Seedance 2.5

30 seconds, 4K, 50 references: ByteDance's most advanced video model.

modellane/seedance-2.5 · Text to video · ByteDance

ByteDance's newest video model. It produces up to 30 seconds of 4K video in one pass and keeps characters, products, and places consistent across scenes with up to 50 multimodal references. Audio is generated along with the picture; with multilingual subtitles and in-frame text support it is a complete production pipeline.

Vendor: ByteDance
Task: Text to video
Output: Video
Price: $0.168 · per second
Longest run: 30 seconds

Input parameters

Prompt prompt — required, at most 20000 characters Describe what you want produced as concretely as you can.
Duration duration — optional, default: 5, range: 4–30 sn, affects price A longer duration raises the price directly.
Resolution resolution — optional, default: 720p, options: 480p, 720p, 1080p Higher resolution raises the per-second price.
Aspect ratio ratio — optional, default: adaptive, options: adaptive, 16:9, 9:16, 1:1, 4:3, 3:4 "adaptive" picks the ratio from the source.
Generate audio generate_audio — optional, advanced, default: true When off, the video is produced silent.
Output format output_format — optional, advanced, default: mp4, options: mp4, mov

Recommended use cases

Long single-shot scenes — Up to 30 seconds of uninterrupted footage; ideal for short films, ads, and promos.
Character and brand consistency — Keep the same character, product, and place across scenes with 50 references.
Multilingual content production — Video in 10+ languages with in-frame subtitles and text; no localisation step needed.
Cinematic production — Broadcast-ready output with 4K resolution, 10-bit colour, and synchronised audio.

Where it is strong

30 seconds in a single shot — the longest in its class
Support for 50 multimodal references
4K resolution, 10-bit colour, synchronised audio
Multilingual in-frame text and subtitles

Limits

High per-second cost — budget long clips carefully
No cuts, transitions, or multi-shot editing
Does not run in sync mode, queue only

Example prompt

A slow tracking shot through the Grand Bazaar in Istanbul, golden shafts of light through ancient windows, shopkeepers arranging colorful lanterns, the camera gently weaves between stalls, cinematic, 24fps, atmospheric dust particles in the light

Example request

POST /v1/queue/modellane/seedance-2.5

Playground

The identifier, price, parameters, and example requests on this page can be read without signing in. Running the model from the panel sends a real generation request and charges the project balance; the playground therefore requires an account.