midjourney
π―SkillInstallation
npx skills add g1joshi/agent-skills --skill midjourneyRun it in a terminal; it asks which AI coding tools to install it for (Claude Code, Codex, Cursor and more).
Uses skills.sh Β· Node.js 22.20+
After installing
Ask for related work and the agent uses it on its own, or call it yourself.
/midjourneyInstall by hand
Copy the whole skill folder from GitHub to the place below. Copying only SKILL.md can leave out files the skill uses.
Open the skill folder β- This project
.claude/skills/midjourney/ - All projects
~/.claude/skills/midjourney/
These paths are for Claude Code. For other tools, the command above puts it in the right place.
SKILL.md
View on GitHub βExpert Midjourney generative AI assistance covering prompt engineering, parameters (--v, --ar, --stylize, --chaos), and style referencing. Use when designing photorealistic concept art and production visuals.
Midjourney
Midjourney is an advanced generative image platform recognized for artistic composition, fine detail, photorealism, and parameter-driven prompt manipulation.
When to Use
- High-End Concept Art & Creative Exploration: Synthesizing cinematic illustrations, character design, and game environments.
- Photorealistic Architectural & Product Visuals: Generating high-fidelity mockups with precise lighting and texture prompts.
- Style Consistency with Style References: Replicating specific artistic styles and palettes across multiple prompts using
--sref. - Multi-Prompt Weighting & Composition: Blending concepts with explicit semantic weights using
::syntax.
Quick Start
/imagine prompt: cinematic wide shot of a futuristic data center with neon liquid cooling, architectural digest style, volumetric lighting, unreal engine 5 render --ar 16:9 --v 6.1 --style raw
Core Concepts
Core Command Parameters (v6 & v6.1)
Directing aspect ratio, stylization, and rendering engines:
/imagine prompt: cinematic photography of a sleek glass server rack in a minimalist data center, volumetric teal lighting, shallow depth of field, architectural digest style --ar 16:9 --v 6.1 --style raw --stylize 250 --quality 1
Parameter breakdown:
--ar 16:9: Sets wide landscape aspect ratio (9:16 for mobile, 1:1 for square).--v 6.1: Specifies the Midjourney v6.1 rendering algorithm.--style raw: Reduces Midjourney's default aesthetic bias for more photographic realism.--stylize 250(--s): Controls strength of artistic flair (range 0 to 1000).--chaos 15(--c): Adds variation to initial grid generations (range 0 to 100).
Style References (--sref) & Character Consistency (--cref)
Transferring aesthetic signatures across scenes:
# Style Reference: match aesthetic of an existing reference image
/imagine prompt: an autonomous delivery drone flying over a futuristic metropolis at sunset --sref https://example.com/style_sample.jpg --sw 800
# Character Reference: maintain subject identity
/imagine prompt: the detective sitting at a rainy cafe table reviewing case files --cref https://example.com/detective_face.jpg --cw 90
Multi-Prompting with Explicit Weights (::)
Preventing concept bleed and tuning semantic emphasis:
# Concept separation: 'hot' and 'dog' instead of a frankfurter
/imagine prompt: hot::2 dog::1 running on beach --ar 3:2
# Negative weighting: suppress specific elements without prompt negative words
/imagine prompt: vibrant Tokyo street market at night, cinematic neon reflections --no blur, text, watermark, vehicles
Common Patterns
Character Consistency with Style and Character Reference (--sref, --cref)
Problem: Generating the same character or style across multiple scenes produces completely different visuals.
Solution:
Use --cref (character reference) and --sref (style reference) parameters:
/imagine prompt: cybernetic detective walking down a rainy alley, night time, cinematic lighting --cref https://url-to-character-face.png --cw 80 --v 6.1
/imagine prompt: cozy cafe interior with warm morning light --sref https://url-to-style-reference.png --sw 100 --v 6.1
Best Practices
Do:
- Use
--style rawwhen aiming for accurate photorealism and literal prompt adherence. - Specify lighting, camera lenses, and film stock (e.g.
35mm lens, f/1.8, golden hour illumination) for realism. - Leverage
--sref(Style Reference) with--sw(weight) to maintain brand aesthetic across a series of images. - Use
--nofor negative conditions rather than writing phrases like "without people" in the main prompt.
Don't:
- Use buzzwords like "photorealistic", "hyperrealistic", or "4K"; describe physical lighting and textures instead.
- Use long, rambling paragraphs; concise, comma-separated descriptive descriptors yield better results.
- Exceed
--stylize 750unless abstract, highly stylized, or artistic interpretation is desired.
Troubleshooting
| Error | Cause | Solution |
|---|---|---|
Invalid parameter: unrecognized flag | Typo in parameter name or space between double dashes. | Ensure format is exactly --ar 16:9, --v 6.1, or --stylize 250. |
Image reference URL cannot be accessed | Direct link to image URL is private, expired, or not ending in image extension. | Ensure image URL is public and ends in .png, .jpg, or .webp. |
Banned prompt / Content filter trigger | Prompt contains words flagged by Midjourney community guidelines. | Rephrase prompt using descriptive, non-violent, non-explicit terms. |
References
More from this repository10
A template skill providing a starter framework for creating new agent skills with structured sections for triggers, quick start examples, and capability descriptions.
Skill
Skill
Skill
Skill
Skill
Skill
Skill
Skill
Skill