Seedance 2.5
Seedance 2.5 is a video model family by ByteDance. Seedance 2.5 supports longer scene planning and several reference types in Vini. Use its actual mode limits below. A longer available duration is room for action, not a reason to fill a clip with unrelated events.
Resolution and sound
Section titled “Resolution and sound”Choose 480p, 720p or 1080p. New nodes start at 480p with sound on; a saved explicit resolution remains selected. Turning sound off is supported. Generation accepts 4–30 seconds; editing keeps the source length.
The 1080p output uses 10-bit H.265/HEVC and may require a compatible player. Review the quote again after changing resolution, duration or reference videos: the price depends on output tokens and the input video. The vendor’s two-step Draft-to-final workflow and whitelist-only Premium model are not connected by this resolution selector; normal Seedance 2.5 does not offer 4K here.
Settings were checked against the BytePlus Seedance 2.5 documentation and pricing on 20 October 2026. This check establishes available settings, not a new rendered result.
| Setting | Value |
|---|---|
| Model ID | seedance25 |
| Name | Seedance 2.5 |
| Availability | available |
| Default for new nodes | no |
Common declared limits
Section titled “Common declared limits”| Setting | Value |
|---|---|
| Capabilities | text to video, image to video, native audio, dialogue, reference video, reference audio |
| Maximum prompt length | no fixed maximum declared |
| Duration | 4–30 seconds; fallback 8 seconds |
| Audio | model-generated audio, off; fallback model-generated audio |
| Aspect ratio | 9:16, 16:9, 1:1, 4:3, 3:4, 21:9; fallback 9:16 |
| Declared language support | prompt: English; speech: English, Chinese |
| Combined reference limits | video frame: min pixels: 407696; images: 30; videos: 10; audio inputs: 10; combined maximum: 50; audio without visual input: yes; video reference seconds: minimum: 2; maximum: 30; audio reference seconds: minimum: 2; maximum: 30; trim: head; av seconds total: 30 |
| Controls | resolution: 480p, 720p, 1080p; new node 480p; fallback 480p |
| Element input form | frames |
| Content constraints | people: restriction level: risk; applies to: first frame, reference image, reference video |
| Negative prompt strategy | instruction in prompt text |
Mode: from_first_frame
Section titled “Mode: from_first_frame”| Setting | Value |
|---|---|
| Aspect ratio | same as source video |
| First frame | required |
| Last frame | accepted; optional |
Mode: from_references
Section titled “Mode: from_references”| Setting | Value |
|---|---|
| Video reference | style |
| Source video | accepted; optional |
| References | maximum: 30 |
Mode: text_to_video
Section titled “Mode: text_to_video”Uses the common declared limits above.
Mode: edit_video
Section titled “Mode: edit_video”| Setting | Value |
|---|---|
| Duration | same as source video; fallback same as source video |
| Aspect ratio | same as source video |
| Source video | required |
| References | maximum: 30 |
Mode: extend_video
Section titled “Mode: extend_video”| Setting | Value |
|---|---|
| Aspect ratio | same as source video |
| Source video | required |
| References | maximum: 30 |
Unavailable modes: avatar_audio, motion_transfer, recast.
Limits apply together. initial is the studio choice for a new node; default is the fallback for an existing node without that value. Declared language or input support does not guarantee a successful result.
Prompting playbook
Section titled “Prompting playbook”- For several shots, assign a purpose and a visible event to each shot. Keep the same character and place descriptions across the sequence.
- Describe an extension as the next action after the source ends, rather than restarting the scene.
- Use the analysis of a style sample to describe camera and light. Exact imitation of its motion is a separate, unproven requirement.
The built-in Seedance 2.5 authoring skill adapts a scene request for this family through Prompt for the model. It prepares instructions; it does not render video or certify the result.
Studio example
Section titled “Studio example”Start with an approved clip of a gardener holding a pot. Select extension in Video with Seedance 2.5. Ask Prompt for the model: “Continue from the existing pose. The gardener carries the pot to the windowsill and gently sets it down.” Check that the source ends before this action begins.
A run may consume credits. Review the prompt and inputs, then inspect the finished clip for action, identity, framing and sound. For automation, use the supported MCP workflow.
Known failure modes
Section titled “Known failure modes”A character sheet was refused in the 30 September extension test; clean frames passed. The passing extension approximated a gesture rather than reproducing it exactly. On 4 October, adding a style clip also brought unwanted landscape content in some prompts. Text describing how a scene was filmed can be a better starting point than attaching that clip.
Evidence and limits
Section titled “Evidence and limits”Character-reference and replacement-edit cases passed on 28 September 2026. A clean-frame extension retained continuity on 30 September. The 4 October video-style result was partial: cutting rhythm and light transferred, but the exact camera moves did not. A later words-only version avoided unwanted sample content, with remaining camera and framing differences.
A reference sound longer than 30 seconds is not refused: Vini sends its first 30 seconds and says so on the Video card before the run. The combined 30-second budget for reference video and sound counts what is sent, so a 10-second clip with a 60-second song is still refused. The cut is made by the studio; no render was made to judge where it falls in the music.
The studio registry establishes available inputs and modes. A recorded render establishes only its observed outcome. The applicability guide explains the distinction. See also character identity, static camera and video style samples.
See the model–technique applicability matrix for connected operations, dated outcomes and configurations that remain untested.
Reviewed 14 October 2026 against the current studio contract and dated studio research. Official model-author documentation checked on that date; author capabilities may exceed Vini’s connected controls. Recheck this playbook when the model version, quality or mode changes.