Midjourney Now Lets You Request Edits to AI-Generated Images

pcmag.com:

Midjourney has released a new editing option that allows people to modify a portion of generated art rather than having to create a new piece, a process known as "inpainting."

It's one of the biggest updates to its AI image generator since the platform's debut in July 2022, and it's winning rave reviews on the Midjourney Discord server. "This is the greatest addition to Midjourney since...forever," one user writes. "This is one of the coolest product releases I've seen in any product in the past five years," another adds.

As a text-to-image platform, Midjourney converts written prompts like "fish jumping out of a lake" into high-quality images. Now, by selecting a new "Vary (Region)" button, users can enter a new prompt to customize one piece of the already-generated artwork: "shark jumping out of a lake."

One of the most common uses for Vary Region will likely be correcting misshapen hands and limbs, something AI image generators struggle with, as BuzzFeed notes. "I can now fix those extra limbs that pop up so often," one user writes on Discord. Others reference fixing people with six fingers or more than two hands.

Vary Region will also help with iterating on an image and trying out new combinations. This could include "adding a top hat on a character," Midjourney suggests. One user calls it "invaluable for character portraits of the same person with differing expressions/emotions."

There are a few limitations. Midjourney says the feature works best on "large areas (20% to 50% of the image)" and for small prompt tweaks. "Changing the prompt will work best if it's a change that's more matched to that image versus something that's extremely out of place (a dolphin in a forest)," it says. One user also noted the feature