StyleCLIP

The first two demo videos are interesting examples of using StyleCLIP's global directions to guide an image toward a "smiling face" as noted in that paper with smooth interpolation: https://github.com/orpatashnik/StyleCLIP I had ran a few chaotic experiments with StyleCLIP a few months ago which would work very well with smooth interpolation: https://minimaxir.com/2021/04/styleclip/

The approach he used didn't really make sense. I get the impression Tom is so dedicated to his old custom hackedup NN framework that he tailors the approach to what he could easily do in it. If you asked someone else, "I want to automatically turn characters into 'more uppercase' or 'more lowercase' versions, and it has to involve neural networks; what do?", they would say something like "Easy enough! dump character maps into a standard StyleGAN2-ADA, train for a few days, then find the latent direction corresponding to uppercasing & lowercasing, and edit whatever font you please." (People have been generating fonts with RNNs or GANs for ages.) You could do this over a week or so, the tooling has gotten pretty easy to use. And if you asked someone on the cutting-edge, they'd suggesting using OA CLIP through Aleph/BigSleep/etc to automatic