Finally, I can start working on denoising!
Recently added MIS, HDR environment map importance sampling, Russian roulette, and support for glass and clearcoat materials to my renderer.
Here's a showcase of the latest progress!
Big thanks to @Kiiiri7 for the feedback and suggestions on the video!
@bunnylever Thanks so much, Matt! I find your work really inspiring too! I’d love to have you test the renderer and hear your feedback. It’s still a WIP, but I’ll definitely reach out when it’s a bit more ready!
New update to glTF Research Scenes!
Fixed missing normals on some meshes
Corrected roughness factor interpretation
Added a new scene
Check it out:
https://t.co/SimeQYrApp
@garrettkjohnson Agreed! That would make it a more useful benchmark for comparing renderers. I’ll include alternative env maps and new reference renders in the next update!
@Bubba_TX1 For example, the original Tungsten Veach MIS scene uses Beckmann and scales copper reflections by 0.3, while the PBRT export uses GGX and loses that multiplier. I'll eventually need to re-render the scenes using Cycles or some other path tracer for updated references.
@Bubba_TX1 Thanks for the feedback! You're right about the reference mismatches. There was previously an issue with PBRT's roughness² convention that I fixed, but some mismatches remain due to other material differences.
Thanks for the feedback! I agree about the reference images, and I’ll try to include renders of the converted glTFs in future updates.
As for the environment maps, some have licensing restrictions that prevent redistribution. But I’ll look into converting and including the ones that allow it!
Some of the reference environments already provide equirect HDR versions on their website.
Racing Game Prototype Update 🍂🏎️
Leaves that settle on the car and swirl in its wake, improved engine sound synthesis and two AI cars racing.
This time the whole video was edited fully inside the engine, no external editor!
Been working on adaptive exposure for my path tracer.
The exposure adapts over time toward a target luminance of 0.18.
Thanks to Alex Tardif for his histogram-based method for averaging luminance:
https://t.co/BnG6W3O4JB