Midjourney Character Consistency with --cref
Keep the same character across many images using Midjourney's character reference parameter.
Prompt template
/imagine prompt: [scene description] --ar 3:4 --cref [URL of reference image] --cw 100 --v 6
How to use
- 1Generate a base character image you love
- 2Copy the image URL or use the image in --cref
- 3Add --cref <image URL> and --cw (0-100) for consistency strength
- 4Keep --ar and style parameters consistent across all shots
Pro tips
- --cw 100 keeps look + style; lower values keep only the face
- Use --sref for consistent art style across a series
Works best with
Ratings & Reviews
Loading reviews…
Related Skills
Generative Media Skills — 41+ agent recipes for images, video, and audio
A high-performance, schema-driven collection of 41 AI agent recipes for generating, editing, and displaying professional-grade images, videos, and audio. Powered by muapi-cli with 100+ models including Midjourney, Flux, Seedance, Kling, and Veo.
Punk Cover — generate article cover images from 32 curated visual styles
Turn articles, notes, tweets, and topic drafts into polished cover images for Xiaohongshu, WeChat public accounts, X/Twitter, and more. Selects exactly one reusable style from a shared library of 32 curated visual styles — including minimalist black-and-white, retro torn collage, pixel art metaphor, industrial OSB blueprint, business magazine front page, avant-garde geometry, layered paper-cut concept, and more — then fuses the style atom into a complete cover-specific prompt. Supports custom aspect ratios, multi-size suites, and prompt-only output mode.
Punk Avatar — generate profile pictures, pet portraits, and surreal paper-art from 8 styles
Generate avatar images from photos or text descriptions for people, pets, and objects. Choose from 8 curated styles including pixel avatar, grotesque soul sketch, messy crayon pet portrait, fashion sketch observation, polaroid keepsake card, minimal paper acrylic block illustration, and surreal pop-up paper landscape. Preserves recognizable subject traits while translating them into the selected style. Default 1:1 ratio with custom ratio support.