Fetching the paper…

CINEMA: Coherent Multi-Subject Video Generation via MLLM-Based Guidance · Around