A video model in ByteDance’s Seedance family, supporting clips up to 30 seconds with audio, endpoint guidance, multimodal references, editing, and extension for social content, products, and advertising.
Bytedance
Seedance 2.5 Flash is a video creation model in ByteDance’s Seedance 2.5 series. It combines written directions with image, video, and audio references for workflows spanning generation, revision, and extension, targeting 480p or 720p video tasks.
Flash retains endpoint guidance, omni references, editing, and extension, while differing from the standard version in resolution, reference capacity, and Draft support. Evaluate it against delivery requirements and sample results rather than transferring standard-version settings unchanged.
| Feature | Value |
|---|---|
| Clips up to 30 seconds | Provides room for a short story or product explanation. |
| Multimodal guidance | Communicates requirements that are difficult to express through text alone. |
| Endpoint constraints | Helps organize openings, endings, and shot transitions. |
| Generation and editing | Supports continued revision of existing results. |
| Multilingual audiovisual support | Enables dialogue and content production across languages. |
| 480p and 720p output | Offers resolution choices for sample review and intended channels. |
Start with delivery resolution, reference capacity, and workflow needs. Flash suits 480p and 720p tasks; evaluate the standard version for more references or a Draft workflow, comparing results with identical inputs.
Current BytePlus documentation lists only 480p and 720p for Flash output. Input-video resolution requirements do not establish equivalent output capabilities, so it should not be described as a 1080p or 4K model.
No: Flash does not support the official Draft mode. A 480p clip can be reviewed manually, but that does not provide a Draft task ID workflow for producing a higher-resolution final video.
Give each asset a clear purpose and avoid conflicting subjects, actions, or styles. More references are not automatically better; start with a small, clear set and add assets gradually to make problems easier to diagnose.
Assets, prompt intent, and parameters must match the selected task; rules from different modes should not be mixed. Validate a simple change first, then add camera or narrative requirements while retaining each result for comparison.
Open its LinkModel page and configure the published model identifier, authentication method, and request format. Complete a simple video task before enabling references, editing, or extension within the documented scope, and follow the specified result-retrieval flow.
Use fixed prompts and assets to assess subject consistency, motion continuity, sound, and editing success. Also verify output resolution, duration, task status, and failure handling rather than reusing standard-version parameters or another provider’s interface fields.
Video tokens equal (input video duration + output video duration) × output width × output height × 24 FPS ÷ 1,024. Without video input, input duration is zero.
| Output specification | Our pricing | Official |
|---|---|---|
| 480P | $0.060527 / second | $0.134505 / second |
| 720P | $0.13608 / second | $0.3024 / second |