I want to generate an image of a person given the face and for rest of the feautres like body shape, clothing, attire, color, etc by a prompt in a photo realistic way.
Using dalle2, i am only getting image editor to cleanup the image and not more.
You would use the edits endpoint. You send a 1024x1024 transparent 32-bit (24 bit with alpha) PNG image, where the area for the AI to extend is made completely transparent (or you can use a transparent mask).
Give lots of transparent room for what you want filled in by AI as it doesn’t resize.
$0.02 to see the AI not do an awesome job of following your prompt.
@_j, do you happen to catch if they said anything about dall-e-3 coming to edits or variations endpoint soon?
Nothing was discussed during the keynote about modifications to edits or variations.
We can see in the new API documentation dump for edits:
model: string
Optional
Defaults to dall-e-2
The model to use for image generation. Only dall-e-2 is supported at this time.