| name | midjourney-reference |
| description | Use when the user asks about Midjourney parameters, settings, prompt syntax, aspect ratios, style references, model versions, or needs help crafting a Midjourney prompt. Also use when user asks "what does --sref do", "how do I use character reference", "midjourney tips", or similar questions about MJ features and capabilities. |
Midjourney Parameter & Feature Reference
Complete reference for all Midjourney parameters, prompt techniques, and features. Use this to craft effective prompts and advise on parameter choices.
Quick Parameter Reference
| Parameter | Syntax | Range | Default | Purpose |
|---|
| --ar | --ar W:H | Any ratio | 1:1 | Aspect ratio |
| --stylize / --s | --s N | 0-1000 | 100 | Artistic interpretation strength |
| --chaos / --c | --c N | 0-100 | 0 | Variation between grid images |
| --weird / --w | --w N | 0-3000 | 0 | Unconventional/surreal qualities |
| --no | --no item1, item2 | Text | None | Exclude elements (negative prompt) |
| --raw | --raw | Flag | Off | Literal prompt, less MJ aesthetic |
| --seed | --seed N | 0-4294967295 | Random | Reproducibility |
| --tile | --tile | Flag | Off | Seamless tileable pattern |
| --stop | --stop N | 10-100 | 100 | Stop generation at percentage |
| --quality / --q | --q N | 1, 2, 4 | 1 | GPU time per image |
| --repeat / --r | --r N | 2-40 | 1 | Run prompt multiple times |
| --iw | --iw N | 0-3 | 1 | Image prompt weight vs text |
| --sref | --sref URL/code | URL or number | None | Style reference |
| --sw | --sw N | 0-1000 | 100 | Style reference weight |
| --cref | --cref URL | URL | None | Character reference (V6) |
| --cw | --cw N | 0-100 | 100 | Character reference weight |
| --oref | --oref URL | URL | None | Omni reference (V7+) |
| --ow | --ow N | 0-1000 | 100 | Omni reference weight |
| --p | --p | Flag | Off | Apply personalization profile |
| --exp | --exp N | 5-100 | Off | Experimental creative boost (V7+) |
| --v | --v N | 1-8 | 7 | Model version |
| --niji | --niji N | 5-7 | None | Anime/Eastern aesthetic model |
Aspect Ratio Guide
Common ratios and their use cases:
| Ratio | Use Case |
|---|
| 1:1 | Square, social media, icons |
| 4:3 | Classic monitor, presentations |
| 3:2 | Photography, prints |
| 16:9 | Desktop wallpapers, widescreen |
| 21:9 | Ultrawide monitors (MJ normalizes to 7:3) |
| 9:16 | Mobile wallpapers, stories, vertical video |
| 2:3 | Portrait photography, posters |
| 3:4 | Tablets, book covers |
MJ auto-simplifies ratios: 1920:1080 becomes 16:9, 21:9 becomes 7:3. No decimals allowed (use 139:100 not 1.39:1).
Parameter Deep Dives
Stylization (--s)
- 0-50: Very literal, close to prompt. Good for technical/architectural prompts.
- 100 (default): Balanced. Good starting point.
- 250-500: Noticeably artistic. MJ adds its own aesthetic choices.
- 750-1000: Highly stylized. MJ takes creative liberty. May drift from prompt.
Weirdness (--w)
- 0 (default): Normal output.
- 50-250: Subtle quirky touches. Interesting textures or unexpected compositions.
- 500-1000: Distinctly unusual. Surreal elements emerge.
- 1000-3000: Very experimental. Results can be unpredictable.
Chaos/Variety (--c)
- 0 (default): All 4 grid images are similar interpretations.
- 10-30: Moderate diversity. Good for exploring directions.
- 50-100: High diversity. Each image may be radically different.
Quality (--q)
- 1 (default): Standard quality. Best for most uses.
- 2: More detail and texture. 2x GPU time.
- 4: Maximum detail. 4x GPU time. Not compatible with --oref.
