Midjourney Starts Testing a Dedicated Edit Model for V8
Midjourney began testing its first dedicated V8.2 image edit model on August 27, 2026, adding a proper editing workflow on top of the V8.2 generation model it shipped earlier in the summer. The update rolled out simultaneously to the main Midjourney site and the alpha.midjourney.com redesign, and the team is actively soliciting feedback on rough edges through Discord.
Details
- Text-instructed editing: users can modify an existing image by describing the change in words, rather than only re-prompting from scratch
- Multi-reference generation: up to four image references can be used at once, replacing the previous single-image omni-reference feature
- Canvas tools: targeted region edits via inpainting, and composition expansion via outpainting
- Style control carryover: personalization profiles and style references (srefs) work with the new edit model, though Midjourney notes moodboards/srefs “may work best with some extra prompt direction”
- Multiple entry points: drag-and-drop into the prompt bar, an Edit button in the lightbox viewer, a dedicated edit tab for uploads, and a Discord
--edit url...command - Staged rollout: available now on both the standard site and the alpha redesign, which Midjourney is using as a testbed for broader UI changes
What happened next
Midjourney framed this as an early test rather than a finished feature, and is collecting community feedback on edge cases through dedicated Discord channels before wider refinement. The move continues a pattern from the company’s recent updates: rather than adding new parameters, Midjourney is pushing image editing toward more visual, direct manipulation instead of memorized prompt syntax.