Lock Brand Consistency Across Visuals With 4-Image Refs

Midjourney's V8.2 Edit Model-released in August 2026 and now the default-supports attaching up to four reference images at once, replacing the old workflow where you'd need separate generations to lock a character's consistency, apply a visual style, and maintain product details. This single-pass approach dramatically speeds up professional creative work where visual consistency matters.

Here's the workflow: upload or drag 4 reference images into the create panel. The first might be a hero product shot, the second a brand color palette reference, the third a previous campaign asset showing the desired lighting, and the fourth a character model you want to carry through. Write a natural-language instruction describing what you want-"create a new product shot with the same lighting, colors, and overall feeling, but showing the item in a lifestyle context." The model processes all four reference contexts together and returns a generation that respects identity markers across all of them.

The practical advantage: a marketing team can now generate 12 localized campaign visuals (6 languages × 2 formats) in a single session, attaching the same 4 brand references and simply changing the text overlay instruction for each locale. Previously, this required checking each result against the original assets and running edits. Now the consistency is baked into the prompt itself.

Style weight is controlled with --sw (default 100, range 0-1000). At higher weights like 500-800, the reference images tightly constrain the output; at 50-100, they provide subtle guidance. For professional asset production, start at 300-400 and adjust based on how much control you need versus creative variation.