Midjourney Consistent Character: The V8.2 and V7 Workflow That Holds a Face

To keep a Midjourney consistent character on V8.2, attach up to four reference images of the character to your prompt through the Edit Model and repeat one fixed identity description in every prompt. On V7, Omni Reference does the same job from a single image with --oref and a strength set by --ow; the older --cref is a legacy V6 parameter, and --sref carries a look, not a face.

Parameters and limits checked October 2026 against Midjourney's docs for the Edit Model, Omni Reference, Style Reference, versions, legacy features, seeds and the community guidelines.

Most Midjourney character tutorials still teach --cref, which is now a legacy parameter. This page is the Midjourney-only route for an AI influencer persona: which reference type to use for which job, the step-by-step build, the prompt pattern, and a sample 30-day shot set. If you are still choosing between Midjourney, a LoRA, Nano Banana or FLUX, start with our consistent character AI guide, which compares every method; come back here once Midjourney is your pick.

What you will have at the end

• One hero image that defines the character.
• A locked set of up to four reference views (front, three-quarter, profile, full body).
• A two-block prompt: a fixed identity block and a scene block you change per post.
• A planned month of feed shots that all reuse the same references.

Every Midjourney plan is a subscription. The plan comparison lists Basic at $10 a month with 3.3 hours of Fast GPU time, and Standard at $30 a month with 15 Fast hours plus unlimited image generation in Relax Mode (checked October 2026). A persona build burns through generations quickly, so Standard is the practical floor if you plan to post daily.

Midjourney --cref vs --oref vs --sref vs the Edit Model

Midjourney has had four ways to carry something from one image to the next, and they do different jobs. According to the version docs, V8.2 became the default on July 24, 2026 and its Edit Model replaces Omni Reference, Character Reference and the Retexture tool. Omni Reference still works on V7, and the legacy features page lists --cref and --cw as legacy parameters.

The rule that follows from the table: the face always goes in a subject reference (Edit Model on V8.2, Omni Reference on V7), and the feed's look goes in a style reference. The Style Reference docs say it captures colour, medium, texture and lighting and does not copy objects or people, so it cannot hold identity on its own. Used together, they give you the same person in the same visual world.

How to create a consistent character in Midjourney, step by step

These steps use the V8.2 Edit Model. The V7 variant follows in its own section.

1. Generate the hero image

Start without references. Write the persona as a plain description and generate until one image is clearly the character: face fully visible, neutral expression, even light, simple background. This image becomes the anchor for everything else, so pick it slowly. Add --raw if the default styling pushes the face towards a glossy look; the Edit Model docs say Raw reduces Midjourney's added flair so the text has more influence.

portrait photo of a 27-year-old woman, shoulder-length dark brown wavy hair, light freckles across the nose, small silver hoop earrings, plain cream sweater, soft window light, neutral background --ar 4:5 --raw

Expected result: one image you would be happy to see in every post. If two candidates compete, pick the one with the clearest view of the eyes, nose and jaw.

2. Make the other views from the hero

Attach the hero image and prompt for new perspectives. Midjourney's own Edit Model examples use plain instructions such as "let's see this image from the front", and the docs confirm the model accepts instruction-style prompts as well as descriptions. Generate a three-quarter view, a profile and a full-body shot in a plain outfit.

Expected result: three extra images where the character is still clearly the same person. Reject any view where a landmark has moved; a drifted reference teaches every later image the drift.

3. Lock the reference set

Put the hero and the best views in the Attach to prompt section of the Imagine bar. The Edit Model takes up to four images, and the lock icon keeps them pinned across prompts. In Discord, add --edit to the end of the prompt followed by the image URLs, separated by spaces. Note that the Edit Model tries to match the aspect ratio of the first image, and your default aspect ratio setting does not apply, so add --ar to every prompt.

4. Write the two-block prompt

Keep a fixed identity block word for word at the start of every prompt and change only the scene block. This is the same discipline the hub guide recommends for every method, and on Midjourney it matters more because the text is weighted alongside the references.

