Skip to main content

Why did my generation fail?

Every generation in Studio, the editor, and Workflows goes through the same pipeline, so the same handful of causes explain almost every failure. Failed items stay in your gallery — open one to see the error that stopped it, then match it against the causes below.

Not enough credits​

Generation is the main thing credits pay for, and a request that would overdraw your balance is refused up front rather than charged and failed.

  • Your own balance is empty. The composer shows the estimate ("~30 credits") next to your remaining balance before you click Generate. Top up or wait for your monthly reset — see How credits work and Top-ups.
  • You're working in a shared project. Work billed to a project is paid by the project owner's credits, not yours. If the owner's pool runs dry you will see that the project owner does not have enough credits — even if your personal balance is healthy. The owner needs to top up; see Pooled billing.

The model needs a purchase​

Models marked Pro in the picker (Seedance 2, Kling 3.0, and friends) unlock once the paying account has bought anything — starting one before that returns "This model unlocks with your first top-up or plan." Either switch to a non-Pro model in the picker or top up (from $5) or subscribe in Settings → Usage. See Plans & pricing and Choosing a model.

The model can't accept your attachments​

Not every model consumes reference images, first frames, or video input. When your prompt includes an attachment the selected model can't use, the composer warns you before you generate and offers a one-click swap to a model that can. Take the swap, or remove the attachment. Details in Mentions & references.

It's been queued for a long time​

Queued means the request is accepted and waiting on the model provider. Provider-side congestion happens, especially on newer video models.

  • Leave it — queued work completes without you watching it, and you can keep editing meanwhile.
  • Or click Cancel on the queued item and retry, possibly on a faster model variant.
  • If every generation queues indefinitely, the provider may be having an incident; trying a model from a different family is the quickest check.

The provider rejected the content​

Model providers apply their own content rules, and some are stricter than Corte's. Rejections commonly involve real people's faces, brands, or content the provider considers sensitive. Rephrasing the prompt, swapping the reference image, or trying a different model usually resolves it.

The agent paused mid-task​

The AI editing agent spends credits too — both for its own turns and for any generation tools it calls. If it stops with a credit error, the fix is the same as above: top up, or (in a shared project) have the owner top up. The agent picks up where it left off when you ask again.

Still failing?​

If a generation fails repeatedly with the same error and none of the causes above fit, contact us with the failed item's prompt, model, and rough time — see Getting support for what to include. The failed attempt itself costs nothing: you're only charged for work that completes.