Midjourney Omni Reference

Midjourney Omni Reference is a generative image tool. Midjourney's Omni-Reference system places a chosen character, object, vehicle or creature into new generations using an image reference and an adjustable omni-weight, available through the V7 web interface and the Discord --oref parameter.

Omni-Reference is a Midjourney image reference system that tells the model to place a specific subject into a newly generated image. Where earlier versions relied on character references that mainly handled faces, Omni-Reference is pitched as a broader instruction: put this thing here. The subject can be a person, a piece of clothing, a vehicle, an object or a non-human creature, and the aim is to carry its identity into the output rather than merely borrow its palette or mood. The feature is accessed differently depending on where you work. On the web app, users switch their model version to V7 in the settings menu, drag an image into the prompt bar, and drop it into the bin labelled Omni-Reference; a slider icon sets the strength. On Discord, the same result comes from typing --oref followed by the image URL, with the omni-weight parameter --ow controlling how strictly the reference is followed. Weights run from 0 to 1000, with 100 as the default. Lower values around 25 suit style changes such as turning a photo into anime, while higher values around 400 help preserve a face or an outfit. Midjourney warns that going beyond moderate weights without very high stylize or exp values can make results worse. In practice Omni-Reference is aimed at illustrators, character designers, comic and game artists, and anyone producing repeatable imagery across a set of prompts. It is designed to combine with personalization, stylization, style references and moodboards, though stylize and exp compete with the reference for influence, so high values need a higher omni-weight to balance out. Prompts still matter: a sword only appears if the text says the character is holding one, and low-weight style transfers need explicit description of hair, clothing and other features worth keeping. Users can sometimes place two referenced subjects together by naming both in the prompt, a capability Midjourney describes as somewhat untested. Against alternatives, Omni-Reference stands out because it is native to an existing Midjourney subscription rather than a separate add-on, and it handles objects and creatures as readily as people. The trade-off is that Midjourney itself calls the system experimental and asks users to report what works.