zlacker

Do you know that the final trained model is only 2GB? There is no way it can reproduce anything verbatim. There is also Riffusion that can generate music after being trained on FFTs of music.