How to Get Consistent Characters in Midjourney (2026)

How to Get Consistent Characters in Midjourney (2026 Guide: Edit Model, Omni Reference, and –cref)

Quick answer: The method depends on which Midjourney version you’re using, and that changed in 2026. On the current V8.1 and V8.2 models, upload your character as a reference image to the Edit model (on Discord, add --edit plus the image URL) and describe the new scene in plain instructions. On V7, use Omni Reference (--oref plus --ow for weight). On V6 and Niji 6, use the older Character Reference (--cref plus --cw). Whichever you use, the real secret is the same: start with one clean, well-designed reference image, change only the scene in your prompts, and build a small library of approved references.

This guide walks through each method, the settings that matter, how to design a character that stays consistent, and how to fix the most common problems. Midjourney’s tools have been changing quickly, so we point to the official documentation wherever we can and flag which instructions apply to which version.

Key Takeaways

• Consistency in Midjourney comes from reference images, not from prompts alone.
• On V8.1 and V8.2, the Edit model replaces Omni Reference and Character Reference and accepts up to four reference images.
• Omni Reference (--oref) works only with V7, accepts one image, and costs about twice the GPU time of a regular V7 image.
• --cref is a V6 and Niji 6 feature and isn’t supported on V7 or V8.
• Keep a fixed description of your character and change only the setting, pose, or action.
• No method guarantees a perfect match every time. Plan to generate several options and pick the best.

Step 0: Check Which Midjourney Version You’re Using

This is the number one reason people follow a tutorial and get nothing. Many guides online were written for older versions.

Version	Consistent character method	Key parameters
V8.1 and V8.2 (current models)	Edit model with up to 4 reference images	--edit on Discord; image attach on the website
V7	Omni Reference	--oref [image URL] and --ow (weight)
V6 / Niji 6	Character Reference	--cref [image URL] and --cw

At the time of our November 2026 review, V8.2 was the default model. Defaults can change without warning, so confirm your version: on the website, open the settings in the Imagine bar and check the Version field. On Discord, type /settings. To lock a version for a project, add it to the end of your prompt, for example --v 8.2 or --v 7.

One trap to know about: Omni Reference is a V7 feature. If you use --oref while working in V8, the job may be handled by V7, so the result can look different from your other V8 images. If you want V8 and consistency, use the Edit model.

Step 1: Design a Character That’s Easy to Keep Consistent

A good reference image does most of the work. Before you try to reuse a character, create the best possible “source” version of them.

What makes a strong reference image:

• One character only, clearly visible, with the face and key features unobstructed.
• Simple, plain background so the model focuses on the character instead of the scene.
• Even lighting without heavy shadows that hide the face or outfit.
• A clear, neutral pose, facing the camera or at a three-quarter angle.
• A defined outfit and a few memorable details, like a distinct hair color, a scar, glasses, or a signature jacket.
• High enough quality that fine details are visible.

Keep the design simple. Characters with three or four distinctive, easy-to-describe features stay consistent far more reliably than ones with intricate patterns, many accessories, and complicated outfits.

Example starter prompt:

full-body portrait of a young woman with short silver hair, round gold glasses, and a green canvas jacket, neutral expression, plain light gray background, soft even lighting, front view

Generate several versions, pick the best, and treat that image as your “master reference.” Many creators also generate a character sheet (front, side, and back views) as a second reference.

Step 2: Use the Edit Model (V8.1 and V8.2)

According to Midjourney’s documentation, the Edit model can generate new images from up to four reference images, and it replaces both Omni Reference and Character Reference on V8.1 and V8.2. It works with Style References, Moodboards, Personalization, and --hd.

On the website

1. Click the image icon in the Imagine bar to open the images panel.
2. Upload your master reference or choose it from your uploads, then drag it into the Attach to prompt section. You can attach up to four images.
3. Optionally click the lock icon to pin your images to the Imagine bar so they stay attached across prompts.
4. Write your prompt and generate.

You can also start from an image you’ve already made: open it from your Create or Organize page and choose Quick Edit (adds it to the Imagine bar) or Open Editor.

On Discord

Add --edit to the end of your prompt, then paste your image URL. For multiple images, separate each URL with a space. The image must already be online with a valid URL, and if it’s on your device, you can upload it to Discord first to get one.

How to prompt the Edit model

The Edit model accepts instructions, similar to Conversational mode, so you can write things like “let’s see this image from behind” or “make this character sit at a cafe table” instead of only describing a finished scene. Midjourney encourages experimenting with both instruction-style and traditional descriptive prompts.

