01
A complete scene in one take
Setup, action, reaction, and payoff in one 30-second clip, instead of four short generations that do not quite match.
Long-form AI Video Model
Seedance 2.5
30-second scenes from up to 50 references
Write the whole scene — who is in it, what happens second by second, what it sounds like — and get one continuous clip of up to 30 seconds, with up to 50 references holding faces, places, motion, and voices in place.
Creative workspace
Seedance 2.5
68
What is Seedance 2.5?
Seedance 2.5 is ByteDance's long-form video model. It generates up to 30 seconds in one continuous clip, reads up to 50 references at once — 30 images, 10 clips, and 10 audio files — and follows timestamps to the second, so a whole scene can be written, cast, and timed in a single request.
How the people who built the model describe it:
“Seedance 2.5 raises the maximum generation length from the 2.0 series’ 15 seconds to 30 seconds, generating a coherent video of up to 30 seconds in one pass so a complete story can be told without stitching clips together.”
Seedance 2.5 at a glance
- Duration
- 4–30 seconds
- References
- Up to 50
- Output
- Up to 1080P
What you can make with the Seedance 2.5 tool
These are the jobs people actually bring to Seedance 2.5, and the settings each one needs.
02
Large casts and worlds
Give each character their own reference images and voice, add the locations and props, and keep all of them consistent across the scene.
03
Localized campaigns
Generate dialogue natively in 11 languages, from English and Chinese to Arabic, Japanese, and Portuguese, without dubbing afterwards.
Seedance 2.5
A workspace shaped for the model
Seedance 2.5 puts the handful of controls that change the result most in front of you, and keeps the rest out of the way.
01
30 seconds without a seam
Seedance 2.0 stops at 15 seconds; 2.5 generates up to 30 in one pass. Because it is one generation rather than several stitched together, faces, lighting, and motion carry through the whole clip instead of jumping at each join.
02
Up to 50 references in one request
Up to 30 images, 10 clips, and 10 audio files, with clips and audio each totalling up to 30 seconds. Audio can be used on its own, and multi-angle character sheets work as references — two things Seedance 2.0 does not handle well.
03
Timed to the second
Write "0–3s", "3–8s", or "at 5s, a fast whip-pan" and the scene follows it — Seedance 2.0 only understands shot numbers. Keep the timeline continuous and give each segment enough to fill it.
04
Carry a shot on
Upload a clip and ask for what happens before or after it. The continuation keeps the source clip’s framing, so the join does not jump — for the best match, give it the clip you want extended at the aspect ratio you want out.
05
Speaks 11 languages
Dialogue and lip-sync are generated natively in Chinese, English, Spanish, Indonesian, Malay, Thai, Arabic, Portuguese, Vietnamese, Japanese, and Korean — write the line in the language it should be spoken in.
Seedance 2.5
Choose with context
Where Seedance 2.5 is the right call, and where a different model does the job better.
Maximum duration
Seedance 2.5: 30 seconds · Seedance 2.0 series: 15 seconds
Reference capacity
Seedance 2.5: 30 images, 10 clips, 10 audio (50 total) · Seedance 2.0: 9 images, 3 clips, 3 audio
Timing control
Seedance 2.5: second-level timestamps · Seedance 2.0: shot numbers only
Audio-only references
Seedance 2.5: supported · Seedance 2.0: audio must come with an image or clip
Maximum resolution
Seedance 2.5: 1080P · Seedance 2.0: 4K
Choose based on the job
Choose 2.5 for length, large casts, and precise timing; choose 2.0 when you need 4K
How to write a Seedance 2.5 prompt
01
Map the references
Number each file by upload order within its type and bind it to a role: "images 1–2 are character 1, who speaks with audio 1; images 3–4 are character 2, with audio 2." Say which part of a reference to use — "the spell-casting motion from video 1, the orbiting camera from video 2".
02
Summarize in one line
Subject, place, event, genre or style, and any signature camera move.
03
Write the timeline
Split the action into continuous whole-second ranges — "0–3s … 3–7s … 7–15s" — with the picture, camera, action, dialogue, and sound for each. Describe what should happen rather than what should not.
04
Finish with what runs throughout
Add the camera position, environment, sound, and mood that apply to the whole clip, and any exclusions such as "no subtitles" or "no background music, only ambience".
Seedance 2.5 FAQ
Compare the Seedance 2 models
Compare output length, resolution, references, aspect ratios, and audio support before choosing a model for your next video.
| Capability | Seedance 2.0 | Seedance 2 Mini | Seedance 2.5Current model |
|---|---|---|---|
| Best for | High-resolution production | Fast, economical iteration | Longer, coherent storytelling |
| Output duration | 4–15 seconds | 4–15 seconds | 4–30 seconds |
| Output resolution | 480P / 720P / 1080P / 4K | 480P / 720P | 480P / 720P / 1080P |
| Input and references | Text + image / video / audio references | Text + image / video / audio references | Text + image / video / audio references |
| Reference capacity | 9 images · 3 videos · 3 audio files | 9 images · 3 videos · 3 audio files | 30 images · 10 videos · 10 audio files |
| Aspect ratios | 21:9 · 16:9 · 4:3 · 1:1 · 3:4 · 9:16 | 21:9 · 16:9 · 4:3 · 1:1 · 3:4 · 9:16 | 21:9 · 16:9 · 4:3 · 1:1 · 3:4 · 9:16 |
| Native synchronized audio | Supported | Supported | Supported |
Start Creating With Seedance 2.5
Write the brief, submit it here, and follow the result in Studio.