ByteDance's Seedance 2.5 turns 50 reference files into one 30-second film
ByteDance released Seedance 2.5, a video model that generates clips of up to 30 seconds in a single run and accepts as many as 50 multimodal references — images, text, style frames, character shots — to steer the result. Audio and video are generated jointly, and the model keeps character appearance, lighting and motion style consistent across the whole clip. It is rolling out on ByteDance's Jimeng AI and Doubao Pro apps, with an API coming to the Volcano Engine platform.
Why it mattersDays after MiniMax's H3, China's video-model race keeps compressing film-style control into consumer apps. Thirty coherent seconds in one run plus heavy reference control is what creators need for real storytelling — not just pretty five-second demos.
Read in the app — free, in 9 languages
Related stories
DeepSeek's cheap workhorse can now see — and on agent tasks that need eyes it says it is close to Anthropic's best2026-08-21Adobe will now generate the music, the voiceover and the door slam — and it says the licence covers you
2026-08-21Show the robot once — three to twelve seconds — and it gets the job right 59 times out of 100 with no training at all
2026-08-21Six months ago Hollywood sent ByteDance a cease-and-desist. This week they signed a truce — and no money changed hands.
2026-08-20The model invents its own tasks, builds the rig to test them, then trains on the results — and DeepReinforce gave the weights away
2026-08-20