AI image generation content policy violation warnings

[dosmore]

Hi,

I'm using the AI image generator to create illustrations relating to cleaning services, but I'm getting frequent content policy violation warnings, even though my prompts seems very innocuous and don't seem to violate any of Openai's guidelines. Here is an example prompt that was flagged:

line drawing of a professional cleaner in a large residential shower room scrubbing the shower controls


If anyone can advise me where I might be going wrong or what policy I might be violating, I'd be very grateful. Thanks in advance.

[Pablo]

The software does not have any prompt restrictions. Everything is directly send and receive via the OpenAI server. Technically, all that WWB does is provide the input and output boxes.
So, if you think the error is wrong then you will have contact OpenAI about this.

[gofrank]

Try adding "uniformed" in your prompt, as in:

line drawing of a uniformed professional cleaner in a large residential shower room scrubbing the shower controls

This produced four images for me with no problem.

[onlye]

While I have been playing around I have gotten similar errors on request that I can't even imagine how it could be misinterpreted. Guess this is part of being an earlier adopter.

[Pablo]

While I have been playing around I have gotten similar errors on request that I can't even imagine how it could be misinterpreted. Guess this is part of being an earlier adopter.

Please note that these messages have absolutely nothing to do with WYSIWYG Web Builder.
All content and error messages are generate by OpenAI.

[wwonderfull]

As @pablo said I think he is totally right and I understand him. This AI feature is more like an internal browser or a frame which is bringing the results and most importantly people should know that WWB is not the AI it is using the API to connect the AI features to the software like a modular browsing unit so the results can be utilized in the builder. So the AI feature we have is a modular feature which is fully dependent on internet, Open AI and the subscription the user has and how they are using it.

But if the software would have been having the problem then it would have been a bug not an error which has not been reported or found because all the features are working and if it does not work then it is only when the open ai server which creates the problem depending on certain services for the user.

[onlye]

While I have been playing around I have gotten similar errors on request that I can't even imagine how it could be misinterpreted. Guess this is part of being an earlier adopter.

Please note that these messages have absolutely nothing to do with WYSIWYG Web Builder.
All content and error messages are generate by OpenAI.

Sorry, I should have been more specific - I got similar warnings working directly with the AI (outside of Wizzy). No, this is not a Wizzy thing; it is totally within the AI.

[BaconFries]

Using your prompt via Bing and GPT-4 gave the following return. But with what you are saying about the error (unrelated to the software) it does seem a issue with in OpenAI.

[dosmore]

Many thanks for the tips. I completely recognise that the content violation errors are nothing to do with WWB. Openai do state that Dall.e does err on the side of caution, but I still can't think why my specific prompts would generate a caution (if I re-run the same prompt, sometimes it's ok, other times it isn't). Thanks again.