Seeddance APIModels and inputs
Choose a public Seedance model. The API selects the workflow from the media URLs you provide.
Use one of the public model IDs below. Seeddance keeps provider model names internal, so they can change without affecting the public API.
| Model | Typical use | Image inputs | Video refs | Audio refs | Quality |
|---|---|---|---|---|---|
seedance-1.0-pro-fast | Fast generation with lower credit use | Up to 1 | No | No | 480p to 1080p |
seedance-1.5-pro | Start and end frame control | Up to 2 | No | No | 480p to 1080p |
seedance-2.0 | High-quality multimodal generation | Up to 9 | Up to 1 | No | 480p to 1080p |
seedance-2.0-fast | Faster Seedance 2.0 generation | Up to 9 | Up to 1 | No | 480p to 1080p |
seedance-2.0-mini | Seedance 2.0 with lower credit use | Up to 9 | Up to 1 | No | 480p to 720p |
seedance-2.5 | Text, frame, and reference generation | Up to 30 | Up to 10 | Up to 10 | 480p to 720p |
Your account plan determines which models and resolutions you can use. The API returns a 403 or 422 error before it creates a task with an unsupported combination.
Automatic generation mode
Do not send a provider model ID. Seeddance selects the workflow from image_urls, video_urls, and audio_urls:
- With no media, Seeddance uses text-to-video generation.
- With one image, Seeddance uses image-to-video and treats the image as the start frame.
- With two images, Seeddance uses first and last frame generation when the model supports it.
- With three or more images, Seedance 2.0 and 2.5 models use reference-to-video generation.
- With a video, supported models use reference-to-video generation.
- With audio, Seedance 2.5 uses reference-to-video generation.
Set reference_mode to true when one or two images should be treated as reference assets instead of first and last frames. Seedance 2.0 and 2.5 support this option, and the request must include at least one media URL.
All media must use public HTTPS URLs. Seeddance rejects localhost, private network addresses, embedded credentials, and hostnames it cannot resolve.
Parameters
Each model supports its own duration, quality, and aspect_ratio values. The API checks these values against the selected model and your account plan. generate_audio works only with models that support synchronized audio.
Seedance 2.5 is available
Use seedance-2.5 for 4 to 30 second output at 480p or 720p. It supports
text-to-video, one or two frame inputs, and reference generation with up to 30
images, 10 videos, and 10 audio files. Seedance 2.5 also accepts
output_format, content_filter, reference_mode, and text-only
web_search.
Seedance 2.5 media limits
- Images: JPEG, PNG, or WebP; 300 to 6000 px per side; 0.4 to 2.5 aspect ratio; up to 30 MB each.
- Videos: MP4 or MOV; 2 to 30 seconds each and 30 seconds total; 24 to 60 FPS; 300 to 6000 px per side; up to 200 MB each.
- Audio: MP3 or WAV; 2 to 30 seconds each and 30 seconds total; up to 15 MB each.
- Reference prompts can address assets as
@image1,@video1, and@audio1in array order. - Use URL inputs rather than Base64. The provider limits each request body to 64 MB.