Deleted, because of censorship.
That is interesting. Thanks for sharing, I will have to do some testing
Let me test this
It really works.
You can also generate four images and pick the best one.
Ask which seed is used for that particular image and which prompt and dimensions.
You will retrieve the exact JSON formatted API call from ChatGPT to Dall·E3.
Then you can copy / paste that code and request to build an image with it.
The image will be the one you picked before (there can be some non visible numeric noise, but it is an 99.99% match).
Request the same image with the same code, but a different seed.
Than you see a variation on the previous image.
If your are known to the JSON object array (jquery - How to write a nested multi dimensional json object - Stack Overflow), you can format it with more prompt and seeds of code.
Development meets design.
I figured it out by this thread, which is updated already;
Nice. Thanks for sharing. I remember them mentioning during Alpha that there was a single seed being used… good to know we can change it now.
Who is this “we” that you mentioned here?
LoL at last sorry for earlier, I think community new a better delete:wastebasket:
“we” could apply to anyone using DALLE-3. Please provide more context to your question as I do not believe I fully understand it.
Okay, He do not have any questions, he is sharing a way to generate a consistent images. Though i am not sure whom you are referring this me or foo-bar.