Now available from ByteDance

Seedance 2.5 — 30 Seconds of Video in One Take

ByteDance's latest video model: a full 30 seconds in one take, up to 50 reference files, and native audio in 10+ languages. Start with free credits.

30 Seconds50 ReferencesTimestamp Control

30 Seconds, One Take

Renders a full 30-second scene in a single generation — no stitching short clips, no continuity breaks between cuts.

Write what happens at second 4, 11, 20. Seedance 2.5 follows the timing you specify instead of averaging the whole prompt.

Attach up to 50 images, video clips, and audio files to lock character identity, products, movement, and tone.

Select a label, face, or background area and describe the correction — timing, composition, and audio elsewhere stay untouched.

What's new

30-second single take keeps lighting, faces, and motion continuous — never separate segments.

Fix the beat at the wrong moment instead of regenerating the whole scene.

Up to 50 reference files so characters and products hold across every clip.

Upload a music track — Seedance 2.5 matches cuts, motion, and transitions to your beat.

Characters speak in 10+ languages with mouth movement matched to audio.

Upload a clip, describe your edit in text. Keep original audio or switch to AI audio.

Create an account and get free credits — enough to test with real prompts. No card needed.

Describe what happens and when. Calling out the moment an action lands gets you closer on the first render.

Render the full 30 seconds, then fix any region without touching the rest. Export when ready.

Native audio, 50 references, region-level editing — free credits on signup.