Seedance Video Workflows
Seedance is a video model family in Redbit’s current registry. In the dashboard, Seedance-specific controls appear when the active card is avideo card and the selected model is seedance or a doubao-seedance-2.0* variant.
Who Should Read This
Read this page when a workflow needs Seedance-specific reference roles, when video generation is slow or failing, or when you need to decide whether a reference image, first/last-frame setup, or multi-reference setup is appropriate.Before You Read
Confirm provider access in Models and Provider Configuration. Seedance requests may go through direct provider settings or a compatible relay, and generation can be slower than image generation because the provider may queue, poll, and post-process video jobs.Current Variants and Output Settings
| Setting | Current values in the codebase |
|---|---|
| Model ID | seedance |
| Variants | doubao-seedance-2.0, doubao-seedance-2.0-fast |
| Ratios | 9:16, 16:9, 1:1, 4:3, 3:4, 21:9 |
| Durations | 4, 5, 6, 8, 10, 15 seconds |
| Resolutions | 480p, 720p, 1080p |
| Default output options | 720p, audio enabled, web search disabled unless changed in the card settings |
| Reference media roles | first frame, last frame, reference image, reference video, reference audio |
Modes
| Mode | Required input | Use for |
|---|---|---|
| Text to Video | Prompt only | Prompt-led motion ideas with no source frame |
| Image to Video | A first-frame image | Animate from a selected starting image |
| First/Last Frame | First-frame and last-frame images | Guide a transition between two fixed visual anchors |
| Multi-Reference | At least one image, video, or audio reference | Let multiple media references influence the video task |
Reference Rules
Seedance reference handling is stricter than generic video reference handling:- text-to-video does not use references;
- image-to-video requires a first-frame image;
- first/last-frame mode requires both a first-frame and a last-frame image;
- multi-reference mode requires at least one reference asset;
- frame mode references cannot be mixed with free multimodal references in one task;
- reference tokens such as
@image1or@video1can be inserted into prompts after references are collected.
Recommended Workflow
Configure provider access
In Settings, configure Seedance direct credentials or a compatible relay. Seedance’s registry entry currently lists
wavespeed and other as relay options.Create a video card
Open the
Video tab, select Seedance, and confirm the variant, ratio, duration, resolution, and audio setting.Choose the mode
Use text-to-video for prompt-only work, image-to-video for a start frame, first/last-frame for a transition, or multi-reference for mixed media guidance.
Attach references
Upload assets or select them from Asset Dock/SmartPicker. Check that the assigned roles match the selected mode.
Troubleshooting
| Symptom | Likely cause | What to check |
|---|---|---|
| Missing first-frame error | Image-to-video mode has no first-frame reference | Attach an image and keep its role as first_frame |
| Missing last-frame error | First/last mode has only one endpoint frame | Attach both first and last frame images |
| Reference mix error | Frame references and free references are mixed | Use either first/last frame roles or multi-reference roles, not both |
| Provider error | Credentials, relay capability, endpoint, quota, or model variant mismatch | Re-check Settings and provider console |
| Slow generation | Provider queue or long video settings | Try shorter duration, lower resolution, or fast variant where appropriate |