Seedance 2.5 makes 30-second video with synced audio, $0.10/second
ByteDance's model takes text, an image, or up to 50 reference clips, speaks 10+ languages lip-synced, and is live on OpenRouter — no enterprise gate.
Copy markdown
30 seconds in one pass, audio baked in
Seedance 2.5 renders up to 30 seconds of video in a single native pass — with synchronized sound, voiceover, and lip movement — then extends the shot twice for longer sequences. You get a finished, scored clip instead of stitching a separate TTS track on top.
It talks in 10+ languages, lip-synced
Native audio generates captions, voiceover, and localized dubs in 10+ languages, timed to the mouth movement it draws. That means multilingual product demos or dubbed shorts from a single prompt, with no separate lip-sync pass.
$0.1028 a second, out of enterprise beta
Model ID bytedance/seedance-2.5 bills at $0.1028 per generated second on OpenRouter — about $3.08 for a full 30-second clip. It left enterprise-only beta on Aug 7, so any account can call it today.
Pin the frames, feed it 50 references
You can lock the first frame (or first and last) and hand it up to 50 image, video, and audio references for character and style consistency. Reference-to-video plus frame control makes a recurring character across shots actually repeatable.
Build this weekend: a narrated explainer
Drop in a script and one product screenshot and get a 30-second narrated walkthrough with synced voice — then re-run the same seed for a five-language dubbed set. ByteDance's Dreamina markets up to 4K on the final render.