What a fascinating topic! I've always been curious about the potential of video models for synthetic data generation - can't wait to dig into this.