Most generation failures fall into a small set of causes.
“401 invalid_api_key”
The API key for the provider is wrong, expired, or revoked.
- Open Settings → AI Models.
- Click the eye icon next to the failing provider’s key and verify.
- If wrong, re-copy from the provider’s dashboard and paste.
- Click Save, retry.
“429 quota_exceeded” or “rate_limit_exceeded”
You hit a rate limit or spent your quota.
- Short-term — wait a minute and retry. Most rate limits reset quickly.
- Quota exhausted — top up at the provider’s billing dashboard. The app cannot do this for you.
- Repeated rate limits — slow down. Pause-and-batch helps.
“Model not found” or “model unavailable in your region”
The selected model is not available to your account or in your region.
- Pick a different model from the picker.
- For some providers (Kling Pro tier, certain Google Veo models), region restrictions apply. Check the provider’s docs.
Video generation hangs
Some video providers (Kling) queue jobs. The “Generation History” panel shows the queue state. Wait — typical queue times are 1–5 minutes. If a job is queued for over 20 minutes, it is probably stuck on the provider’s side; cancel and retry.
Generated images are all black / corrupt
Usually one of:
- Out-of-quota mid-render — the provider returned a partial file. Check billing.
- Model glitch — re-roll usually solves.
- Corrupted reference image — try without the reference.
Cost surprise
Each panel shows an estimate before you generate. If a model’s real bill is much higher than the estimate, check the provider’s dashboard for unusual usage and compare against the in-app generation history.
When all else fails
Open server.log (see Backend Will Not Start for the log locations) and copy the relevant lines into a bug report. AI provider errors are passed through verbatim, so the log usually contains the full provider response.
Leave a Reply