Video task status is failed with no error detail: a debugging checklist for Seedance

When a Seedance video task returns status failed without an error detail, poll GET /v1/video/generations/{task_id} for the final state, then verify your request payload: model, prompt, seconds, and metadata (ratio, resolution, reference media). Confirm reference limits (up to 9 images, 3 videos, 3 audio tracks) and clip length (up to 15 seconds). Failed requests carry no cost, so you can safely retry after fixing the payload.

Confirm the failure is not billed before retrying

A failed Seedance task is never charged, so retrying costs you nothing extra during debugging. According to the official relaydance.com docs, 「Failed or errored requests are never billed」 (据 relaydance.com 官方文档). This means you can resubmit a corrected request without worrying about accumulating charges from earlier errors. Billing on RelayDance is pay-as-you-go per generated video, with live per-model rates published at relaydance.com/models. For reference, Seedance 2.0 720p is about $0.190 / second and Seedance Fast is about $0.152 / second, but neither rate applies to a task that ends in the failed state.

Verify your base URL and authentication

Many silent failures trace back to an incorrect base URL or a missing API key, not the model itself. According to the official relaydance.com/docs docs, 「Set base_url to https://relaydance.com/v1 and keep your OpenAI SDK」 (据 relaydance.com 官方文档). Confirm that requests carry the header Authorization: Bearer YOUR_API_KEY, using a key created at the console. RelayDance is OpenAI-compatible, so the SDK you already use should work once the base URL points to https://relaydance.com/v1. If authentication or routing is wrong, the task may terminate without a descriptive error, so check these two settings first. See relaydance.com/docs for the full request format.

Check reference media and clip limits

A task can fail when reference media or clip length exceed the documented limits. Reference media go in metadata.content[] and are cited in the prompt as @image1 to @imageN. The per-request caps are:

1. Up to 9 reference images.
2. Up to 3 reference videos.
3. Up to 3 audio tracks.
4. Clips up to 15 seconds.

If your request cites more media than the model accepts, or requests a duration above 15 seconds, correct the count in metadata.content[] and the seconds value, then resubmit. Because failed requests are not billed, iterating on these limits carries no cost.

Compare polling and webhook delivery

Choose polling or webhook mode based on how you want the final state delivered, since a missed result can look like a failure.

If you rely on a webhook and never receive the POST, poll the task directly to read the real status. For pricing context while testing, Seedance 2.0 1080p is about $0.470 / second and native 4K is about $4.90 per 5-second clip, per relaydance.com/models.

FAQ

Does a failed task cost anything? No. Failed or errored requests are never billed, so retries during debugging are free of charge.

What endpoint do I poll for status? Call GET /v1/video/generations/{task_id} until the status is succeeded or failed; a succeeded result includes the video url.

How many reference images can one request include? Up to 9 reference images, along with up to 3 reference videos and 3 audio tracks, placed in metadata.content[].