Seedance 2.5

A next-generation multimodal audio-visual generation model by ByteDance, featuring 30-second single-clip generation, 50-slot multimodal reference conditioning, and timestamp-based local editing.