A generation failed or was blocked
If a generation fails or is blocked, the credits are returned to your balance automatically (and if a job was cancelled before it started, nothing was charged). Below are the most common messages and what to do when you encounter them.
“This request was blocked by the content safety filter”
The AI model’s safety system flagged your prompt or image. This can happen with violent, sexual or hateful content, and sometimes with harmless requests that look similar.
- Rephrase the prompt and remove words that could be misread.
- Try a different image, or a different model — each model has its own filter.
Related messages work the same way: “The uploaded image was blocked by the content safety check” (use a different image), “Your prompt was blocked by the content safety check” (reword the prompt), and “The model declined this result because it looked too close to copyrighted material” (avoid well-known characters, logos and brands in the prompt or references).
“The model declined to generate a result”
The model refused the request, most often because the input contains photos of real people or minors. Try different images or prompts.
Photos of people are not accepted
Some models don’t accept uploaded photos or videos of real people, and AI Studio can’t process images that include faces of real people with private information. If you see “[model] doesn’t support directly uploaded images of people. Use an original image generated by Seedream, or choose another AI model.” (or the same message about videos):
- Choose Create an image with Seedream to make the person with AI first and use that image, or use a ready-made Character.
- Choose another model. For Seedance videos, Continue with Kling retries the same request with Kling 3.0.
“The AI model provider ran into a temporary problem”
The model had a temporary outage. Wait a few minutes and try again, or switch to another model.
Other messages
| Message | What to do |
|---|---|
| Not enough credits | Your balance is lower than the cost shown. Lower the length, resolution or number of outputs, or add credits. |
| AI model credits have changed | The price was updated while the page was open. Check the new price and press Generate again. |
| Temporarily disabled for maintenance | The model is paused for a short time. Use another model or try again later. |
| Rate limit exceeded / try again later | You sent too many requests in a short period of time. Wait a few minutes and try again. |
| Your cloud storage is full | Delete older creations or uploaded files, or subscribe or upgrade your plan for more storage. See How long are my files kept? |
| This job was cancelled after waiting too long to start | No credits were charged. Try again. |
| No speech was found in this file | Voice and subtitle tools need a file with spoken words. Use a different file. |
| This model is temporarily unavailable in your region | Choose another model. |
| The reference videos and the result together are longer than this model allows | Shorten the reference videos or the result length (Wan 3.0 allows 30 seconds in total). No credits were used. |
It’s taking a long time
Videos, especially long or high-resolution ones, can take several minutes, and longer when a model is busy. Once a generation has started, you can close the page — the job keeps running and the result appears in My Creations. In Super Director, keep the conversation open until every clip has started. Super Narrator’s local rendering needs the tab to stay open; download the finished MP4 before you leave. If a job never finishes, it eventually fails and the credits are returned.
Still not working?
Contact support with the Job ID from My Creations › Activity Log and the exact error message.