It’ s not only the API I am referring to here, but in ChatGPT using Dall-E 3 gives you a picture with a lot more information that the user didn’t mentioned in his prompt (see picture).
In the documentation for Dall-E 3 they mentioned the use of SSC and DSC. Is this the result - making each user prompt more detailed with this method?
Hey there and welcome to the community!
So, if I’m understanding this correctly, you’re noticing the prompt given by the user and the prompt that was actually used to generate the model was different, right?
This is expected behavior. The model will typically modify the prompt before sending it to DALL-E in ChatGPT unless you specify to use theprompt verbatim (which I can’t even verify if that works anymore).
Thank you! 
Yes, you are right. I was just wondering, if it has something more to do with the SSC and DSC, which are described in the dall-e-3 research paper:
I just thought, the process of adjusting the given prompt and modify it in ChatGPT looks a lot like the SSC and DSC process mentioned in the paper and I wanted to know, if these are one and the same thing!
