What albedo / normals AOVs to feed the denosier with "near"-specular materials (roughness < 0.15 for example)?

[TomClabault]

It makes sense that when the primary ray hits a perfectly specular material, the albedo & normals AOVs will only consider the secondary hit if it is itself non perfectly specular.

What about materials that are not exactly using a delta BSDF but very close to it? The OIDN guide does give some directions for handling layered / metallic / coated materials but it is still unclear to me what I should do exactly for close-to-specular materials.

Should I average the first hit and the second hit? But at this point, why not continue on the third hit?

[atafra]

Sorry for the late reply. It's hard to say what would be the best approach in your case. There are no strict rules for how to produce the albedo & normal AOVs, so it's best to try multiple approaches and compare the results. However, introducing additional noise in the AOVs might actually degrade quality, especially if you're not prefiltering the buffers.