Troubleshooting Failed or Stuck Video Generations

A generation that fails or sits in "processing" forever is frustrating, especially when credits are on the line. Here is how to read what happened, and what to do in the right order — most cases resolve themselves or cost you nothing.

First: what does each status actually mean?

• Queued / processing — the task is in line or rendering. Rendering a 15-second video typically takes a few minutes; queues at busy times can push it longer. Give it time before assuming failure.
• Failed — the renderer or the safety review stopped the task. Your credits are returned to your balance automatically, every time, with no action needed from you.
• Succeeded but the video looks wrong — the pipeline worked; the inputs or template did not. This is a quality issue, not a failure, and the fix is regeneration with adjustments.

My task failed instantly. Why?

Instant failures are almost always the safety review rejecting the prompt or a photo before rendering starts. Common triggers: realistic photos of identifiable people in sensitive contexts, and prompts asking for content the models refuse. Change the prompt or swap the photo, then retry — the rejected attempt already refunded itself. The details are in the photo privacy and safety guide.

My task failed after rendering started. Why?

Mid-render failures usually mean the model choked on the inputs: a very low-resolution or corrupted photo, an extreme aspect ratio, or a temporary overload on the provider side. Check your photos against the photo selection guide, regenerate, and if it fails again immediately, wait a few minutes — provider outages come and go.

My task has been processing for 15 minutes. Is it stuck?

Maybe, maybe not — queues are real. Wait up to about 30 minutes before treating it as stuck. If a task eventually times out on our side, it is marked failed and the credits come back automatically. You can check your balance to confirm the refund in the credits guide.

The smart retry loop

When something fails or looks off, work from cheapest to most expensive:

1. Retry as-is. Provider variance is real; the same inputs sometimes succeed on the second try.
2. Retry with adjusted inputs. Better photo, calmer prompt, or a 480p draft to isolate the problem — a cheap test tier exists exactly for this.
3. Escalate with the task ID. If three attempts with clean inputs all fail the same way, contact support with the task ID from your dashboard. That pattern usually means something on our side worth fixing.

What never costs you credits

• Safety review rejections — auto-refunded.
• Renderer crashes and timeouts — auto-refunded.
• Provider outages — tasks fail closed and refund.

You only ever spend credits on videos that actually render. When yours succeeds, open the generator is where the next attempt starts.

Frequently asked questions

Where do I find my task ID? In your account dashboard, on the task entry — it is included in any support request so we can trace exactly what happened.

Will retrying the exact same inputs give the exact same result? No. Outputs vary between attempts, which is why retrying is a legitimate strategy, not a coin flip you already lost.

Can a generation partially succeed and still charge me? No. Credits are deducted only when a video is produced; every failure path refunds in full.