Hollow teapot with global illumination. The only light is coming from the emissive sphere so no direct lighting. 5 bounces, 4-32 samples, 64 monte carlo samples which can bounce 4 times each. Thanks to global illumination the balls in the teapot are actually lit now, which makes me very happy. Render took 11:17:25 which is ridiculous so Im going to try and figure out how to use the GPU.Render took 11:17:25 which

The project 11 scene but with the point light swapped out for an emissive sphere, meaning that this scene now only has the indirect lighting. 5 bounces, 4-16 samples, 32 MCGI samples, allowing MCGI rays to bounce up to 4 times but only casting 1 extra sample on the bounces. 2:34 render.

Sample count image for the below. I dont really like how many extra samples are being taken on flat areas like the walls, but I dont know if that can be helped.

Same parameters as below but with uniform sampling. I think it looks much better, though it is much darker than the reference which is odd.

Fixed the issue from below. This uses cos-weighted distribution. 8-32 shadow samples, 16-256 primary ray samples, 128 Monte Carlo samples. 2:09:36 render. I dont like the lobe on the right ball, I think it really needs that cosine term from uniform distribution.

First attempt with uniform hemisphere sampling, only 4 per pixel. Ray directions are messed up somehow.

After doing everything new except global illumination. 16-256 samples, 8-32 shadow samples, 5 bounces. 22 seconds. Just saving the state of it before the global illumination absolutely nukes render time.

How project 11 renders out of the box without attenuation. Just thought it was really funny