Key takeaways
- Use one clean reference image that shows the subject’s important facial, clothing, shape or material details.
- Repeat a short identity description in every prompt instead of expecting the reference image to carry every detail.
- Start Omni Strength near its normal setting, then raise it in small steps only when the subject changes too much.
- Keep the reference, aspect ratio and core prompt stable while testing one new scene or setting at a time.
Quick answer
For anyone searching “midjourney omni reference how to use,” upload a clear reference image, assign it as the Omni Reference, then describe the new scene while repeating the subject’s defining traits. Keep the same reference and core description across generations, and adjust Omni Strength gradually rather than changing several settings at once.
Midjourney Omni Reference: what does it actually do?
Omni Reference helps Midjourney carry a person, character or object from one generated image into another. You give Midjourney one source image, then ask for the same subject in a new setting, pose, outfit, camera angle or art style.
It is useful for recurring characters in story illustrations, product concepts, social posts, thumbnails and client mockups. It can preserve recognizable features, but it does not copy every pixel or create a fixed digital model of the subject.
Omni Reference is broader than a style reference. A style reference guides features such as color, lighting and visual treatment, while an Omni Reference tells Midjourney what subject should appear.
| Reference type | What it guides | Best use |
|---|---|---|
| Omni Reference | Person, character or object identity | Keeping a recurring subject recognizable |
| Style Reference | Color, texture, lighting and visual treatment | Matching a consistent illustration or photo look |
| Image Prompt | Composition, content and general visual ideas | Borrowing several features from an image |
Results are strongest when the source has one obvious subject. A crowded photo forces Midjourney to guess which face, outfit or object matters.
Access requires a paid Midjourney plan. The entry plan is around $10 per month at the time of writing, although available features, generation limits and prices can change.
Midjourney omni reference how to use, step by step
The easiest method is through Midjourney’s website, where you can place an uploaded image into the Omni Reference slot. The exact icon placement can move as the interface changes, but the process remains similar.
- Choose the source. Pick a sharp image with one main person, character or object. Use even lighting and make sure defining details are visible.
- Upload the image. Open the Create page, click the image control beside the prompt box and add your source file.
- Assign the reference. Select the uploaded image and set it as the Omni Reference, not only as a regular image prompt or style reference.
- Write the new scene. State who or what appears, the action, location, framing, lighting and visual style. Mention two to four defining subject details in words.
- Generate a first batch. Keep the normal Omni Strength for the first test. Compare all results for face shape, hair, clothing, colors, proportions and distinctive marks.
- Change one setting. If identity is weak, raise Omni Strength a little. If the result copies the source too closely or loses the requested style, lower it.
- Save a working recipe. Record the source image, core subject wording, aspect ratio, style choice and strength that produced the best result.
A useful prompt structure is: subject description, action, setting, shot type, lighting, style, then exclusions. For example, write: recurring red robot with a square head, chipped white chest panel and round blue eyes, carrying groceries through a rainy city market, full body street photo, soft shop lighting, no extra robots.
How should you choose a strong reference image?
Your source image sets the identity information Midjourney has available. A clean, informative image usually matters more than pushing the strength setting to an extreme.
For a person
Use a sharp image where the face is large enough to inspect, with eyes, hairline and jaw visible. A three quarter view often gives Midjourney more facial shape information than a strict front view, while still keeping the person recognizable.
If clothing must stay consistent, include the upper body or full body. If facial identity matters most, use a closer portrait and describe the required outfit in the prompt.
For a character
Choose the most finished design, not an early sketch with uncertain details. Show signature colors, silhouette, face, accessories and any unusual body proportions.
A character sheet can work, but multiple views inside one image may cause mixed limbs, duplicated accessories or conflicting proportions. Start with one strong view, then test a sheet only if the single view omits essential details.
For an object or product
Use an angle that reveals the object’s main shape, material and identifying marks. Avoid hands covering the product, harsh reflections and a background with similar colors.
- Good source: One subject, sharp focus, even light, visible defining details and little obstruction.
