Midjourney SREF vs CREF: Style vs Character Reference

Midjourney Style Reference: What Visual Style It Preserves

SREF captures the overall aesthetic like colors, textures, and lighting of a reference image without copying specific objects or people. Use Midjourney Style Reference with --sref and --sw to apply that visual vibe to new scenes while your text prompt defines content.

Official Midjourney docs define Style Reference as a way to capture the visual vibe of an existing image and apply it to new creations.

It preserves overall style attributes such as colors, medium, textures, or lighting.

It does not copy objects or people from the reference.

Look:

That is the whole job of Midjourney SREF.

You get a consistent visual theme across new scenes without dragging unwanted subjects into the frame.

SREF is compatible with Midjourney versions 6 and later.

On Discord and the web, apply it with--sref plus one or more image URLs.

You can also pass a style code with--sref <code>, or use--sref random for a random code.

Tune influence with--sw on a 0 to 1000 scale.

Watch this official Midjourney beginner guide to Style References to see the control work in real prompts.

The clip shows how sref captures overall vibe such as colors, medium, textures, and lighting without copying objects or people.

Pay attention to style intensity, image-based references versus style codes, and how blended styles stay cohesive across new scenes.

Default--sw 100 is the normal starting point for most style locks.

A text prompt is required with Style Reference every time.

Official best practices keep that prompt simple and content-first.

Describe what should appear in the scene.

Avoid style words that conflict with the reference aesthetic.

Skip commands like "copy this style."

Doc-style examples stay plain: "detailed portrait of a dog" and "ballpoint pen sketch of a bunny."

The practical result:

Content lives in the text prompt, while Midjourney Style Reference carries the look.

On Discord, you can weight individual SREF URLs with::notation when several style sources compete.

Style Reference can also be combined with Character References, Omni References, and Image Prompts.

Even then, SREF only locks visual style.

It does not become a character-identity control.

Midjourney Character Reference: Preserving Character Identity

CREF was designed to recreate specific character features like hair, face, and clothes across scenes, not overall style. Midjourney Character Reference works best with Midjourney-made characters using --cref and --cw, while V7 routes newer identity work to Omni Reference.

Official Character Reference docs describe Midjourney CREF as a way to recreate a specific character across multiple images.

It tries to match features such as hair color, clothes, and facial traits.

The official announcement frames it as similar to Style Reference, except it matches a character image instead of a style.

That distinction matters in production work.

CREF is identity control, not aesthetic control.

Common mistake:

Creators expect Midjourney Character Reference to lock the full visual look of a scene.

It will not.

On supported versions, apply it with--cref plus a character image URL.

Optional--cw sets strength from 100 down to 0.

Default--cw 100 uses face, hair, and clothes.

--cw 0 focuses mainly on the face, which helps when outfits or hair should change.

This Woollyfern CREF deep dive is worth watching here if you want Character Reference beyond a quick parameter checklist.

The walkthrough covers basics, character weight, multi-scene poses, multi-character methods, and pairing Character Reference with Style Reference.

Watch for the --cw examples that show when face-focused lock is better than full face, hair, and clothes lock, plus what makes a strong character reference image.

Multiple URLs can blend characters with--cref URL1 URL2.

A text prompt is still required with the reference.

CREF works best with Midjourney-created single characters.

It is not designed for real people or photos and will likely distort them.

Precision is limited on fine details.

Exact dimples, freckles, or t-shirt logos often will not match.

For V7, official docs replace Character Reference with Omni Reference.

CREF remains a Midjourney and Niji version 6 workflow.

Treat Midjourney CREF as legacy for newer character-consistency jobs.

For deeper character-consistency guidance, see the Midjourney CREF Tutorial: Stop Character Face Drift.

Midjourney SREF vs CREF: Style vs Character Reference

Use SREF for visual style lock and CREF for character identity lock. In Midjourney SREF vs CREF workflows, Style Reference steers colors, medium, texture, and lighting, while Character Reference targets face, hair, and clothing. For V7 identity work, route newer character jobs to Omni Reference.

They solve different Midjourney problems by design.

SREF preserves aesthetic attributes from a reference image without copying people or objects.

CREF was built to recreate character identity traits across new scenes, not the overall look.

Source-reported V6 teaching often frames the same split: SREF as style control and CREF as character-feature control.

