Not sure I follow. What do you mean?
Maybe he is talking about your technique? I’m not sure what he wants to say either…
Highly doubt it, from what I know Pixar uses real raytracing, mainly nurbs ifrc, never saw them talking about voxels(well, read a talk or two about using it for atmospheric , but that’s about it).
Maybe he meant about using albedo?
Test map is brillant. everything works.
But when i’m trying to create my own indoor map scene.
Everything is always too bright.
By the way scene is being created from a blank map with nothing.
With black boxes around the white walls.
No lights so there’s not even a possibility of leaking.
When i add lights not much difference.
I gave the latest build a try with a similar scene setup - exactly the same problem. Everything was super bright, and all the corners have a cyan glow. I’ve tried zeroing all ambient lighting options in the AHR settings, but to no avail.
Ok, that’s probably a bounds. Try changing the lost ray color, if you see it gone then is a bounds. When you are outside the bounds all rays are “lost” at start, so its like a constant ambient term in that case. Did you created another post pro vlume and set Defines AHR Bounds?
i created two Postprocess like the video you posted.
One at 3,000(xyz) and another at 2,000 (xyz) with ‘set AHR Bounds’ enabled.
My lost ray color was set to Black.
Playing with it has no effect, and changing the bounds dimensions doesn’t solve the problem.
I should add that i have successfully made several open scene maps to test and it works perfectly. just not the closed scene it seems.
Hmm odd. Thinnk you can upload the problematic project somewhere so I can give it a look?
Here is the project. giMap is the name of the map.
https://mega.nz/#!QgdgzJyB!L2LN7Wae0g2OvdLM-9iLxdRPlSfxPgXR0Id6wfm4CaA
Will test it and tell you in a minute
Adding, as you call it, “a bit of the albedo” to irradiance. Or rather, as I did it while messing around with some SVOGI stuff, made the ambient light affect the irradiance GI output rather than the normal albedo. Tried to convince some people on Gamedev.net that it was a good idea, but they didn’t go for it for whatever reason. Worked well when I tried it, and when I looked it up later I found out it was something Pixar had done for a lot of their movies.
Not only does it make “non lit” areas show up in specular reflections, but it gives a nice and frankly realistic color bleed. Of course, the way your doing it is straight adding albedo to the irradiance, which I imagine would make the secondary color bleed a little too strong and saturated. For a more realistic look you could try shading your voxel albedo as if it was being lit by a greyish ambient light (user selectable ambient light) and then having that be added to the irradiance output.
Save as reference, nice tool!
I did a few tweaks, is more like it?
I just played a bit with the settings. Will upload the modified project later. kind of scenes are where at least 2 bounces are required, look strange without them. Also, 5 rays here are too little. When I finish the re-work of the algorithm, the number of rays will be tweakable parameter, so that will make it much easier to control quality/performance. Quite confident that kind of scenes with multiple bounces and about 8 rays should look really good.
Frenetic Pony, that kind of scenes is where adding the ambient light break stuff, with that corners get a white light, that’s just appearing magically.
EDIT : Can you give me some link to the disney stuff? Would love to read it.
Will add a slider to change the amount of ambient in the grid, as it works quite good on some scenes, but it breaks in others
For comparison, here’s a pic with and without the ambient light (exaggerated to show the problem)
Looks better. Well i was trying to replicate
Realistic Render Demo with Enlighten
Opening the map ‘room with no lights’ led to a very bright room instead of a pitch black room, even after deleting all the lights.
That’s why i attempted to re-create the room in a new project from scratch thinking that would solve the problem.
Hmm that’s odd. Not sure why it was so bright.
Unfortunately, I don’t think my thing will work with your trace distance, at least not naively. I forgot to mention the way I did it and to make it look good was to brute force the trace distance all the way to the end of Sponza, so that everything was lit by global illumination. What you’re seeing is the limited trace distance brightening areas that receive trace hits, but incorrectly not receiving light that’s “too far away” even though it should be lit.
Pixar used “fill” lights to light there GI point clouds, then used the point cloud to indirectly light the scene, Fig. 7 Pg. 9: http://graphics.pixar.com/library/PointBasedGlobalIlluminationForMovieProduction/paper.pdf
Same thing as a uniform ambient light, but they just set up a bunch of light manually to control the look. And unfortunately of course they can trace that entire tiny scene, and control everything shot by shot so it “works right”. I can’t find the PDF of the lighting from Up specifically, only reference to it, but you get the idea. Light point cloud/voxels/etc. separately with ambient/“fill” then use lit point cloud/voxels to indirectly light scene.
Now how to do with a limited trace distance… one could try having a clamped minimum “normal” ambient term as well, such that everything receives a minimum ambient amount of light without overbrightening the areas that recieve the “ambient bounce” nor darkening areas that receive “actual” bounce light.
I can actually have the trace distance I want, is just increasing the samples, but I tried and it didn’t improve much. Like I said, it works on some scenes, but that color bleeding can backfire on you.
In any case, giving options is always good, and thanks for the link
Just wanted to drop by and say happy 1 year anniversary AHR!
I’ve only made a few posts in thread but I’ve always followed it everyday.
Oh I’m a happy person. My Titan X came in the mail early today. Was ale to get the beast installed but had to take my fiancée to school before I could play with it. When I get home though I plan on testing it out and seeing what it can do with AHR. I can’t wait. Will post results as soon as I get them.
Huh, well if a large trace distance didn’t work then I don’t know what’s wrong. Ohwell, keep up the good work either way.