- Weak source: Several people, tiny face, motion blur, hidden features, heavy filters or dramatic shadows.
- Better preparation: Crop unused background, keep enough space around the subject and export a normal quality JPG or PNG file.
Do not use a real person’s likeness deceptively or without permission. For paid work, also confirm that you have the right to use the source image, logos, character designs and product photography.
Which settings improve consistency without freezing the image?
Omni Strength controls how firmly Midjourney follows the subject reference. The normal setting is a sensible starting point, because very high values can preserve more identity while reducing your control over pose, clothing, scene or art direction.
| Problem | First adjustment | Why it helps |
|---|---|---|
| Face or object changes too much | Raise Omni Strength in small steps | Gives the reference more influence |
| Pose stays too close to the source | Lower Omni Strength slightly | Gives the scene prompt more room |
| Style looks weak | Lower strength or simplify the style request | Reduces competition between identity and treatment |
| Extra accessories appear | Name required items and exclude unwanted ones | Makes the prompt less open to guessing |
| Body proportions vary | Use a full body source and state proportions | Provides information missing from a portrait |
Raw mode can help when Midjourney adds too much decorative interpretation. A lower stylization setting can also make literal prompt details more dependable, although the image may look less polished.
Keep the aspect ratio stable across a series when possible. Moving from a square portrait to a very wide scene changes how much room the subject occupies, which can affect clothing details, face size and pose.
A seed can help you compare related prompt tests, but it is not an identity lock. Treat it as a way to reduce random variation during testing, not as a replacement for Omni Reference.
How do you build a repeatable character or product workflow?
Consistency comes from controlling your inputs, not from finding one magic setting. Create a short identity block that you paste into every prompt, such as “middle aged woman, oval face, short black curls, gold round glasses, navy jacket.”
Keep that block unchanged while you alter the action and background. If you rewrite the identity on every attempt, Midjourney receives different instructions and may produce a different person.
Build a reference pack
Start with the original Omni Reference and generate a neutral portrait, a full body view and one side view. Pick only images that still match the intended subject, because using a weak generated result as your new source can slowly change the design.
Midjourney generally uses one Omni Reference at a time. If you need information from several views, combine approved views into one clean reference board, but keep it simple and avoid conflicting outfits or facial expressions.
Use a scene template
Separate the prompt into fixed and changing parts. The fixed part contains identity, core clothing or product features and visual treatment, while the changing part contains action, location, camera angle and mood.
For a client project, save each prompt and label approved images by scene. Also record the aspect ratio, strength, stylization choice and whether Raw mode was used.
Fix failures in the right order
First check the source image, then the written identity block, then Omni Strength. Only after those are stable should you adjust style, camera wording or exclusions.
Correct small defects with Midjourney’s editing tools rather than regenerating the entire scene. For major identity errors, return to the original prompt and source, since repeated edits can change the face, object shape or signature details.
Expect to make several batches for a usable set. Omni Reference reduces variation, but hands, tiny logos, exact patterns and back views can still require selection, editing or a stronger source image.
FAQ
Can Omni Reference keep the same face in every image?
It can make a face recognizably similar across scenes, but it does not guarantee an exact match. Use a sharp portrait, repeat key facial traits and avoid extreme angles until you have a reliable setup.
What Omni Strength should I use?
Start with Midjourney’s normal setting, then move up or down in small steps. Raise it when identity drifts, and lower it when pose, clothing or style stays too close to the source.
Can I use more than one Omni Reference?
Midjourney generally accepts one Omni Reference at a time. If several views are essential, place a few approved views on one simple reference board and test for mixed or duplicated details.
Why does my character’s outfit keep changing?
Your source may not show the full outfit, or the prompt may leave clothing open to interpretation. Use a full body source and repeat the garment type, main colors and signature accessories in every prompt.
Is Omni Reference the same as Style Reference?
No. Omni Reference guides the identity of a person, character or object, while Style Reference guides the image’s visual treatment, such as color, texture and lighting.