Essential Parameters
| Parameter | Syntax | Range | What It Does |
|---|
| Aspect Ratio | --ar 16:9 | Any ratio | Controls image dimensions |
| Stylize | --s 100 | 0-1000 | Artistic freedom (higher=more artistic, less literal) |
| Chaos | --c 20 | 0-100 | Variation between grid images (higher=more diverse) |
| Weird | --w 500 | 0-3000 | Introduces unconventional/experimental elements |
| Style Reference | --sref URL | Image URL | Transfer style from reference image |
| Character Reference | --cref URL | Image URL | Maintain character consistency across generations |
| Model Version | --v 6 | v1-6 | Which MJ model to use |
Prompt Structure
| Item | Description |
|---|
Subject | Main focus first: 'a cyberpunk samurai warrior...' |
Details | Clothing, expression, pose, setting: '...wearing neon armor, battle stance...' |
Style | Artistic direction: '...photorealistic / oil painting / anime / 3D render' |
Lighting/Color | '...cinematic lighting, blue and orange color palette, rim light' |
Parameters | --ar 16:9 --s 250 --v 6 |
Multi-Prompts | :: operator weights prompt sections: 'dog::2 cat::1' = dog twice as important |
Style Reference (--sref)
| Item | Description |
|---|
--sref URL | Apply style from image: colors, composition, mood — not content |
--sw 100 (0-1000) | Style weight — 0 = barely influenced, 1000 = strongly stylized |
Multiple --sref | Combine styles from multiple images with weighting |
Style Code | --sref random generates a style code you can save and reuse |
Moodboards | Use --sref for consistent visual identity across a project |
Advanced Workflows
| Item | Description |
|---|
/describe | Upload image → MJ describes it in prompts — reverse engineer any image |
/blend | Combine 2-5 images — MJ merges content and style |
Pan/Zoom | Extend image in any direction — outpainting for panoramas |
Vary Region | Select region → regenerate — fix specific areas without changing rest |
Remix Mode | Change prompt while varying — keep composition, alter style or subject |
Permutation Prompts | {red, blue, green} generates one image per option — A/B test prompts |
Pro Tip: Use --style raw for photorealistic results — it bypasses Midjourney's default aesthetic bias. For artistic/stylized images, omit --style raw and use --stylize 250-750.