Hi,
I can load and describe what’s in an image.
If there is an area of particular interest in the image I can use an image editor and draw an outline around that area of interest and prompt “tell me what’s in the highlighted area”.
Is there a way to tell openAI to create a mask image by prompting openAI to delete the highlighted area to use as a mask?
What you describe is not possible at this time.
It is good to understand individual scopes of imagery functions and features they furnish — and how the two do not meet or combine to provide any aspect of image revision or alteration.
gpt-4-vision:
Uses computer vision to identify and describe imagery.
DALL-E 3:
DALL-E 3 turns words into visions, generating scenes that favor a computer-rendered appearance.
DALL-E 2 via images API
EDITS endpoint:
Allows for transparent areas and (almost completely ignored) prompting, where imagery can be filled by AI where missing. Strictly square images @ 1024px.
(no sunbathing woman created per prompt…just infill and outfill)
Useful here for just fixing my deleted third chair…
Thank you for the response.
I am performing the describe and mask functions successfully. I was trying to see if I could combine the functions. That is, instead of deleting part of the image to create a mask image, just create an outline and tell AI that the inside of the outline is the mask.
Just a way to save an editing step and only having to pass one image in the post.
Dave
1024x1024 ok. What image formats are ok ? .jpg .png .bmp .gif ?
Not possible at this time …