Edit Model for V8

Midjourney is opening up testing of a new edit model built specifically for its V8 pipeline, currently labeled V8.2. The rollout is available to all users, positioning it as a broad public test rather than a limited alpha. The model is designed to bring several image manipulation capabilities into one unified system tied to Midjourney's latest generation quality.

The core features include instruction-based image editing, where users can describe changes in plain language and have the model apply them to an existing image. Also included is a multi-image reference mode that replaces the older omni-reference feature - users can now supply up to four separate reference images at once to guide generation. The third capability is inpainting, which allows targeted edits to specific regions of an image while leaving the rest intact.

Bundling these tools under the V8 model matters because editing quality is closely tied to the underlying generation model. Earlier editing features in Midjourney were built on older model versions, which could mean inconsistencies between freshly generated images and edited ones. Aligning the edit model with V8 should bring more coherent results, particularly for users who want to iterate on V8 outputs without a visible quality gap.

The move also signals that Midjourney is working to close the gap with other platforms that have offered robust in-image editing for some time. With up to four reference inputs and inpainting now available alongside text-driven edits, the V8.2 edit model gives users a more complete toolkit for post-generation work. As this is still a test phase, further refinements to the model are expected based on user feedback.