I had no idea that could you do the variations or the brush stuff. Maybe I’m just glossing right over it? But that seems to give the tool more utility. I just try a phrase and I either like it or I don’t.
The fact that you can integrate on it seems to make it much more useful.
Yes! A really cool way to work with it is to generate a bunch of images, arrange them on a transparent canvas (i.e. in Affinity Designer), and then ask Dalle to fill in the gaps.
For example see here[0], where I've combined a picture of a flying whale, a tardigrade in space, and a bunch of flying turtles.
That is neat. So are commas an official way to blend different schools of thought together for the image? Is there any documented way it’s supposed to work? Like [main subject], [art style], etc? Or is it something you picked up from trial and error?
But overall, you have to think about the context Dalle has seen similar images in the training set. If it's seen them on an art sharing site, then it's probably good to mention such sites and tags it could hypothetically have there. Or if it's more like a photo in an article, think about what could be written about it in the article.
You definitely glossed over it - the Dall-E 2 homepage has the three main features of the program with examples (Image Generation, Edits to Existing Images, and Variations of the image)
The fact that you can integrate on it seems to make it much more useful.