That framing matches official docs and keeps the decision simple.

Here’s the decision rule:

Need style consistency only? Use Midjourney SREF alone.

Need the same character only? Use CREF on V6, or Omni Reference on V7.

Need both? Combine them, but assign each control one job.

The catch:

Creators often feed a face into SREF and expect identity lock.

SREF will not copy that person. It only pulls the visual vibe.

The reverse mistake is expecting CREF to set the whole aesthetic.

CREF matches character traits, not brand look or lighting language.

Omni Reference migration matters for modern character work.

On V7, treat CREF as a legacy mental model and move identity consistency to Omni Reference while keeping SREF for aesthetic lock.

Combining SREF and CREF for Cohesive Image Generation

Combine Midjourney SREF and Midjourney CREF when both style and character consistency matter in the same scene. Official Style Reference docs allow pairing with Character Reference so you can lock identity and aesthetic together without forcing one control to do both jobs.

Use each reference for one role only.

SREF owns colors, medium, texture, and lighting. CREF owns face, hair, clothes, and other identity traits.

You can also attach multiple Style Reference URLs when several style cues matter.

Three practical scenarios cover most production work.

Style lock alone: run SREF when the brief needs a stable series look and the subject can change freely.

Character lock alone: run CREF on V6, or Omni Reference on V7, when identity must hold across new scenes.

Both together: pair SREF with CREF (or Omni) when the same character must stay on-brand inside one visual system.

A clean combined workflow looks like this:

1. Write a content-first prompt that describes the scene, pose, and setting.

2. Add--cref [character-image-url] --cw 100 for full identity, or--cw 0 when you want face focus only.

3. Add--sref [style-image-url] --sw 100 to pull the aesthetic without restating style words in the text.

4. Raise or lower--sw and--cw only after you see which control is dominating.

Example shape:

red-haired explorer standing in a rainy alley at night --cref [character-image-url] --cw 100 --sref [style-image-url] --sw 100

Keep the text on content. Avoid “copy this style” instructions that fight the Style Reference.

The better move:

If style starts washing out identity, lower--sw before you rewrite the whole prompt.

If the face drifts while the look stays perfect, raise character weight or switch V7 identity work to Omni Reference with SREF still handling style.

Common mistakes still break combined runs.

Using SREF for character needs will not lock a face. Expecting CREF to set the full aesthetic will not lock brand lighting or medium.

Trade-off to plan for: combined weights can compete.

Strong style influence can push against character detail, so tune one parameter at a time and keep the text prompt simple.

Midjourney Omni Reference: Replacing Legacy CREF for V7

For V7, use Midjourney Omni Reference with --oref instead of legacy Midjourney CREF. Omni Reference is the newer direction for character consistency and can lock broader subjects like objects and logos. Pair it with --ow and a clear text prompt for stronger identity control.

Treat Midjourney CREF as a legacy path on V7.

Current character-consistency guidance is Omni-first, so older --cref tutorials sit in non-V7 context.

You keep the same identity goal.

You just swap the control to Midjourney Omni Reference for new V7 jobs.

Official Omni docs frame it as a broader “put this into my image” tool.

It can carry characters, objects, logos, animals, and similar subjects, not only face and clothing cues.

Practical use stays tight.

Attach one reference image with --oref, then set --ow from 1 to 1000.

Default weight is 100.

Stay below 400 unless high stylize is intentional and you want the reference to dominate.

Public demos suggest Omni is more accurate than V6 CREF on subjects beyond faces, including smarter camera-angle handling.

The catch: Omni is expensive and constrained.

It costs 2x more GPU time than regular V7 images.

It is not compatible with Fast Mode, Draft and Conversational Mode, or --q 4.

It also conflicts with V6.1 inpainting and outpainting, plus Vary Region, Pan, and Zoom Out unless you strip oref and ow in the Editor.

Fine details can still drift.

Intricate freckles, logos, and accessories often mismatch unless the text prompt restates them.

Best practice is simple.

Combine Omni with a clear text prompt, and add SREF when style lock is also required.

If the desired style differs from the reference, name style at the start and end, drop Omni weight, and reinforce body traits in words.

One Omni image only.

For multi-character work, put both people in that single image and describe each person in the prompt.

Frequently Asked Questions: Midjourney SREF vs CREF