Generation Failed
When a generation fails, modelBridge shows a clear error message explaining what went wrong and what to do next. Here are the most common causes and how to handle them.
Common Reasons a Generation Fails
Section titled “Common Reasons a Generation Fails”The model is temporarily overloaded
Section titled “The model is temporarily overloaded”AI models run on shared GPU infrastructure. During peak hours, a model may not have enough capacity to process your request.
What to do:
- Wait a minute and click Retry
- If it keeps failing, try a different model — similar models often run on different hardware
Your input doesn’t meet the model’s requirements
Section titled “Your input doesn’t meet the model’s requirements”Each model has specific requirements for the media you send — minimum resolution, maximum duration, supported formats, and more.
What to do:
- Read the error message — it tells you exactly what the model needs
- Adjust your clip or image to match (e.g., use a higher-resolution source)
- modelBridge remembers these requirements, so the same issue won’t cost you money twice
Network or connection issue
Section titled “Network or connection issue”Your connection to fal.ai dropped during generation, or fal.ai’s servers returned a timeout.
What to do:
- Check your internet connection
- Click Retry — network issues are almost always temporary
API balance or authentication
Section titled “API balance or authentication”Your fal.ai account may have run out of credits, or your API key may have expired.
What to do:
- Check your balance at fal.ai/dashboard
- Top up your account if needed
- If your API key expired, generate a new one and update it in modelBridge (API & Support tab)
Content policy
Section titled “Content policy”Some models have built-in content moderation. If your input is flagged, the model will refuse to generate.
What to do:
- Review your prompt and input media
- Adjust and try again
Dual Mode — When One Model Fails
Section titled “Dual Mode — When One Model Fails”In dual mode, two models generate from the same input simultaneously. Because each model has different requirements and runs on different infrastructure, one model can succeed while the other fails.
Why this happens in dual mode specifically
Section titled “Why this happens in dual mode specifically”- Different input requirements — Model A may accept your 480p clip, but Model B requires at least 720p
- Different availability — One model may be overloaded while the other has capacity
- Different processing limits — Your 30-second clip may be fine for one model but exceed the other’s maximum duration
What you can do
Section titled “What you can do”When one model fails in dual mode:
- Use the successful result — click Import to timeline or Save selected to keep the result that worked
- Retry the failed model — click the Retry button to try again
- Close the preview — click the X button on the preview panel to dismiss it
The successful result is fully usable regardless of the other model’s failure. You don’t lose anything.
Retrying a Failed Generation
Section titled “Retrying a Failed Generation”You can retry any failed generation directly from the preview panel — no need to set everything up again.
- In single mode: click Retry in the error message
- In dual mode: click Retry [Model Name] to retry just the failed model, or Retry both if both failed
Retrying uses the same prompt, settings, and source media as the original attempt.
Still Not Working?
Section titled “Still Not Working?”If a model keeps failing after multiple retries:
- Try a different model — browse the catalog for alternatives
- Check fal.ai status — visit status.fal.ai for platform-wide issues
- Report the issue — go to API & Support tab in modelBridge and use the error reporting tool