Dall-e3 can now read PDF, see photo's and use them to generate new ones

You can now do this (without switching models);

  1. Upload a sketch
  2. Tell Dall-e3 to reproduce it
  3. Let ChatGPT describe the output
  4. Gather data from the created image

Or at least, they are rolling out this feature.

And because Dall-e3 can now “see” it’s own creations, it’s also capable to fine-tune some details inside the image much better.