5. Add the feed look

Pick one style reference or sref code for the whole account and keep it. Midjourney recommends also describing the style in words when you use style references, to activate the look more consistently. Keep style words out of the identity block so they never compete with the face.

6. Review before you post

Compare each image with the hero at full size: eye shape and spacing, nose, lips, jawline, hairline and the signature details. Midjourney warns that intricate details like freckles or logos may not match the reference exactly, so choose signature details that survive that, such as hair colour and cut or a jewellery piece, rather than a precise freckle pattern.

The Midjourney consistent character prompt

Here is the pattern with references attached. The identity block stays the same in every post; the scene block is the only thing you edit:

27-year-old woman, shoulder-length dark brown wavy hair, light freckles, small silver hoop earrings, reading at a bookshop window seat, oversized olive jacket, overcast afternoon light, 50mm photo --ar 4:5 --raw

Our free AI influencer prompt generator turns a persona brief into an identity block you can paste in front of every scene.

Midjourney consistent character on V7: the Omni Reference workflow

If you prefer V7's look, add --v 7 and use Omni Reference. The Omni Reference docs set the rules:

• One image only. Use the hero. If you need two characters, Midjourney suggests one reference image that contains both, described in the prompt.
• Weight with --ow. Any value from 1 to 1,000, default 100. Keep it below 400 unless you use a very high stylize value, or results get unpredictable. Raise it when high --stylize or --exp values compete with the reference.
• Changing style. Lower the weight, name the style at the start and end of the prompt, and restate the physical features you want kept.
• Cost and limits. Omni Reference uses twice the GPU time of a regular V7 image and does not work with Fast Mode, Draft Mode, Conversational Mode or --q 4. Its results cannot go straight into Vary Region, Pan or Zoom Out.

27-year-old woman, shoulder-length dark brown wavy hair, light freckles, walking through a flower market, linen shirt, morning light --oref [hero image URL] --ow 150 --v 7 --ar 4:5

A 30-day posting set from one reference set

Here is how one locked reference set covers a month of feed posts. This is an illustrative plan, not a test log: four weekly themes, five shots each, all generated from the same hero, views and identity block. Changing the theme weekly keeps the feed varied while the face stays fixed.

Generate in batches ahead of the posting date rather than the morning of, so a drifted image is replaced calmly. Our AI face consistency guide covers the seven causes of drift if a week's batch starts to wander, and how to make AI influencer videos picks up when you want the same character in motion. The AI Influencers program covers the full persona system, from brief to posting schedule.

Troubleshooting a drifting Midjourney character

• The face is close but softer or more generic. On V7, raise --ow in steps. On V8.2, check that the hero image is first in the attached set and that the identity block has not changed.
• The frame keeps coming out square or portrait when you wanted another ratio. The Edit Model follows the first reference's aspect ratio; add --ar.
• The look changes between posts. You are probably carrying style in the prompt words. Move it to one --sref and keep it fixed.
• Small details vary. Expected; Midjourney says freckles and logos may not match. Rely on larger signature details.
• Jobs get blocked. Midjourney's moderation can block prompts that look innocent; blocked jobs do not cost GPU time. Rephrase rather than repeat.

Rules for persona work on Midjourney

The community guidelines require safe-for-work content, ban imagery of real people that could harass, defame or otherwise harm, and ask users not to mislead people about the nature or source of generated images. The external image rules on the Omni Reference page add that you must hold the rights to images you upload and must not create sexualised deepfakes. For an AI influencer, that means a fictional character, clear AI disclosure and no uploads of real people's photos.

When Midjourney alone is not enough

Midjourney is strong on aesthetics and weaker on exact facial geometry across hundreds of images. If your persona needs precise likeness in every frame, or explicit control over pose, the open-model route gives you a LoRA trained on your character. Our Stable Diffusion vs Midjourney comparison weighs the two, and the character LoRA dataset guide shows how Midjourney images can become training data. For every parameter beyond character work, see our Midjourney prompts guide.

Midjourney consistent character: FAQ