vpance
Member
There are 1001 papers on inverse rendering for lightsource determination.
Long story short, the issue is ill-posed and approaches np hard if you have as little information as this model has at the point it currently comes into play. It's quite literally impossible, mathematically and from a runtime perspective, that this will ever deliver accurate results if it doesn't get an architectural rework with a full integration into the pipeline.
Just depends if the fudge factor of accuracy can produce results that are acceptable for the compute cost vs a high quality PT option. Theoretically it should be a viable alternative otherwise why go down this path at all if PT is the only acceptable way forward? But this goes beyond lighting too ofc.
To the average gamer's eye they may already be there apart from the unfortunate exaggerated studio portrait faces.
For anyone who missed it
Last edited:



