Troubleshoot generation failures

Start with the status and error shown on the affected node or run. It is the best indication of whether Fuzzbucket rejected the request before it started, a provider failed while processing it, or the result is still pending.

Common causes

Generation can fail because:

• the selected model does not support the requested input, format, duration, or aspect ratio;
• a required source image, video, or public media URL is missing or unavailable;
• the model or provider is temporarily unavailable;
• a provider or moderation system rejected the request;
• the account does not have enough credits for the run;
• the browser connection was interrupted while the run continued elsewhere.

An interrupted browser request does not prove that the provider stopped. Check the node or run status again before starting a duplicate generation.

After a disconnect or reload

Return to the same workspace and allow its active run state to recover. Check the existing node, progress, or error before sending the prompt again. The recovered run can continue to deliver progress or output under its original identity; submitting a duplicate can spend credits and create competing results.

Retry safely

Before retrying:

1. Confirm that the previous run is failed rather than queued or processing.
2. Recheck the prompt, references, and any required source media.
3. Confirm that the model supports the selected media type and settings.
4. Check the visible credit balance and estimated cost.
5. If the request is complex, try a smaller or shorter test first.

Avoid repeated expensive retries when the same validation or provider error keeps returning. Change one relevant input at a time so you can tell what resolved the problem.

Credits changed but no usable result appeared

Capture the workspace name, approximate time and timezone, media type, model if visible, final status, and the expected versus observed credit change. This gives support enough context to trace authorization, settlement, and provider behavior.

See the Media generation overview for the normal workflow, Models and generation settings for capability checks, or Report a problem when the failure remains unresolved.