Example prompts to adapt:

• this character walking through a rainy street at night, cinematic lighting
• let's see this character from the side, holding an umbrella
• this character laughing at a kitchen table, warm morning light --ar 3:2

Tips from the documentation

• Aspect ratio: the model tries to match the first image you upload. To change it, add --ar to your prompt, since your default aspect ratio settings won’t apply automatically.
• Use --raw for more precise control. It reduces Midjourney’s added creative flair so your prompt has more influence.
• Multiple characters: you can attach up to four reference images. If features start getting mixed between characters, Midjourney suggests combining your characters into a single reference image.
• Change the style: pair your reference with a Style Reference or Moodboard, and describe the style in the prompt to help it take hold.
• Change perspective: ask to see the character from the front or behind.

Limitations to know

Per the documentation, the Edit model isn’t currently compatible with --tile, and Edit results aren’t currently compatible with Remix. Features and compatibility keep changing, so check Midjourney’s Edit Model article for the latest.

Step 3: Use Omni Reference on V7

If you’re on V7, Omni Reference is the official tool. It lets you place a character, object, vehicle, or creature from a reference image into new scenes.

On Discord: add --oref to the end of your prompt, then paste your image URL. You can only use one image with Omni Reference.

On the website: attach your image in the Omni Reference slot and use the Omni Strength slider to adjust weight.

Example:

the same character reading in a sunlit library, soft film photography --oref [image URL] --ow 100

Controlling strength with –ow

The omni weight (--ow) controls how strongly the reference affects the result. The documented range is 1 to 1,000 with a default of 100. Midjourney advises keeping the weight below 400 unless you’re using a very high stylize value, because higher values can produce unpredictable results. A practical approach is to start at the default, then adjust in steps, raising it if the character drifts and lowering it if the result looks stiff or copies the reference’s pose too closely.

Style and other settings

If you want the new image to look different from the reference, say so in your text prompt, and consider a Style Reference. Omni Reference can be combined with Style References, Personalization, Moodboards, and stylize.

Limitations

Based on the official documentation:

• Omni Reference costs about twice the GPU time of a regular V7 image.
• It isn’t compatible with inpainting or outpainting that still use V6.1.
• Omni Reference results aren’t currently compatible with Vary Region, Pan, or Zoom Out. To edit them, load them into the Editor and remove the reference parameters.
• It isn’t currently compatible with certain modes, such as Draft Mode and Conversational Mode, or --q 4. Check the current article for any changes.

Step 4: Character Reference (–cref) on V6 and Niji 6

If you work in V6 or Niji 6, Character Reference is the older method. Add --cref [image URL] to your prompt, and control strength with --cw. The character reference was designed mainly around characters, and V6 workflows can accept more than one image URL. It isn’t supported on V7 or V8, so use it only when you deliberately choose those older versions.

Step 5: Keep Your Prompts Consistent

The reference does the heavy lifting, but your prompt can still break consistency. A few habits help:

1. Write a fixed character description and reuse it word for word, such as “short silver hair, round gold glasses, green canvas jacket.”
2. Change only the scene, action, or emotion. If you rewrite the whole prompt every time, you add variables.
3. Keep the style language steady. Switching between “anime,” “oil painting,” and “photo” will change how the character looks.
4. Match aspect ratio and framing when comparing results, since close-ups and wide shots reveal different details.
5. Don’t rely on seeds. Seeds help reproduce results but don’t reliably keep a character consistent when the scene changes. Midjourney notes that in V8, seeds are about 99% identical, not exact.
6. Generate batches. Even good workflows produce some misses, so make several versions and choose the best.

Step 6: Build a Character Bible and Reference Library

Treat your character like a real project asset. Keep a simple document or folder that holds:

Item	What to store
Master reference	The single best image of the character
Character sheet	Front, side, and back views
Fixed description	The exact text block you reuse
Style anchor	The style reference code, moodboard, or style words
Settings	Version, aspect ratio, stylize, and raw settings
Approved scenes	Final images that match your target look
Notes	Fixes that worked, and things to avoid

This also helps when you return weeks later after a default version change, because you’ll know exactly which setup produced the look you like.

