I’ve been using Designer to experiment with images, and I’ve found it particularly impressive when you feed it as much detail as possible. Asking Designer for “an image of a dachshund sitting by a window staring at a slice of bacon” generates some good examples, but you can get Designer to do much more with some additional prompting. Adding in more descriptive language to generate a “hyper-real painting” with “natural lighting, medium shot, and shallow depth of field” will greatly improve image results.

As you can see in the two examples below, Designer gets the natural lighting correct, with some depth of field around the bacon. Unfortunately, there are multiple slices of bacon here instead of just one, and they’re giant pieces of bacon.

Like most things involving AI, the Designer feature isn’t perfect. I generated another separate image of a dog staring at bacon, and a giant piece of bacon was randomly inserted. In fact, I’d say most times only one or two of the four images that are produced are usable. DALL-E 3 still struggles with text, too, particularly if you ask Designer to add labels or signs that have text written on them.

It did a good job of “an illustrated image of a UPS delivery man from 1910. In the style of early Japanese cartoons,” though, adding the UPS logo in — even if it’s a little wonky. Copilot Pro lets you generate 100 images per day, and it does so much faster than the free version.