midjourney-consistency

name

midjourney-consistency

description

Use when generating multiple images that need to share a consistent visual style, theme, or design language -- such as a set of icons, a series of illustrations, branded assets, UI elements, or any batch of images for the same project. Also use when the user has existing project assets, screenshots, or design files and wants new images that match the established look and feel. Triggers on "match the style", "keep it consistent", "same look as", "make more like this", "design system", "themed set", "series of images", "brand-consistent", "match our existing".

Maintaining Design Consistency Across Multiple Midjourney Generations

When generating multiple images for the same project, theme, or brand, use these techniques to keep everything visually cohesive.

Strategy Overview

There are four main tools for consistency, in order of strength:

Tool	What it controls	Strength
Style Reference (--sref)	Color palette, texture, mood, rendering style	Strongest visual consistency
Omni Reference (--oref)	Character/subject appearance across scenes	Strongest subject consistency
Seed (--seed)	Reproducible starting point for generation	Moderate -- same seed + similar prompt = similar feel
Prompt structure	Shared descriptive language	Foundation -- always use alongside other tools

Method 1: Style Reference Anchoring (Recommended)

The most reliable way to keep a consistent look across a set.

Step 1: Generate your anchor image

Create the first image in your set. Get it exactly right -- this becomes the style source for everything else.

Step 2: Extract the style

From the anchor image's detail view, click Use > Style. This copies the image as a --sref parameter you can paste into future prompts.

Alternatively, use the anchor image URL directly:

new subject or scene --sref https://cdn.midjourney.com/your-anchor-image-url.png --sw 200

Step 3: Generate the rest of the set

Use the same --sref on every subsequent prompt. Only change the subject/content description:

a login page hero illustration --sref URL --sw 200 --ar 16:9 --s 500
a pricing page hero illustration --sref URL --sw 200 --ar 16:9 --s 500
an about page hero illustration --sref URL --sw 200 --ar 16:9 --s 500

Tuning style weight (--sw)

• --sw 100 (default): Moderate influence -- picks up color palette and general mood
• --sw 200-400: Strong influence -- closely matches rendering style, textures, lighting
• --sw 600-1000: Very strong -- may override prompt details in favor of style

For a themed set, --sw 200-400 is usually the sweet spot.

Style Codes

If you find a style you love, save its --sref code (numeric codes from the Style Creator at /style-creator). These are persistent and reusable across sessions:

icon design --sref 6281858160 --sw 300

Method 2: Using Existing Project Assets as Inspiration

When a project already has established visual assets (logos, screenshots, illustrations, UI mockups), use them to ground new generations in the existing design language.

Uploading project assets via Playwright

1. Navigate to /imagine
2. Click the Add Images button (left side of prompt bar)
3. Upload the existing asset from the project directory:
   • Use browser_fill_form or interact with the file upload dialog
   • Alternatively, if the image is accessible via URL, paste the URL directly
4. The uploaded image appears as a thumbnail in the prompt area
5. Add your text prompt after it

As image prompt (composition + content influence)

Upload or reference the existing asset, then describe what you want:

[uploaded project screenshot] a new onboarding screen in the same visual style --iw 1.5 --ar 16:9

• --iw 0.5-1: Light influence -- picks up general vibe
• --iw 1.5-2: Strong influence -- closely follows layout, colors, feel
• --iw 2.5-3: Very strong -- almost reproduces the original with modifications

As style reference only

If you want the aesthetic but completely different content:

a mountain landscape --sref [uploaded-asset-url] --sw 300

This extracts only the color palette, texture, and rendering style from your existing asset.

Practical workflow for project assets

1. Read the project to find existing assets:
   - Glob for images: **/*.png, **/*.svg, **/*.jpg
   - Check public/, assets/, static/, images/ directories
   
2. Pick the most representative asset (hero image, key illustration, etc.)

3. Upload it to Midjourney as a style reference

4. Generate new assets using that reference + consistent parameters

Method 3: Seed Pinning

Use --seed to anchor the random generation starting point. Same seed + similar prompt structure = similar aesthetic output.

minimalist icon, cloud upload, flat design, blue tones --seed 42 --ar 1:1 --s 300
minimalist icon, user profile, flat design, blue tones --seed 42 --ar 1:1 --s 300
minimalist icon, settings gear, flat design, blue tones --seed 42 --ar 1:1 --s 300

Seed alone gives moderate consistency. Combine with --sref for best results.

Method 4: Prompt Template Strategy

Create a reusable prompt template with fixed style descriptors, varying only the subject:

Template: "[SUBJECT], [STYLE BLOCK] [PARAMS BLOCK]"

Style block (keep identical):
"clean vector illustration, soft gradient background, rounded shapes, 
warm pastel palette, subtle shadow, modern SaaS aesthetic"

Params block (keep identical):
"--ar 1:1 --s 400 --raw --no text words"

Then for each image, only swap the subject:

cloud storage icon, clean vector illustration, soft gradient background... --ar 1:1 --s 400 --raw
user authentication icon, clean vector illustration, soft gradient background... --ar 1:1 --s 400 --raw
data analytics icon, clean vector illustration, soft gradient background... --ar 1:1 --s 400 --raw

Combining Methods (Maximum Consistency)

For the tightest consistency across a set, combine all approaches:

[subject description], [shared style block] --sref [anchor-image] --sw 300 --seed 42 --ar 1:1 --s 400 --raw

1. --sref locks the visual style
2. --seed anchors the generation
3. Shared prompt structure reinforces consistent language
4. Fixed parameters (--s, --ar, --raw) prevent per-image drift

Workflow for Generating a Themed Asset Set

Planning phase

1. Decide the set: what images do you need? (e.g., 6 feature icons, 4 hero illustrations)
2. Define the style: colors, mood, rendering approach, aspect ratio
3. Pick parameters: --s, --ar, --raw, --no, --chaos

Anchor phase

4. Generate the first image with careful prompting until you're happy
5. Upscale it
6. Extract style reference via Use > Style

Production phase

7. Generate remaining images using the style reference + consistent params
8. Use Draft mode for quick iteration, then re-run at full quality for keepers
9. Upscale all final selections with the same upscaler (Subtle or Creative, pick one)

Quality control

10. Review all images together -- look for outliers in color, tone, or detail level
11. Re-generate any that drift too far, increasing --sw or --iw
12. Download the full set

Moodboards for Ongoing Projects

For projects where you'll generate assets across multiple sessions:

1. Navigate to /moodboards
2. Create a moodboard for the project
3. Add your best/anchor images to it
4. In future sessions, reference the moodboard images as --sref sources
5. This preserves your style anchors across conversations

Common Pitfalls

• Don't change --stylize between images -- different stylization levels create different rendering intensities
• Don't mix --raw and non-raw -- they produce fundamentally different aesthetics
• Don't vary aspect ratio unnecessarily -- different ratios change composition, which affects perceived style
• Watch out for --chaos -- any value above 0 introduces variation between grid images, making it harder to match across sets
• Use Vary > Subtle, not Strong when iterating -- Strong can drift from your established style