08-14-2009, 09:20 PM
Try here (http://forums.cgsociety.org/forumdisplay.php?f=132). In the search, type "mental ray", and it should help. Or perhaps "interior mental" would narrow the search down pretty well. Even "interior mental GI".

Some quick examples:

http://forums.cgsociety.org/showthread.php?f=132&t=784250&highlight=interior+mental

http://forums.cgsociety.org/showthread.php?f=132&t=781677&highlight=interior+mental

http://forums.cgsociety.org/showthread.php?f=132&t=772512&highlight=interior+mental

http://forums.cgsociety.org/showthread.php?f=132&t=771772&highlight=interior+mental


Then if you search a bit more, you'll find some that state outright they only used FG. Tend toward the bland, even if they are awesome renders, the dynamic range really isn't there:

http://forums.cgsociety.org/showthread.php?f=132&t=752283&highlight=interior+mental

A prime example of the "photorealistic" light blowout, but still a wonderful render:

http://forums.cgsociety.org/showthread.php?f=132&t=704873&highlight=interior+mental

3d165p05
08-15-2009, 04:17 PM
So, probably it all depends on the scene... Don't know :)

I did some test with my current work in progress. And here are results.

1. just FG
FG settings
accuracy: 100
point Density: 2
point Interpolation: 25
secondary Diffuse Bounces: 3

render time: 45:24 (crappy one core AMD (!?), crappy crippled gforce 5700LE)

http://i385.photobucket.com/albums/oo291/3dzanimacija/testing/testing_00.jpg

2. FG + GI
FG settings
accuracy: 50
point Density: 1
point Interpolation: 15
secondary Diffuse Bounces: 1 (when using GI in conjuction with FG, than FG secondary bounces defaults to 1 anyway)

GI settings
accuracy 500 (default)

render time: 20:08 (crappy one core AMD (!?), crappy crippled gforce 5700LE)

http://i385.photobucket.com/albums/oo291/3dzanimacija/testing/testing_01.jpg

3. FG + GI (trying to match brightness of the only-FG-solution, I bumped up Primary FG Scale from 1.4 to 1.8 value, and I raised FG settings a bit for better result)

FG settings
accuracy: 100
point Density: 1
point Interpolation: 20
secondary Diffuse Bounces: 1

GI settings
accuracy: 500 (default)

render time: 32:37 (crappy one core AMD (!?), crappy crippled gforce 5700LE)

http://i385.photobucket.com/albums/oo291/3dzanimacija/testing/testing_02.jpg



Personaly, I like FG+GI solution more. Esspecialy walls around windows and door. And note that, because of lower FG settings in that case, render times decrease. Also, I have to play a little bit with the render settings to get better results in all of renderings, but this is WIP - I'll do that in final render.
opinions?

cheers

edit: It's all from renderview with mia_exposure_simple, no postprocess

royterr
08-16-2009, 04:43 AM
opinions?

actually in your case the difference is very subtle.
the Gi/FG solution is a bit more convincing to me.

i dont agree that the FG/GI formula always works better for interiors. This is not a general rule, it is very scene dependent i think, specially if you have beveled edges (instead of perfect unrealistic sharp edges where the floor and the walls meet for exemple). The proof for this is great interiors with no GI:
http://marlas.cgsociety.org/gallery/

spiralof5
08-16-2009, 05:17 AM
Ever seen caustics? Photorealistic glass? Liquids?

Yes, but these can be separated in passes. The GI would be only to drive the caustics.

As far as exposure is concerned, who cares? That doesn't really have much to do with this, as far as I'm concerned, because exposure can and is usually controlled in post.

You do have to admit that the GI+FG solution really does cramp and dull the effects of using a physically accurate sun and sky solution.

The thread on Importons and IP is http://forums.cgsociety.org/showthread.php?f=87&t=659021

I dunno, use the tools you have to make the best picture possible. Do we need a thread on that? haha. My problems are more technical, and most likely come from a lack of understanding. I will say I'm enjoying Importons very much more than just GI alone.

InfernalDarkness
08-16-2009, 08:44 AM
As far as exposure is concerned, who cares? That doesn't really have much to do with this, as far as I'm concerned, because exposure can and is usually controlled in post.

I would say that given the current atmosphere and demand for a proper linear workflow in Maya, which was the #1 requested update for 2010 (followed closely by working passes), everyone cares. If you don't, that's your business of course. But everyone else does.

That's the stinker about Maya as well as its beauty. It's such a broad program, with so much room for expansion through scripting, that most users don't ever touch 100% of the features it offers, or even 50%.

3d165p05
08-16-2009, 10:13 AM
actually in your case the difference is very subtle.
the Gi/FG solution is a bit more convincing to me.
[/url]

I agree.
And I was surprised by such subtle difference.


i dont agree that the FG/GI formula always works better for interiors. This is not a general rule, it is very scene dependent i think[/url]

Again, I agree, and I stated that in my post ;)

royterr
08-30-2009, 06:16 AM
Why does the FG solution resolves the beveled corners/edges much better then the FG/GI solution where they loock black and weird?
http://sor.typepad.com/.a/6a00d8350600cb53ef0120a4c9b02a970b-800wi

InfernalDarkness
08-31-2009, 04:48 AM
Honestly, I can't find the GI only scene you're using to be a decent comparison. Unless you're equalizing rendertimes, and saying that at x rendertime, this is how the FG looks vs. the GI only render.

I think ramping up your GI configuration there could make a huge difference, is all I'm saying. If it was a clean GI render it would be a better comparison, and might be more worthwhile as a solution...

grafmishurov
08-31-2009, 12:04 PM
Multibounce FG increases rendering time and color bleeding for indoor scenes. GI+FG better. :wip:

grafmishurov
08-31-2009, 12:14 PM
Why does the FG solution resolves the beveled corners/edges much better then the FG/GI solution where they loock black and weird?

You can tune interpolation settings for FG rays, and use oldschool radius control, or use ambient occlusion for small details and corners.

CGTalk Moderation
08-31-2009, 12:14 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.