content_policy_violation in DALL-E 3 API For Non-English Prompts

API
1

Thank you for announcing the DALL-E 3 API.
however, I encountered a problem with content_policy_violation when using it, sometimes the error is triggered when entering the Chinese dalle prompt

{“error”:{“code”:“content_policy_violation”,“message”:“Your request was rejected as a result of our safety system. Your prompt may contain text that is not allowed by our safety system.”,“param”:null,“type”:“invalid_request_error”}}
or
{“error”:{“code”:“content_policy_violation”,“message”:“This request has been blocked by our content filters.”,“param”:null,“type”:“invalid_request_error”}}

When I meet content_policy_violation ,i try again with the same Chinese input, sometimes the call succeeds but does not trigger the content_policy_violation error
i notice you fix for the issue of non-English prompts being flagged incorrectly for safety(Concerns Over Stringent Content Policy Blocks in DALL-E 3 API, Especially For Non-English Prompts - #4 by owencmoore)
I guess is that the image generated by dalle is prompt1-prompt2-image.
my prompt1 is Chinese,my prompt2 is generated by gpt, maybe prompt2 triggers content_policy?

How can I avoid this error?
Looking forward to your reply, thank you！

2

That could be it. Is there any way to translate your Chinese prompt to english and check it then send?

4

Thanks for replying, i’ll give it a try

5

Thank you for your suggestion. However, this adds an extra step to my process and may not always accurately capture the nuance of my original Chinese prompt. Is there a way to adjust the content filters to better accommodate non-English prompts, or perhaps a mechanism to review and approve prompts that are mistakenly flagged? I appreciate your help and look forward to further improvements in the DALL-E 3 API.

6

With more than 100 million active users and billions of messages, a manual review process like that would not be practical. Whatever solution you implement will have to come from your end.

7

I’m not looking for manual review, but I will try to find solutions on my end to ensure that they adhere to the content policies while still conveying the intent I aim for. Thank you for your suggestion.