Troubleshooting Common Problems

Problem	Likely cause	What to try
Face keeps changing	Weak reference or too much scene change	Use a cleaner, closer reference; in V7 raise --ow in steps; simplify the scene
Outfit changes	Outfit isn’t clear in the reference or prompt	Show the full outfit in the reference and repeat the clothing description
Character looks stiff or copies the pose	Weight too high	Lower --ow; describe the new pose clearly
Style doesn’t match	Reference style is overpowering the prompt	Describe the style in the prompt and add a Style Reference or Moodboard
Two characters merge	Features mixing between references	Combine them into one reference image or generate them separately and composite
Results look “off” compared to other V8 images	--oref routed through V7	Use the Edit model on V8, or work fully in V7
Reference ignored	Incompatible mode or version	Check version and mode against the current documentation

Practical Uses for Consistent Characters

• Comics and storyboards: keep a hero recognizable across panels.
• Children’s books: maintain one look across a full set of illustrations.
• Brand mascots: build a recognizable character for ads, packaging, and social media.
• Games and animation concepts: explore poses and scenes before commissioning final art.
• Social content: create a recurring character for a series.

If you’re building finished assets, our guides to AI illustration generators, AI poster generators, AI flyer generators, and AI icon generators can help you turn character art into layouts and graphics. For mascots that double as brand marks, see our guides to creating a logo with AI and the best AI tools to create a custom logo, and use background removal tools to isolate your character for different layouts.

Legal and Ethical Points to Keep in Mind

A consistent character is often a business asset, so a few cautions are worth knowing:

• Ownership isn’t the same as copyright. A character made purely by AI generally can’t be copyrighted in the U.S., and your rights may depend on how much human creative work you add. See our guides on whether AI-generated images are copyright free and whether you can copyright AI art if you edit it.
• Check the commercial terms. Midjourney’s plan, the $1 million revenue rule, and public-by-default galleries all matter for brand characters. We explain them in our guide on using Midjourney images commercially.
• Consider trademark for mascots. If your character will represent a business, trademark rights and a clearance search may matter more than copyright.
• Use references you have the right to use. Don’t upload other people’s artwork, copyrighted characters, or photos of real people without permission, especially for commercial work.
• Be transparent. If you present a consistent AI character as a real person, such as an “influencer,” disclose it where platforms or advertising rules require.

Frequently Asked Questions

How do I get consistent characters in Midjourney?

Use a reference image. On V8.1 and V8.2, attach your character to the Edit model (or add --edit and the image URL on Discord). On V7, use Omni Reference with --oref. On V6 and Niji 6, use --cref. Then keep your text description steady and change only the scene.

What replaced –cref in Midjourney?

On V7, Omni Reference (--oref) is the documented replacement. On V8.1 and V8.2, the Edit model replaces both Character Reference and Omni Reference.

Does –oref work on V8?

Omni Reference is a V7 feature. Midjourney’s documentation says adding an Omni Reference automatically runs the prompt in V7. On V8, use the Edit model instead.

How many reference images can I use?

Omni Reference accepts one image. The Edit model accepts up to four.

What is a good –ow value?

The range is 1 to 1,000 and the default is 100. Midjourney recommends keeping it below 400 unless you’re using a very high stylize value. Start at the default and adjust in steps.

Why does my character keep changing?

Common causes are a messy reference image, changing too much in your prompt, an incompatible version or mode, or mixing styles. Use a clean reference, keep a fixed description, and check your version.

Can I make a character sheet in Midjourney?

Yes. Prompt for multiple views, such as front, side, and back, on a plain background, and use the best result as an additional reference.

Can I keep two characters consistent in one image?

It’s harder. The Edit model supports up to four reference images, and Midjourney suggests combining characters into one reference image if their features get mixed.

Can I use my own photo or drawing as a reference?

Yes, if you own it or have permission. Your own drawings are especially useful because they add original human work. Avoid uploading images of other people or protected characters without permission.

Will the same seed give me the same character?

Not reliably. Seeds help reproduce a result but aren’t a character-locking feature, and Midjourney notes that V8 seeds are about 99% identical rather than exact.

Is a consistent Midjourney character safe to use commercially?

Often yes with a paid plan that fits your situation, but you should check the terms, avoid lookalikes of protected characters, and understand that copyright protection for AI-generated characters is limited.

The Bottom Line

Consistency in Midjourney is less about a magic prompt and more about a repeatable process: design one strong reference, use the right tool for your version, keep your descriptions fixed, and save everything that works. Because Midjourney keeps changing its tools, confirm your version and read the current documentation before you commit to a long project.

Sources and Further Reading

• Midjourney documentation: Omni Reference and the Edit Model article
• Midjourney documentation: Version article (V7 and V8.x compatibility)
• Midjourney documentation: Character Reference (V6 and Niji 6)

Disclaimer: Midjourney features, parameters, and compatibility change frequently. Verify current instructions in Midjourney’s official documentation. This article is general information and is not legal advice.