Midjourney 101 — Visual Continuity: Mastering SREF, CREF, and Omni References

The "Secret Sauce" for branding and storytelling. Master the reference system to maintain character and style consistency across different prompts so your visuals always belong to the same campaign.

The One-Off Trap

The biggest frustration for new AI creators is the lack of "memory." You generate a perfect character in a cafe, but when you ask for the same character in a boardroom, they look like a completely different person. In the professional world, this is a deal-breaker. A brand needs a consistent soul.

In May 2026, Midjourney has solved this with a trio of continuity tools. We are moving beyond the prompt and into Reference Architecture. This module teaches you how to "pin" your visual DNA so that every image you generate feels like it was shot on the same day by the same photographer.

A digital DNA strand glowing in cyan and purple, connecting four different scene images showing a consistent mascot across desk, cafe, park, and boardroom settings

What You Will Learn

Section 1: SREF—The Aesthetic Glue (Style Reference)

Section 2: CREF—Character Integrity (Character Reference)

Section 3: --oref—The 2026 Omni Link (Combined Reference Authority)

Section 4: Strength Controls—The --sw and --cw Dials

Section 5: Building Brand Universes—Real-world campaign workflow

Section 1: SREF – The Aesthetic Glue (Style Reference)

The Concept: Style Reference (--sref) is how you copy-paste a vibe without copying the subject. If you have an image with a beautiful "80s Retro-Vaporwave" aesthetic, you can apply that exact vibe to any new subject—a robot, a landscape, a product—while maintaining the original style language.

How SREF Works

Step 1: Identify Your Aesthetic

Find or create an image with your desired style

Could be: color palette, lighting mood, texture, composition approach

This becomes your aesthetic authority

Step 2: Get the Image ID

In Web Alpha: Right-click image → "Get Image ID"

In Discord: Extract from the image URL after /file/

Copy the full ID (32-character alphanumeric code)

Step 3: Apply to New Prompt
\\\/imagine prompt: A sleek robot in a corporate boardroom, professional lighting --sref [IMAGE_ID] \\\

Expected Result: The robot inherits the vaporwave color palette, lighting treatment, and overall aesthetic from your reference image—but remains a completely different subject.

SREF Use Cases

Brand Color Consistency

Generate all marketing assets with the same color palette

Ensures visual unity across social media, website, print

Lighting Language

Apply the same cinematography style to different subjects

Golden hour + volumetric haze for all product shots

Cinematic drama for all hero images

Texture & Grain

Film photography aesthetic across all outputs

Maintains "grain signature" and color warmth

Professional, cohesive brand voice

Art Direction

Apply a photographer's style to multiple subjects

"Shot by the same person on the same day" feeling

Luxury, editorial, or documentary consistency

A split-screen showing 80s retro vaporwave aesthetic on the left, and the same aesthetic applied to a futuristic robot on the right, maintaining pastel colors and grid patterns

Section 2: CREF – Character Integrity (Character Reference)

The Concept: Character Reference (--cref) is how you lock in a character's appearance across different scenes, outfits, and contexts. Generate a character once, then use them consistently in any environment—cafe, boardroom, spaceship—without them looking like a different person.

How CREF Works

Step 1: Create Your Hero

Generate your character image with perfect features

This is your character authority

Quality matters—get the best possible version

Step 2: Extract Character ID

Same process as SREF—get the image ID from your reference

This ID becomes your character's genetic code

Step 3: Deploy Across Scenes
\\\/imagine prompt: The character sits in a Tokyo cafe, sipping coffee, rain outside, cinematic lighting --cref [CHARACTER_ID] \\\

Step 4: Scene Variations
\\\/imagine prompt: The character in a boardroom, presenting to investors, professional attire, corporate lighting --cref [CHARACTER_ID] \\\

Result: Same character, different environments. Facial features, body proportions, and distinctive elements remain locked.

CREF Parameters

Consistency Levels

By default, CREF maintains high fidelity

Character appears in new prompt while adapting to the new environment

Clothing and context change; core identity remains

Clothing & Context Adaptation

CREF respects your new prompt instructions

If you say "business suit," they'll wear a business suit

If you say "pajamas," they'll wear pajamas

Core character identity never wavers

Advanced CREF Techniques

Brand Mascot Strategy

Generate your mascot once (perfect version)

Use CREF for every marketing asset, advertisement, video frame

Creates instant brand recognition

Character Development Arc

Same character across story frames

Early morning → stressed version → relaxed evening

Visual journey feels cohesive because the character is unchanged

Multi-Character Consistency

Use multiple CREF IDs in one prompt

"Character A and Character B in a meeting"

Both maintain their identities while interacting

Four panels showing the same cyber-panda character in different locations (office, cafe, park, boardroom) with identical facial features and proportions

Section 3: --oref – The Omni Link (Omni Reference Authority)

The Game Changer: In 2026, Midjourney introduced --oref (Omni Reference), which allows you to anchor both subject AND style simultaneously with near-perfect fidelity. This is SREF + CREF merged into one hyper-precise tool.

How --oref Works

The Difference from SREF/CREF

SREF locks aesthetic only (style bleeds in)

CREF locks character only (appearance preserved)

--oref locks BOTH simultaneously with unprecedented precision

Single Image, Dual Authority
\\\/imagine prompt: The hero character in a new scene, desert landscape, sunset, action pose --oref [REFERENCE_ID] \\\

What --oref Preserves

Character's facial features and body structure

Aesthetic and color language

Composition and framing philosophy

Lighting approach

Overall "energy" or mood

What --oref Allows to Change

Location and background

Clothing (if specified in prompt)

Pose and action

Environmental context

Atmospheric conditions

--oref Strength Control

The --oresw parameter (Omni Reference Strength Weighting) controls how much influence the reference has:

\\\--oref [ID] --oresw 100 (Maximum fidelity, strict lockdown) --oref [ID] --oresw 75 (High fidelity, some flexibility) --oref [ID] --oresw 50 (Balanced, reference influence is medium) --oref [ID] --oresw 25 (Subtle influence, more variation allowed) \\\

Real-World --oref Workflow

Scenario: You have a perfect hero image of a tech entrepreneur. You need 20 variations for a marketing campaign.

