Seedance 2.5 API: 30-second clips, resolutions, references

On the Sume API, Seedance 2.5 is the model id seedance-2.5 on POST /v1/videos, and it makes 4–30 second clips at 480p, 720p, or 1080p. The rest of the family, seedance-2, seedance-2-fast, and seedance-2-mini, stops at 15 seconds, and of those three the docs list 1080p only for seedance-2.

Every limit below comes from Sume's Video generation and Video Router docs and the Sume API reference, read on 2026-09-27. For Seedance outside Sume and its vendor prices, see Sume vs Seedance; for lengths across all models, see AI video length limits by model.

Which Seedance ids can I call, and what are their limits?

All four ids work on POST /v1/videos and on the legacy POST /v1/video-router/generate, which takes the same ids with flat fields such as image_url and reference_image_urls; the docs say new integrations should use /v1/videos. The documented limits differ by id:

How do I request a 30-second Seedance 2.5 clip?

Send duration: 30 with model: "seedance-2.5", the only Seedance id that goes past 15 seconds. duration is whole seconds, and resolution and aspect_ratio must be values the model advertises. The request below asks for a 30-second 1080p clip guided by a reference video; the submit-and-poll flow and the error for an unlisted length are in AI video length limits by model.

• Send resolution explicitly. For the Video Router, the API reference says an omitted one falls back to provider or runtime defaults.
• The docs' example descriptor, for seedance-2, lists six aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. 2:3, 3:2, and 9:21, which the /v1/videos enum also has, are not among them. Read seedance-2.5's own list from GET /v1/videos/models.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: seedance-25-30s-001" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "A ceramic mug on a stone plinth, the camera moving like the reference clip",
    "duration": 30,
    "resolution": "1080p",
    "aspect_ratio": "9:16",
    "input_references": [
      { "type": "video_url", "video_url": { "url": "https://example.com/camera-move.mp4" } }
    ]
  }'

Which inputs does each generation mode take?

The fields you send pick the mode. Reference-to-video API compares references across models.

• Text-to-video: prompt alone, as in the API reference's seedance-2.5 and seedance-2-mini examples.
• Image-to-video: frame_images, at most two entries, each with a frame_type of first_frame or last_frame. The docs' seedance-2 descriptor lists both types.
• Reference-to-video: input_references, at most 12, typed image_url, video_url, or audio_url. The docs say the Seedance 2.x models honor audio and video references, and a type must appear in the model's supported_input_references.
• Both at once: frame_images takes precedence, and the request runs as image-to-video.

Does Seedance generate audio?

It can. generate_audio defaults to the model's audio capability. The docs describe seedance-2 as “text/image/reference-to-video with optional audio” and show generate_audio: true in its descriptor, and the API reference's examples send generate_audio: true on seedance-2.5 and false on seedance-2-mini. Send false explicitly for a silent clip.

What does Seedance on Sume not accept?

These are refused or not documented for Seedance:

• seed: no v1 model accepts it, and each rejects the field.
• size: pixel sizes return 400 unsupported_parameter; use resolution and aspect_ratio.
• A non-empty provider.options: also 400 unsupported_parameter.
• A clip to edit: the API reference lists video-to-video only for gemini-omni-flash-1.1.
• 1080p on seedance-2-fast or seedance-2-mini: not documented. The docs and the API reference name 1080p only for seedance-2.5 and seedance-2 among Seedance ids.
• More than 15 seconds on any Seedance id except seedance-2.5.

How is Seedance billed?

By the video token. The docs' seedance-2 descriptor lists a per-1000-video-tokens rate in pricing_skus, and GET /v1/videos/models shows each id's own rate. Billing is the provider's list price × 1.25, reserved on submit, plus a 5.5% agent fee by default; Sume vs Seedance compares that with buying Seedance directly. For a 1080p clip, see API pricing and read the poll response's usage.cost, the Sume billable amount, once the job completes.