View Full Version : GI, replicating this image
matmonkfish 08-22-2006, 03:01 PM hi,
ive am pretty familiar with mental ray with final gather and HDR and stuff. not so much with GI
What would be the best technique for going about replicating this?
just some helpful hints, any advice would be appreciated
| |
YourDaftPunk
08-23-2006, 03:18 AM
Ok, you're gonna have to read some threads, but basically after you have your scene built:
1) Materials: Use a DGS material for the ball and a mib_illum_lambert (or DGS with Diffuse only set) for the room walls. With the DGS ball, set the Diffuse to a middle value red. Specular (mirror-like reflection) needs to be set low or to black. Glossy needs a medium value (this is the brightness of the blurry reflection). Shiny is the amount of blur, so set it low (low is more blur, so 5-15 should do). Go to each material's shading group (by graphing output connections in the hypershade) and create a photon shader. Note: for DGS you actually don't have to create a new DGS Photon Shader- just drag the original DGS node with the middle mouse into the PhotonShader slot of the Shader Group to make the connection.
(If want to use Maya materials, you need to read Floze's article, "DGS Exploited," for accurate lighting calculations with native materials. Also, Maya shaders don't need a photon shader attached which is nice.)
2) Lighting: Create a Maya point light. Go to the Mental Ray tab and turn on Area Light. Your light will magically transform to a square in the scene (don't use Maya's default area light in Maya 7). Pump up max samples to 8x8 or so. Then go to the Custom Shaders tab and create a Light Shader. Select the Physical Light shader to create it. Intensity is now controlled by the color ('V' of HSV actually) of the Physical Light. I set it to about 4000. The light has accurate falloff and now we have a number to plug into Photon Intensity (4000) when we do the GI bit. The regular light settings are ignored by mental ray now- make all your changes through the light shader or the Mental Ray tab Area Light settings.
3) GI: Switch on Global Illumination in the Render Globals. The Accuracy setting, counter-intuitively, blurs the result with higher values- higher accuracy, the smoother the photons become. You'll probably be in the range of 500-2000 with the higher accuracy (more blur!) reserved for those renders without Final Gather turned on. The Radius setting is really the critical one- too small and you get an image speckled with photons. Too big and you get a smooth but inaccurate result with artifacts like light bleeding through walls. See my scene for settings or read some of the attached threads at bottom.
4) Lighting Cont: Go back to your light and check Emit Photons. Plug in the intensity number from your Physical Light Shader (V) and set the number of photons to 100,000 (minimum). Final renders will have at least 500,000 photons bouncing around (maybe much more if you are not using Final Gather- I've tried 6 million without issue).
5) Render. The more lighting bounces and the more photons in the scene, the better. By default you have 5 bounces (Max Photon Depth, Max Reflection Photon, and Max Refraction Photon are all set to 5). The example provided doesn't really show the power of GI. You could do it just as good with Multibounce FG. Still, it's fine for testing.
Tip: Don't let your photons fly off into space- its wasteful. That's why people use containing geometry.
Tip: Turn of Rebuild Photon Map if you have just rendered a bunch of photons and you only want to play with Radius and Accuracy settings. Change the light setting though and you'll need to render a new photon map.
Tip: After you get your GI looking good and you add in Final Gather, half the quality of the GI settings. Otherwise your renders will take forever. I find that getting the right balance between GI and FG is by far the hardest part of GI rendering. Takes patience.
The Threads You Must Read:
Taking it slow with GI, Caustics, Final Gather (beginners) (http://forums.cgsociety.org/showthread.php?t=289028)
DGS exploited (http://forums.cgsociety.org/showthread.php?t=228118)
physical_light exploited (http://forums.cgsociety.org/showthread.php?t=293775)
VRay-like interior renders with mental ray??? (http://forums.cgsociety.org/showthread.php?t=190232)
M~R to MR: Getting the Maxwell look in Mental Ray (http://forums.cgsociety.org/showthread.php?t=364576)
Blurred Reflections??? (http://forums.cgsociety.org/showthread.php?t=252304)
http://studentpages.scad.edu/%7Eslipow20/gi_demo.jpg
gi_room.zip (http://www.lickmypixel.com/demo/gi_room.zip)
-shawn
P.S.: You can make the demo scene much faster by reading the blurry reflections thread (that's what is really slowing it down!), tweaking the BSP settings and adjusting the ratio of GI to FG. Have fun.
