Hacker News Reader

Who is going to put each frame into image to image AI model and get out a photorealistic output and recompile?

Lots of photorealistic phalluses are going to be generated.

Only if you're using a model trained on them

I think it would be against the spirit of the project not to do so

I've read white papers seemingly claiming with the right prompt these sketches would turn into the indicated art.

That would be quite the compression scheme.