I don’t know why, but today dall-e is more restrictive than 2 days ago. If I gave a prompt from an image rendered 2 days ago and ask to rerender the same prompt in a new chat, it say now that the prompt is now conflicting with the policies.
If if try the same exact prompt in the same chat used to create it at the first time (chat from past monday) it render it directly. I’m really confused with the policies. Any info ?
It’s a hit or miss. The content policies are strong on this one, almost too strong.
Your issue is probably context. It probably looks through your context and goes like “ah yes, this is acceptable”, while in another chat with another context, it understands that it shouldn’t do it.
Basically gaslight it and it’ll work half of the time, every time.
i didn’t had any specific context the first time except the discussion with chat gpt who help me to create the prompt initially. I’ll retry, but i’ll also keep the history, just in case 
Always ask GTP for the exact JSON data it is sending to Dall-E and those exact data he is retrieving from it.
Ask to show all parameters, formatted in a JSON object / array.
This way you can exactly see what text is send and received (and eventually what went wrong).
The next time, you can use that exact same prompt and make minor adjustments to it.
I did set this in my custom instruction set to it goes automatically (well sometimes he forgets about it and I have to remind him).