Cartesius
08-23-2006, 07:42 AM
The Threads You Must Read:
Taking it slow with GI, Caustics, Final Gather (beginners) (http://showthread.php?t=289028)
DGS exploited (http://showthread.php?t=228118)
physical_light exploited (http://showthread.php?t=293775)
VRay-like interior renders with mental ray??? (http://showthread.php?t=190232)
M~R to MR: Getting the Maxwell look in Mental Ray (http://showthread.php?t=364576)
Blurred Reflections??? (http://showthread.php?t=252304)
Your links aren't working.
/Anders
David_Web
08-23-2006, 10:43 AM
Taking it slow with GI, Caustics, Final Gather (beginners) (http://forums.cgsociety.org/showthread.php?t=289028)
DGS exploited (http://forums.cgsociety.org/showthread.php?t=228118)
physical_light exploited (http://forums.cgsociety.org/showthread.php?t=293775)
VRay-like interior renders with mental ray??? (http://forums.cgsociety.org/showthread.php?t=190232)
M~R to MR: Getting the Maxwell look in Mental Ray (http://forums.cgsociety.org/showthread.php?t=364576)
Blurred Reflections??? (http://forums.cgsociety.org/showthread.php?t=252304)
Correct links. There you go.
Cartesius
08-23-2006, 01:18 PM
Yep, working now, thanks!
/Anders
Rick Davidson
11-21-2006, 01:58 AM
Hey Shawn, just a note to say great post, thanks mate, great to have a file to look at. Been tweaking a file forever without luck. Will take a look at those threads but maybe someone can take a look at this and tell me what my error is here:
http://www.rickdavidson.com/extra/gi.jpg
Doesn't matter how I tweak it, I can't seem to reduce the intensity. I have one direct light with .6 intensity, emitting default photons etc. Arrgh.
I also would like to use Image based lighting (HDRI) to get nice relections on a car but I need quick renders. Is global illumination worth the extra render time or should I give up? Anyone?
YourDaftPunk
11-21-2006, 03:58 AM
Doesn't matter how I tweak it, I can't seem to reduce the intensity. I have one direct light with .6 intensity, emitting default photons etc. Arrgh.
Ok, with photons, you want to emit light from a fixed source like a spot light or area light. Directional lights will waste photons as mental ray won't know where to spawn them- just the angle as they cruise through the scene from oblivion back into oblivion without hitting your geometry. This also means that your scene should probably be contained in a sphere or cube set to the sky color to reflect the photons you emit back into the scene. You can turn primary visibility off it you want, but you need it there.
Second, your photons luminance should be controlled through photon intensity. You can create an accurate photon intensity by creating a light shader (choose 'physical_light') and setting your photon color to the color in the light shader and your photon intensity to the 'v' value of the physical_light (I believe the example scene is setup like that). The other option is to create an area light and set the decay to quadratic. Use the standard maya color and intensity for the light, and then set the photon color to the same value and the intensity to maya intensity*pi.
I also would like to use Image based lighting (HDRI) to get nice relections on a car but I need quick renders. Is global illumination worth the extra render time or should I give up? Anyone?
GI is way over kill. Check out Brendan McCaffrey's making of the Peugeot 907 article on his site: http://www.bmcaff.com/
bmcaff
11-21-2006, 10:03 AM
I also would like to use Image based lighting (HDRI) to get nice relections on a car but I need quick renders. Is global illumination worth the extra render time or should I give up? Anyone?
I definitely aggree with Shawn, GI is fairly pointless for object/exterior renders Like this. There is more than enough direct and indirect light from FG or multibounce to light an exterior car scene. Mainly due to the level of specularity you get with a car. You want to use GI where there is a real need to bounce light around to illuminate a more diffuse scene.. such as the interior of a room.
Rick Davidson
11-21-2006, 11:29 AM
Wow thanks guys really appreciate your advice, tutorials can only teach you so much :) Will have a play. Bye bye GI.
This is where I'm at now - just a standard area light with a blinn floor surface... Working ok I guess, maybe needs a warm / cool light combo.
http://www.rickdavidson.com/extra/area_light.jpg
Is an image based light the best way to get decent reflections or is a basic sphere faster on the render? That's gonna wreck my whole setup isn't it... sigh...