\\\Base Reference: entrepreneur in a sleek office Scene 1: entrepreneur presenting at a conference --oref [ID] --oresw 90 Scene 2: entrepreneur at home working on laptop --oref [ID] --oresw 90 Scene 3: entrepreneur celebrating success --oref [ID] --oresw 90 Scene 4: entrepreneur mentoring a junior colleague --oref [ID] --oresw 85 \\\

Result: 20 perfectly consistent campaign images with visual cohesion and narrative flow.

A central reference image splitting into two streams labeled "Character DNA" and "Aesthetic DNA", both converging into a new generated image that blends both perfectly

Section 4: Strength Controls – The --sw and --cw Dials

The Precision Layer: Not all references should have equal influence. Sometimes you want the style to dominate; sometimes you want the character to shine. The strength controls give you surgical precision.

--sw (Style Strength Weight)

Controls how much your SREF aesthetic influences the output:

\\\--sref [ID] --sw 100 (Full stylistic authority - aesthetic dominates) --sref [ID] --sw 75 (Strong influence - 75% aesthetic, 25% variation) --sref [ID] --sw 50 (Balanced - equal aesthetic and generation freedom) --sref [ID] --sw 25 (Light touch - reference is subtle influence only) \\\

When to Use High --sw Values (75-100)

Brand campaigns requiring strict aesthetic consistency

Color palette must be exact

Lighting signature is critical to brand identity

When to Use Low --sw Values (25-50)

Experimental variations on a style

Allow creativity while maintaining reference flavor

Testing how far you can push while staying "in brand"

--cw (Character Strength Weight)

Controls how strictly your CREF character is locked:

\\\--cref [ID] --cw 100 (Absolute character lock - maximum fidelity) --cref [ID] --cw 75 (Very high fidelity - minor variation allowed) --cref [ID] --cw 50 (Balanced - character essence preserved, some drift) --cref [ID] --cw 25 (Loose interpretation - character DNA present, style flexible) \\\

When to Use High --cw Values (75-100)

Brand mascots that must be immediately recognizable

Character-driven storytelling (same person across scenes)

Legal/trademark consistency requirements

When to Use Low --cw Values (25-50)

Character evolution (aging, weight loss, transformation)

Subtle personality shifts across scenes

Allowing the new environment to "influence" the character's appearance

Combined Strength Strategy

\\\/imagine prompt: Hero character in boardroom, professional lighting --cref [CHARACTER_ID] --cw 95 --sref [BRAND_AESTHETIC_ID] --sw 80 \\\

This locks the character hard (95%) while maintaining brand aesthetic at high fidelity (80%). Result: Unmistakable character, perfectly on-brand.

A horizontal slider showing three outputs: weak influence (barely styled character), balanced (perfect blend), and full influence (style completely dominates)

Section 5: Building Brand Universes – Real-World Workflow

The Campaign: You're launching a luxury tech brand with:

A distinctive brand mascot (cyber-panda)

A color palette (midnight blue, gold, neon cyan)

A photographic style (cinematic, high-end, luxe aesthetic)

50 marketing assets needed across 10 different scenarios

The Reference Architecture

Step 1: Create Your Authorities

Reference A: Brand Aesthetic

Generate a perfect hero image embodying your brand's visual language

Luxury tech, midnight blue + gold color palette, cinematic lighting

This is your --sref authority

Reference B: Brand Character

Generate your cyber-panda mascot in its "canonical" form

Perfect proportions, distinctive features, on-brand styling

This is your --cref authority

Step 2: Build Your Campaign Grid

For each of 50 assets, use both references:

\\\Scene 1: Mascot in modern office /imagine prompt: Cyber-panda working at a minimalist desk, floor-to-ceiling windows, sunset cityscape, professional attire --cref [PANDA_ID] --cw 95 --sref [AESTHETIC_ID] --sw 85

Scene 2: Mascot at tech conference /imagine prompt: Cyber-panda on stage presenting, dramatic stage lighting, conference room, energetic pose --cref [PANDA_ID] --cw 90 --sref [AESTHETIC_ID] --sw 85

Scene 3: Mascot mentoring /imagine prompt: Cyber-panda in discussion with junior team member, modern office, collaborative vibe, warm lighting --cref [PANDA_ID] --cw 90 --sref [AESTHETIC_ID] --sw 80 \\\

The Result

50 assets that are:

✅ Instantly recognizable (same mascot)

✅ Perfectly on-brand (same aesthetic)

✅ Contextually unique (different scenarios)

✅ Ready for deployment (no retouching needed)

✅ Produced in 4-5 hours (not weeks)

A hierarchical architecture diagram showing Reference Authority at top, branching to SREF, CREF, and --oref, all flowing down to multiple consistent generated images

Time Comparison

Without Reference System:

Generate: 20+ attempts per image (no consistency)

Retry with tweaks to "lock" look

Manual brand consistency enforcement

Total for 50 assets: 20-30 hours

With SREF/CREF/--oref:

Generate: 1-2 attempts per image (locked references)

Strength dials fine-tune instantly

Automatic brand consistency

Total for 50 assets: 4-6 hours