[matroid]

That's a good question!

We made the training data for this using a physically based renderer (i.e, Cycles in Blender).

For non-photorealistic art, there are way too many ways in which artists communicate lighting (even in sketch, which is different from pixel-art / video game, there are at least a couple of ways (a) solid shading (b) hatching).

I'm not sure how to extend our methodology (synthetic data in 3D modeling software + train model) to these settings.

Ooh, I didn't know about Myst-style games. I'll definitely check them out!

[vunderba]

That makes sense. I almost thought maybe one way to do it might be to use generative AI with a ControlNet to create a more photorealistic 3D version of each of the pixel-art environments, and then use those images with your approach.... then somehow separate the lighting information and mask it over the original pixel art?

But I’m not sure if that’s actually feasible or how it would work technically. :)

[matroid]

I guess it would be a good experiment these days, to see how Nano Banana or OpenAI's latest image to image models will be at transferring lighting while preserving the original style domain, like you suggest.

Thanks for the thoughts :)