Ash-Man
11-21-2006, 02:03 PM
MY GOD DraftPunkLOL that is well written and Im sure it took some time :):thumbsup:
Rick Davidson
11-23-2006, 03:07 AM
hey guys, trying to optimise this scene now - just wondering if Final Gather renders faster in an enclosed space or if an open scene is fine? Just GI photons need an enclosure?
YourDaftPunk
11-23-2006, 03:45 AM
Funny you should ask that. Final Gather by itself works fastest with an ibl dome or white camera environment- the fg samples have a very uniform intensity. Since FG works from the camera forward, you don't have to worry about the light problem of rays being emitted from nothing into nothing.
In an enclosed space like a room you might have a bright window and a dark wall - the samples have a big range of intensities and you end up needing more fg rays and filtering of 2-4. Also, FG falloff doesn't look as natural in a room as it does in an outdoor lighting situation.
-shawn
Rick Davidson
11-23-2006, 08:21 AM
Thanks again Shawn, wish I had your knowhow ;) Doing all this for a TVC due very soon so your help is awesome. Thanks again.
MasterZap
11-23-2006, 10:43 AM
In an enclosed space like a room you might have a bright window and a dark wall - the samples have a big range of intensities and you end up needing more fg rays and filtering of 2-4. Also, FG falloff doesn't look as natural in a room as it does in an outdoor lighting situation.
Note; Don't misunderstand the FG filter as a "smoothing" filter per se... it is actually a "ray discarding" filter. Basically, the filter is (simply put) the number of "Brightest" FG rays that are thrown away.
The idea is that if most FG rays hit some "medium intensity" and one ray hits by accident something really bright, that ray is thrown out to make the fg overall "smoother" (but also, inherently, darker). Therefore only fg filter = 0 is a completely "unbiased" light level, but is more prone to "white dots" where you get one lucky fg ray hitting something really really bright.
In ray 3.5 you can often subdue those bright points with increasing the number of FG interpolation points (which does much more of what people expect of a "filter" to do, i.e. smooths out the existing light).
Just a piece of info....
/Z
YourDaftPunk
11-23-2006, 09:51 PM
Note; Don't misunderstand the FG filter as a "smoothing" filter per se... it is actually a "ray discarding" filter. Basically, the filter is (simply put) the number of "Brightest" FG rays that are thrown away.
I actually did have the wrong idea about it (and I'm sure other people do too), so thanks :thumbsup: I avoided filtering when possible because I could tell I was loosing accuracy.
In ray 3.5 you can often subdue those bright points with increasing the number of FG interpolation points (which does much more of what people expect of a "filter" to do, i.e. smooths out the existing light).
Would those would be the "final gather points" exposed in miDefaultOptions, set to 20 by default? What are the costs to render time by increasing the interpolation- is it linear?
-shawn
MasterZap
11-24-2006, 08:38 AM
I actually did have the wrong idea about it (and I'm sure other people do too), so thanks :thumbsup: I avoided filtering when possible because I could tell I was loosing accuracy.
Yes, filter=0 is actually the "unbiased" result.
Would those would be the "final gather points" exposed in miDefaultOptions, set to 20 by default? What are the costs to render time by increasing the interpolation- is it linear?
I think that's the one. However I think you may need to use the "automatic" FG mode for it to work "as one would expect".
Cost is negligeble, since it's just a matter of interpolating over more data. So the render time for "20" or "100" is nearly the same, but "100" is much more smoothed. But it can of course be OVER-smoothed (losing details), so turning it up isn't a "solve everything" solution in any way.
/Z
Lamster
11-24-2006, 09:02 AM
Guess we'll have to wait till Maya properly implements the FG Auto and Multi mode for us to be able to use these nifty FG interpolation points then..
Terence
In ray 3.5 you can often subdue those bright points with increasing the number of FG interpolation points (which does much more of what people expect of a "filter" to do, i.e. smooths out the existing light).
Just a piece of info....
/Z
What exacty do you mean with ''increasing the number of FG points'', you mean FG accuracy?
Or is there a new extra /different option like the new 'photon merging' for GI in the new 3.5 mr(I dont have access yet)?
Cheers
CGTalk Moderation
11-27-2006, 01:47 AM
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.
vBulletin v3.0.5, Copyright ©2000-2009, Jelsoft Enterprises Ltd.