Code for a WGAN implementation in JAX and Equinox. Packages managed with astral-sh/uv. Referenced DVD-GAN paper for architecture. Includes dataset of Twitch emotes. The data_collection folder contains scripts to collect the gifs yourself. Run Trainer.py to train the model. Currently, I have not found a stable training configuration. Moreover, compile time for the model is long, likely due to use of spectral normalization, so that can definitely be improved. Generated examples will be posted in the GH page.