Skip to content
eWiki The practical AI reference

How-To Guides

How to Use Midjourney Omni Reference in V7

Two white ceramic bowls with blue bands, one on a white surface and one on a dark stone block.

Midjourney Omni Reference is an experimental V7 image-reference system documented for placing a referenced character, object, vehicle, or non-human creature in a generated image. This guide explains the official web and Discord workflows, Omni Weight, and cautious ways to adjust competing controls.

What Midjourney Omni Reference Does

Midjourney introduced Omni Reference on May 1, 2025, as an experimental image-reference system being tested with V7. The provider describes its purpose as putting a referenced subject into a generated image and says the subject can be a character, object, vehicle, or non-human creature. The announcement also says the system can reproduce much of the earlier Character Reference functionality from V6, while supporting additional kinds of subjects. These descriptions establish the intended reference behavior, but they do not guarantee that every result will preserve identity, appearance, or recognizability. See Midjourney’s Omni Reference announcement for the documented scope.

As a practical suggestion, choose a reference in which the subject and the features important to your concept are easy to identify. Treat the reference as one input to generation rather than evidence that a particular output is assured.

Black geometric sculpture beside calipers and a folded blue cloth, with printed reference photos on a worktable.

Use Omni Reference on the Web or Discord

The documented web workflow begins by selecting V7 in Midjourney’s settings. Drag the reference image into the prompt bar, place it in the area labeled “Omni-reference,” and use the slider control to set its strength. On Discord, the documented syntax is --oref followed by the URL of the reference image; --ow controls reference strength. Both workflows appear in the official usage instructions.

Midjourney advises describing the intended relationship or action in the prompt instead of relying on the reference alone. Its example says that when a character should hold a referenced sword, the prompt should explicitly describe a character holding a sword.

As an editorial suggestion, start with one subject and a plainly described scene. This can make your own visual comparison easier, although it is not a provider guarantee of better output.

A person's hands placing a red apple and a metal key beside a blue book on a wooden table, with a reference photo behind them.

Start With the Default Omni Weight

The --ow parameter determines how strictly the generated image adheres to the reference. Midjourney documents a range from 0 to 1000 and identifies 100 as the default. For a substantial style change, such as moving from a photograph toward anime, the provider gives --ow 25 as an example. For stronger emphasis on a visible face or preserved clothing, it suggests trying a higher value such as --ow 400. These are examples in the official weighting guidance, not guaranteed presets.

A sensible testing suggestion is to begin at 100, review the result, and change the weight in measured increments. Consider a lower value when the reference influence appears too restrictive for the intended transformation. Consider a higher value when features you want emphasized are not sufficiently reflected. These are editorial recommendations for experimentation; any effect remains conditional on the prompt, reference, and other settings.

Account for Other Controls and Prompt Details

Midjourney says --stylize and --exp compete with Omni Reference for influence over an image. Its guidance says that high values for either may require a correspondingly higher Omni Weight. The provider also cautions that going above moderate values such as --ow 400 may make results worse unless extremely high --stylize and --exp values are involved. Omni Reference is also described as working with personalization, stylization, Style References, and moodboards. These documented interactions are summarized in the Midjourney announcement.

For low-weight style transfer, Midjourney recommends specifying the character details that should remain, illustrated with hair color and clothing details. As a practical suggestion, change one parameter at a time and record the prompt and values used. That approach can help you interpret your own comparisons, but it does not establish that one control alone caused every visible difference.

Experimental Status and Current Use

The May 2025 announcement explicitly characterizes Omni Reference as experimental and asks users to provide feedback and examples. It also describes working with two subjects from one combined reference as somewhat untested, so that possibility should not be presented as dependable behavior. The supplied official documentation establishes a V7 testing workflow at the time of publication; it does not establish a guaranteed outcome for every image or setting. Consult the dated provider announcement before treating its interface details as current.

As a suggestion, first check your Midjourney interface and official documentation for the model and controls presently offered. If the documented V7 Omni Reference controls are available, use the workflow above; otherwise, follow the provider’s current interface rather than assuming the 2025 steps still apply unchanged.

Conclusion

Midjourney’s documented Omni Reference workflow uses a reference image, a scene description, and an adjustable strength control in V7. The official starting value is --ow 100, while lower and higher examples illustrate different degrees of reference adherence. Midjourney also warns that --stylize and --exp compete for influence and describes the feature as experimental in its May 2025 announcement.

The practical recommendation is to confirm that the documented controls exist in your current interface, begin with one clear reference and an explicit prompt, and start at the default weight. Then adjust one setting at a time while judging the outputs for your particular subject and scene. Use the published weight values as trial points, not promises of subject preservation or recognizability.

Sources

Disclosure

This article was prepared with AI assistance.

Improve this guide

Report a problem or suggest the next question

Submissions enter a private editorial queue. They are reviewed and deduplicated before any public update or new page is created.