Thank you sir.

I have now:

• divided ret_colour by samples
• Taken out the sqrt
• I take into account chromatic aberration, which is what your guess was
• I am changing the returns to continues
• As you can tell, I'm a complete beginner. I will look more into the BRDF. Thank you.

As you can see, it's a bit better. Thank you!

Hmmm… there's a strip in the middle of the image. Any ideas?

I'm guessing something related to normalization of rays or ray-surface self intersections that only occur when a coordinate is close to zero (assuming origin is aligned with image center). Does it move if you translate the whole scene (and camera) left or right by a significant amount?

The lighting still looks completely wrong to me - the back wall should not be nearly as dark as it is, meanwhile the top one seems to be way too bright. Bottom (floor) also seems wrong in terms of brightness.

There is something wrong in relation of how you generate secondary rays.

Do you explicitly sample light along each step on the path?

Yeah, it's looking better, but stuck in grayscale mode for some reason LOL – for one there are shadows now. Strange.

This looks more correct. Make sure you don't have just white walls! That might be the problem. How is the performance btw.?

The walls should be red and blue.

the performance is horrible 🙂

Are you using triangular models, or analytic?

Acceleration structure? I guess the one driver builds for you?

OK, one more step in the right direction. It's a bit too bright now.

The code is always at: https://github.com/sjhalayka/cornell_box_textured/blob/main/raygen.rgen

Triangles, not a lot of them. 2000 or so. Yes, I am using the default acceleration structures that come with Vulkan.