# Midjourney Midjourney is a generative AI text-to-image service created by Midjourney, Inc., an independent research lab based in San Francisco. Founded by [[David Holz]] (co-founder of Leap Motion), it entered open beta in July 2022. ## Key Features - **Style Reference** (`--sref`): Upload an image to extract and apply its style (color palette, texture, atmosphere) - **Character Reference** (`--cref`): Maintain character consistency across generations - **Vary (Region)**: Inpaint/edit specific areas while keeping the rest unchanged - **Image Weight** (`--iw`): Control influence of uploaded reference images - **Zoom Out**: Generate surroundings to expand an existing image - **Niji Model**: Specialized model for anime-style images (current: Niji 7) - **Raw Mode** (`--style raw`): More literal prompt interpretation, less stylization ## Versions - **V7** (April 2025): Current stable version - **V6** (December 2023): Better text rendering, more literal prompts - **V5.2** (June 2023): Aesthetics system, zoom out feature - **V4** (November 2022): First version trained on Google TPUs ## Usage Access via Discord bot (`/imagine` command) or web interface (launched August 2024). Web editor consolidates editing, panning, zooming, and inpainting in one interface. ## References - https://www.midjourney.com/ - Examples gallery with prompts: https://www.midjourney.com/explore?tab=top - Discord: https://discord.com/invite/midjourney - Wikipedia: https://en.wikipedia.org/wiki/Midjourney - Library of styles: https://midlibrary.io/ - Great resource to explore artistic styles, artists, scenes, characteristics, and more