Dall-e-3 Getting 'content_policy_violation' error for a simple prompt

I’m getting 400 error code with ‘content_policy_violation’ for a simple prompt e.g.
Build an image for Back on the Beach Café at location in santa monica,CA US

Getting the message:
400 : "{ “error”: { “code”: “content_policy_violation”, “message”: “Your request was rejected as a result of our safety system. Image descriptions generated from your prompt may contain text that is not allowed by our safety system. If you believe this was done in error, your request may succeed if retried, or by adjusting your prompt.”, “param”: null, “type”: “invalid_request_error” }}

Any better way to prompt? Many other places with same prompt is working fine but not this one.

1 Like

Welcome to the forum.

Did you try it again as suggested? That’s more of a “soft-filter”… Your prompt is rewritten on OpenAI’s side, so with such a short, generic prompt, it’s likely adding words that might get you in trouble…

My advice would be to me a lot more descriptive… i.e.

Create a vibrant and inviting image of ‘Back on the Beach Café,’ situated on the picturesque Santa Monica, CA, beachfront. The scene captures the essence of a relaxed beachside atmosphere, with the café nestled directly on the sand, offering panoramic views of the Pacific Ocean. Include details like outdoor seating, patrons enjoying their meals, and the iconic Santa Monica Pier in the background, subtly blending into the scene. The composition should convey the warmth and casual elegance of this beloved spot, highlighting its status as a perfect place to unwind and savor the coastal vibe.

Thank you so much! In my case, I don’t have this amazing description you’ve provided because everything happens on the runtime, I only know the name and address of the place. Is there a generic prompt that can be used where it doesn’t result in content_policy_violation?

1 Like

You could fine-tune a smaller cheaper model or even use GPT-3-turbo-instruct to change your short prompt into something better. It would still be GPT changing your prompt, but you might have more control? Or you could add a bit more wording along with your “short prompt”? That’s likely what’s happening, though. I run into it now and then.

I added this prompt to my prompt, and it solved the problem for me: “Avoid any content that may be considered inappropriate or offensive, ensuring the image aligns with content policies.”

1 Like

Nice. Clever. Simple is best sometimes. Let us know how it goes.

We’ve got a wealth of dalle3 threads here if you stick around. Occasionally, we’ve had members of the DALLE team drop in. There’s an AMA they did on Discord that’s archived here as well.

Welcome to the community. Hope you enjoy your stay!