Midjourney Cheat Sheet 2026

Every parameter, ratio, and setting you need — in one place. Updated for Midjourney v6 and v7.

Models & Versions

Parameter	Description	Example
`--v 7`	Midjourney v7 (latest)	`--v 7`
`--v 6.1`	Midjourney v6.1	`--v 6.1`
`--v 6`	Midjourney v6	`--v 6`
`--niji 6`	Anime/illustration model	`--niji 6`
`--style raw`	Reduces default aesthetic bias	`--style raw`
`--style cute`	Niji-specific cute style	`--style cute`
`--style expressive`	Niji expressive style	`--style expressive`
`--style scenic`	Niji scenic style	`--style scenic`

Aspect Ratios

--ar 1:1    Square
--ar 2:3    Portrait (standard)
--ar 3:2    Landscape (standard)
--ar 4:5    Portrait (Instagram optimal)
--ar 5:4    Landscape
--ar 9:16   Vertical (phone wallpaper, Stories)
--ar 16:9   Widescreen (desktop wallpaper, YouTube thumbnail)
--ar 3:4    Portrait print
--ar 4:3    Landscape print
--ar 21:9   Ultrawide (cinematic)
--ar 2:1    Panoramic

Use whole numbers. --ar 1.85:1 works but --ar 185:100 is more consistent.

Core Parameters

Parameter	Range	Default	What it does
`--stylize` (`--s`)	0-1000	100	Controls Midjourney's artistic styling. 0 = literal. 1000 = max creative interpretation
`--chaos` (`--c`)	0-100	0	Variety across the 4-image grid. 0 = consistent. 100 = wildly different interpretations
`--weird` (`--w`)	0-3000	0	Introduces unconventional elements. Start at 250-500 for noticeable effect
`--seed`	0-4294967295	Random	Locks the randomness. Same seed + same prompt = similar results
`--stop`	10-100	100	Stops generation early. Lower = less detailed, more abstract
`--quality` (`--q`)	0.25, 0.5, 1, 2	1	Render quality. 2 = 2x GPU time (Max plan only). 0.5 = faster, lower detail
`--tile`	—	Off	Creates seamless repeating patterns
`--repeat` (`--r`)	1-40	1	Runs the same prompt multiple times (Pro/Max only)
`--video`	—	Off	Saves a video of the generation process

Image Weight (`--iw`)

Controls how much the image prompt influences the output vs the text prompt.

`--iw` Value	Effect
`--iw 0.5`	Text prompt dominates
`--iw 1`	Balanced (default in v6+)
`--iw 1.5`	Image prompt slightly stronger
`--iw 2`	Image prompt dominates
`--iw 3`	Maximum image influence (v7)

Higher --iw when the image reference IS the look you want. Lower when the image is just inspiration.

Style Reference (`--sref`)

--sref applies a style from a reference image. Use it with a URL or Discord image link:

/imagine prompt a futuristic city skyline --sref https://s.mj.run/ABC123

Multiple references (blend styles):

/imagine prompt a portrait --sref url1 url2

Style weight (--sw) controls reference influence:

--sw 0-1000. Default: 100. Higher = stronger reference influence
Use --sw 500 when you want the reference style to be dominant

Note:

For our full collection of curated SREF codes by style category — from photorealistic portraits to vintage film looks — browse the Midjourney SREF section.

Character Reference (`--cref`)

--cref applies a character's appearance from a reference image:

/imagine prompt a knight standing in a meadow --cref https://s.mj.run/XYZ789

Character weight (--cw):

--cw 0 — only the face is referenced
--cw 100 — face, body, and clothing are referenced (default)

Remix Mode

Toggle with /settings or /prefer remix. When enabled, clicking V (variation) opens the prompt for editing before generating. Useful for iterative refinement.

Prompt Weighting

Control how much attention each part of the prompt gets:

/imagine prompt forest::2 path::1 sunlight::0.5

forest::2 — twice the default weight
sunlight::0.5 — half the default weight
Higher numbers = more emphasis
Can use decimals (0.25, 1.5, 3.0)

Negative weighting (v7 only):

/imagine prompt landscape::-0.5 trees

Reduces the influence of "trees" in the output.

Prompt Structure

Optimal prompt order in Midjourney:

/imagine prompt [main subject], [details], [style], [composition], [lighting], [colors], [mood] --parameters

Example

/imagine prompt a warrior cat with armor, intricate filigree patterns on the armor,
digital painting, dynamic pose, dramatic side lighting from a single torch,
deep blues and warm oranges, tense atmosphere --ar 2:3 --s 400 --v 7

The first words carry the most weight. Put your main subject first.

Common Commands

Command	What it does
`/imagine`	Generate images from a text prompt
`/describe`	Reverse: generate prompts from an image
`/blend`	Blend 2-5 images together
`/shorten`	Analyze a prompt and suggest shorter versions
`/settings`	View and adjust default settings
`/info`	View your account info and usage
`/show`	Regenerate using a job ID
`/prefer option set`	Create a custom option
`/prefer suffix`	Set a suffix auto-appended to all prompts
`/stealth`	Generate images privately (Pro only)

Upscaling & Variation

After generation, buttons appear below the 4-image grid:

Button	Label	What it does
U1-U4	Upscale	Generates a high-res version of that image
V1-V4	Variation	Creates 4 new images similar to the chosen one
🔄	Reroll	Regenerates the entire 4-image grid from scratch

After upscaling:

Vary (Subtle) — small changes, preserves composition
Vary (Strong) — significant reinterpretation
Zoom Out — expands the canvas around the image (1.5x, 2x)
Pan — extends the image in one direction
Upscale (2x, 4x) — further resolution increase

Quick Reference Card

General quality:  --q 1 --s 100
Photorealistic:   --v 7 --style raw --s 50
Highly stylized:  --v 7 --s 750 --c 20
Anime/Illustration: --niji 6 --style cute
Consistent character: --cref [url] --cw 100
Square post:      --ar 1:1
Instagram post:   --ar 4:5
YouTube thumb:    --ar 16:9
Wallpaper:        --ar 9:16 (phone) or 16:9 (desktop)
Seamless pattern: --tile
Fast iterations:  --q 0.5
Similar but different: Use V (Variation) buttons
Lock style across prompts: Use --seed and --sref together

Note:

Pro tip: Create presets with /prefer option set. Example: /prefer option set photo --v 7 --style raw --s 50 --ar 4:5. Then use --photo in any prompt to auto-apply those settings.

Midjourney Cheat Sheet 2026