88API88API
User GuideAI ApplicationsAPI ReferenceHelp & Support
Video

SD Cloud Edition Video API

Resolution-based billing for SD2.0 and SD2.5, plus duration, ratio, reference media, keyframe and audio rules.

This page specifies the SD cloud edition video models (SD2.0 and SD2.5). They share the POST /v1/videos create and GET /v1/videos/{id} polling flow with the other video models; authentication, media upload, polling states, downloads and error handling are covered in the Complete Video API Standard. This page lists only the fields and rules specific to this group.

Updated 2026-10-10. These models support polling only: do not send callback_url, or the request is rejected. The SD cloud edition is a different model group from Seedance Official Edition and Seedance Official Relay; parameters and capabilities do not carry over.

Models

ModelResolutionsDurationRatioImage / video / audio limitsKeyframesGenerated audio
SD2.5480p (default), 720p, 1080p4–30s, default 5srequired, see below30 / 10 / 10yes, exclusive with ordinary referencesgenerate_audio
SD2.0480p (default), 720p, 1080p, 4k4–15s, default 5sdefault 1:19 / 3 / 3, 12 in totalyes, exclusive with ordinary referencesnot configurable

Each series sells one model name. resolution selects the output tier and the tier that is billed, for example "model": "SD2.5", "resolution": "720p" or "model": "SD2.0", "resolution": "4k".

Common rules

  • Resolution: send a tier from the table in resolution, in either case (4k and 4K are equivalent). Without it, the lowest tier 480p is used. A tier the series does not offer returns 400; it is never downgraded.
  • Duration: whole seconds in duration (or the string seconds); defaults to 5 seconds.
  • Aspect ratio: ratio (or aspect_ratio) accepts 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9.
    • SD2.5 requires an explicit ratio and does not support auto; omitting it or sending auto returns 400.
    • SD2.0 defaults to 1:1.
  • Prompt:
    • SD2.5: prompt is required.
    • SD2.0: may be omitted when references or keyframes are supplied, but describing the motion and camera is recommended.
  • Generated audio: SD2.5 accepts generate_audio (boolean). SD2.0 has no switch; sending it returns 400.
  • Optional fields: negative_prompt and seed (integer).

Reference media

  • Put reference images in the top-level images array.
  • Put reference videos in metadata.referenceVideos and reference audio in metadata.referenceAudios.
  • SD2.0 allows at most 12 references across all three types.
  • For SD2.5, each reference video is 2–30 seconds and the total is at most 30 seconds.
  • Use one field per media type. If several aliases are sent, only one set is used; they are not merged.
{
  "model": "SD2.5",
  "resolution": "1080p",
  "prompt": "Keep the face and outfit from the image and run through a rainy street with the camera motion of the video",
  "duration": 10,
  "ratio": "9:16",
  "generate_audio": true,
  "images": ["https://cdn.example.com/character.jpg"],
  "metadata": {
    "referenceVideos": ["https://cdn.example.com/camera-motion.mp4"]
  }
}

First and last frames

Put keyframes in metadata.firstFrame and metadata.lastFrame. lastFrame requires firstFrame. Keyframes cannot be combined with reference images, videos or audio.

{
  "model": "SD2.0",
  "resolution": "720p",
  "prompt": "The camera glides from the coast at dawn to the same bay at dusk",
  "duration": 8,
  "ratio": "16:9",
  "metadata": {
    "firstFrame": "https://cdn.example.com/start.jpg",
    "lastFrame": "https://cdn.example.com/end.jpg"
  }
}

File upload

You can also send one local image with multipart/form-data in the input_reference field; do not send images or image in the same request. Reference videos and audio cannot be sent as files; obtain a URL through Local media upload first.

Earlier model names

The earlier names with a tier suffix (such as SD2.5 720P or SD2.0 4K) still work. Their output and billing stay at the tier in the name, and resolution is ignored. They are no longer listed in the model square; new integrations should use SD2.5 or SD2.0 with resolution.