Seedance 2.5 is ByteDance Seed's video model with native audio, released on July 31, 2026: it makes 4–30 second clips at 480p, 720p or 1080p (24 fps) from text, first/last frames or up to 50 references (30 images, 10 videos, 10 audio clips), and can edit or extend videos. This free builder turns your idea into a prompt in ByteDance's recommended format (timed segments, @Image 1 references, {dialogue}, <sound effects>, (music)), outputs the ModelArk API JSON with the right ratio/duration rules, and calculates the exact token cost from BytePlus' list price. It doesn't render video: paste the prompt into Dreamina (global), Jimeng or Doubao (China), or send the JSON to the API. There are no open weights.
How to use this Seedance 2.5 prompt builder
- Pick the task. Text to video, first frame, first + last frame, omni reference (images, videos and audio together), edit a video, or extend a video. The builder applies BytePlus' rules automatically: first-frame, edit and extend tasks use ratio adaptive, and editing uses duration −1 so the output keeps the input video's length.
- Split the clip into timed segments (0–5 s, 5–10 s…). Seedance 2.5 can carry one continuous story for up to 30 seconds.
- Name what each reference does. Write what
@Image 1,@Video 1or@Audio 1should provide (appearance, motion, voice) — ByteDance recommends this explicitly. - Check the cost box, then copy the prompt into Dreamina (global), Jimeng or Doubao (China), or copy the JSON for the ModelArk API.
How the cost is calculated
BytePlus bills Seedance 2.5 by tokens: (input video seconds + output seconds) × width × height × 24 fps ÷ 1,024, at $10.70 per million tokens for 480p/720p and $11.70 for 1080p without video input, or $6.40 / $7.00 when the input includes video. A 5-second 16:9 clip therefore costs $0.514 at 480p, $1.156 at 720p and $2.843 at 1080p — the same figures as BytePlus' own examples. Failed generations (for example, blocked by moderation) are not charged. When the input includes video, BytePlus also applies a minimum charge; this page approximates it by billing at least 4 seconds of input video. Source: BytePlus ModelArk pricing.
Seedance 2.5 at a glance
| Seedance 2.5 | |
|---|---|
| Made by | ByteDance Seed |
| Release date | July 31, 2026 (Jimeng and Doubao in China; Dreamina globally from the same day) |
| Clip length | 4–30 seconds, or −1 to let the model choose |
| Resolution | 480p, 720p (default), 1080p; 24 fps |
| Aspect ratios | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, or adaptive |
| References | Up to 50: 30 images, 10 videos (2–30 s each, 30 s total), 10 audio clips |
| Tasks | Text to video, first frame, first + last frame, omni reference, edit, extend |
| Dialogue languages | 11, including English, Chinese, Spanish, Japanese, Korean and Arabic |
| Open weights | No, hosted only |
Where to use Seedance 2.5
| Option | Who it's for |
|---|---|
| Dreamina | Global users, in the browser |
| Jimeng (Video Generation) and Doubao | Users in China |
BytePlus ModelArk API, model dreamina-seedance-2-5-260628 |
Developers; needs at least $30 of balance, a savings plan or a resource pack before the model can be enabled |
Seedance 2.5 API pricing (BytePlus list price)
Billing is by tokens: (input video seconds + output seconds) × width × height × 24 ÷ 1,024.
| Output | Without video input | With video input |
|---|---|---|
| 480p / 720p | $10.70 per million tokens | $6.40 per million tokens |
| 1080p | $11.70 per million tokens | $7.00 per million tokens |
| 5-second 16:9 clip, no video input | Cost |
|---|---|
| 480p | $0.514 |
| 720p | $1.156 |
| 1080p | $2.843 |
Failed or moderation-blocked generations aren't charged. Resource packs run 5M tokens for $32, 10M for $64 and 100M for $640 (valid 90 days).
Limits worth knowing
- Uploads containing real human faces are blocked; ByteDance points to its portrait-asset options instead.
- Images must be JPEG, PNG, WebP, BMP, TIFF, GIF or HEIC/HEIF, under 30 MB, 300–6,000 px per side, with an aspect ratio between 0.4 and 2.5.
- Editing keeps the input video's length (duration −1) and aspect ratio (adaptive); first-frame and extend tasks also use adaptive ratio. The builder sets these automatically.
Frequently asked questions
When was Seedance 2.5 released?
ByteDance Seed released Seedance 2.5 on July 31, 2026, in Jimeng and Doubao in China and on Dreamina globally. The BytePlus ModelArk API model is dreamina-seedance-2-5-260628.
How long can a Seedance 2.5 video be?
4 to 30 seconds per generation, or set duration to −1 and the model picks. You can also extend an existing video.
How much does Seedance 2.5 cost?
On the BytePlus API it is billed by tokens: $10.70 per million at 480p/720p and $11.70 at 1080p, or $6.40/$7.00 when the input includes video. A 5-second 16:9 clip costs about $0.51 at 480p, $1.16 at 720p and $2.84 at 1080p. Dreamina and Jimeng use their own credits.
Is Seedance 2.5 free or open source?
It isn't open source; there are no downloadable weights. This prompt builder and cost calculator is free, but generating the video uses Dreamina/Jimeng credits or paid API tokens.
How do I reference images, videos and audio in a Seedance 2.5 prompt?
Call them @Image 1, @Video 1, @Audio 1 in the order you upload them, and say what each one provides (appearance, motion, voice). Put dialogue in {}, sound effects in <> and music in ().
Can I upload photos of real people to Seedance 2.5?
Uploads with real human faces are blocked. ByteDance offers portrait-asset options for that use case instead.
Does this tool upload my prompt anywhere?
No. It runs entirely in your browser; nothing is sent until you paste the prompt or JSON into Dreamina, Jimeng or the API yourself.