Midjourney opened testing of its first image editing model for version 8.2 to all users. It edits existing images from written instructions, generates new images from as many as four reference pictures, and changes selected areas or extends the canvas. The company describes it as an early test and is asking users to report cases where results are wrong.
What changed
Image-to-image work relied on omni-reference, which took a single reference image.
What it unlocks
Editing an existing image by written instruction, and generating from up to four reference images at once.
- up to 4 image references at once
What you need to act on it
- a Midjourney account
- access via the web app or Discord
Sources