its IDREES's Highlights on 'FoundationVision/OmniTokenizer: OmniTokenizer: one model and one weight for image-video joint tokenization.' | Glasp