Raw Mode (--raw)
Disables MJ's automatic aesthetic polish. Use when:
- You want photorealistic results with simple prompts
- You have a detailed stylistic prompt and want precise control
- You want less "MJ look" in the output
Style References (--sref)
Apply visual styles from reference images or codes:
cyberpunk cityscape --sref https://example.com/neon-art.jpg
- Multiple refs: separate URLs with spaces
- Style codes: numeric codes from Style Creator (e.g.,
--sref 6281858160)
- Random:
--sref random for surprise styles
- Weight:
--sw 0-1000 controls style strength (default 100)
- Combine with
--raw for style + literal prompt control
Character/Omni References
V6: Character Reference (--cref)
portrait of a warrior --cref https://example.com/character.jpg --cw 80
--cw 100: Full character (face, hair, clothing)
--cw 0: Face only
V7+: Omni Reference (--oref)
Replaces --cref with unified reference:
running through a forest --oref https://example.com/character.jpg --ow 200
- Higher
--ow = stricter adherence to reference
- Competes with
--stylize and --exp for influence
Multi-Prompt Syntax (::)
Weight different concepts independently:
hot::2 dog → emphasizes "hot" (temperature), less "dog"
hot dog → the food item
- No space before
::
- Default weight is 1
- Negative weights allowed:
plants::-0.5
- All weights must sum to positive
Negative Prompts (--no)
Exclude unwanted elements:
landscape photo of mountains --no clouds fog haze people
Warning: MJ reads each word independently. --no modern clothing is interpreted as "no modern" AND "no clothing" which can trigger moderation. Prefer describing what you WANT rather than excluding.
Permutations
Generate multiple variations in one command:
a {red, blue, green} car in a {city, desert}
Creates 6 separate prompts (3 colors x 2 locations).
Works with parameters too:
sunset landscape --ar {16:9, 1:1, 9:16}
Model Versions
| Version | Key Features |
|---|
| V7 (default) | Production model. --oref, --exp, --sref, --p support |
| V8 Alpha | 5x faster, --hd (native 2K), --q 4, better text. alpha.midjourney.com only |
| V6.1/V6 | --cref support, beautiful results |
| Niji 7 | Anime/Eastern aesthetics, better coherency |
| Niji 6 | Japanese text rendering, anime eye detail |
| Draft | Fast iteration, lower quality. Toggle via UI or settings |
Speed Modes
| Mode | Speed | Cost | Availability |
|---|
| Relax | Slowest | Free (unlimited) | Standard+ plans |
| Fast | Normal (~60s) | 1x fast hours | All plans |
| Turbo | ~4x faster | 2x fast hours | V5+, NOT V8 Alpha |
Post-Generation Actions
| Action | What it does |
|---|
| Vary > Subtle | Minor changes to selected image |
| Vary > Strong | Significant changes |
| Upscale > Subtle | 2x resolution, preserves original look |
| Upscale > Creative | 2x resolution, may enhance/alter details |
| Rerun | Generate new set with same prompt |
| Edit | Open in editor for inpainting/outpainting |
| Use > Image | Use as image prompt for new generation |
| Use > Style | Extract style for new generation |
| Use > Prompt | Copy prompt for reuse |
Prompt Crafting Tips
- Front-load important concepts -- MJ weighs earlier words more heavily
- Be specific about style -- "oil painting", "watercolor", "photograph", "3D render"
- Include lighting -- "golden hour", "neon lighting", "dramatic shadows", "soft diffused light"
- Specify camera/lens -- "wide angle", "macro", "telephoto", "35mm film"
- Use artist/style references -- describe aesthetic qualities rather than naming artists
- Keep it concise -- long prompts can dilute focus. 40-60 words is a sweet spot.
- Use --raw for control -- if MJ is adding too much of its own aesthetic
- Iterate with Draft -- use draft mode to quickly test prompt ideas before full generation
- Use --chaos for exploration -- higher chaos shows more diverse interpretations
- Combine --sref with simple prompts -- style refs work best when text describes content, not style
Wallpaper-Specific Tips
- Use
--ar 16:9 for standard monitors, --ar 21:9 for ultrawide
- Use
--stylize 500-800 for aesthetic wallpapers
- Always Upscale > Subtle for clean high-resolution output
- Add
--no text words logos watermark to keep it clean
- Use
--raw for photorealistic wallpapers
- Dark themes work well: "dark background", "moody atmosphere", "deep shadows"