ParentFull threadleumon·Just like for llms for the base model you need quantity not quality. It just needs to learn how to correctly predict the next frame.View on HN