#Error When Using OpenAI Edit Images Endpoint

1 messages · Page 1 of 1 (latest)

limber moat
#

I'm encountering a 400 error when trying to use the openai.images.edit() endpoint with the gpt-image-1 model. The error message is:

```
Error editing image: Error: 400 Your request was rejected as a result of our safety system. Your request may contain images that is not allowed by our safety system.
```

Here are some relevant console logs:

```
✓ Compiled /api/generate-image in 326ms (1294 modules)
Making API call to GPT-Image-1 with:
{
prompt: 'A youtube thumbnail for a green crow in 1280x720 resolution - Style: realistic, Mood: dramatic',
size: '1536x1024',
quality: 'high'
}
Response received. Base64 data extracted successfully.
Image uploaded to S3.
Saved thumbnail to database with ID: cma6ft0sv0005xufbhh3gwm2x

Calling edit endpoint with:
{
hasImage: true,
imageLength: 1048576,
imagePrefix: 'data:image/png;base64,iVBORw0K',
editPrompt: 'add text',
quality: 'hd',
size: '1536x1024'
}

Error:
Error editing image: Error: 400 Your request was rejected as a result of our safety system.
code: 'moderation_blocked'
request_id: 'req_d364b3df592758ca5ac4d1092acfa16d'
```

The image was generated using the same model (gpt-image-1) and the only edit prompt used was "add text". It's unclear why this would trigger the safety system.

Any insight into what might be causing the moderation block when passing a freshly generated image back into the edit endpoint would be greatly appreciated.

Thanks in advance! 🙏

limber moat
half dawn
# limber moat <:PepeTired:590795745553285120>

the edit api is more strict with stuff vs new generations. it does not matter where the image came from. similar happens sometimes with the regular chatgpt in a new thread and only way seems to be is to say it to just imitate/mimic the original pic

limber moat