Independent research lab Midjourney entered open beta on Discord during its V3 era, establishing collaborative prompt iteration and opinionated painterly aesthetics weeks before open-weight Stable Diffusion weights launched.
Midjourney
Also known as: Midjourney V6 · Midjourney V7 · Midjourney V1 Video
Plain English
An independent research lab and proprietary family of image and video generation models renowned for strong aesthetic defaults, artistic coherence, and creative controls for style and character references.
Technical Overview
Midjourney operates proprietary hosted diffusion and flow-based visual generators exposed through Discord and web interfaces. Across versions V1 through V7 and the V1 Video model, the platform added natural-language prompt rewriting, --sref style and --cref/--oref omni-reference conditioning, user-specific aesthetic personalization profiles, and image-to-video motion synthesis.
Timeline (5)
2022
Release2023
ReleaseMidjourney launched V5, delivering a major leap in unprompted photorealism, skin and lighting textures, wider dynamic range, and anatomically coherent hands.
- Release
Midjourney released V6 in alpha with a rebuilt prompt-understanding stack capable of following detailed multi-object instructions and rendering legible quoted text inside generated images.
2025
ReleaseMidjourney rolled out V7 on a redesigned architecture featuring default per-user aesthetic personalization, omni-reference (
--oref) conditioning, and a low-latency Draft Mode that renders concepts at roughly 10x speed.- Release
Midjourney expanded into video generation with its V1 Video Model, an image-to-video generator that animates generated or uploaded still frames into five-second extendable clips with selectable camera and subject motion intensity.
See also (8)
Open full graphMidjourney