Midjourney Development History: From V1 to V7 and the Rise of AI Art
Midjourney has become one of the most influential AI image generators since its open beta in July 2022. Founded by David Holz and the Midjourney team, it evolved from a Discord-based experiment into a staple for artists, designers, and brands. This article traces its development from V1 through V7 and the parallel growth of the Niji anime model line.
Founding and Open Beta
Midjourney launched as an invite-only then open beta on July 12, 2022, operating primarily through Discord. Users type text prompts and receive four image options, then refine with variations and upscaling. The model improved rapidly: V1 (February 2022) was groundbreaking for its time; V2 (April 2022) and V3 (July 2022) added better detail, lighting, and composition. By V3, the service had already established a distinct aesthetic—dreamy, coherent, and often painterly.
| Version | Release (approx.) | Notable change |
|---|---|---|
| V1 | February 2022 | Initial release, foundational style |
| V2 | April 12, 2022 | Improved detail and coherence |
| V3 | July 25, 2022 | Better lighting and composition |
| V4 | November 5, 2022 (alpha) | New codebase, Google TPU training, complex prompts |
| V5 | March 15, 2023 (alpha) | Higher photorealism and prompt adherence |
| V5.1 / V5.2 | May–June 2023 | Refinements and style consistency |
| V6 | December 21, 2023 (alpha) | Retrained from scratch, better text in images |
| V6.1 | July 31, 2024 | Stability and quality polish |
| V7 | April 4, 2025 (alpha) | Latest generation, continued improvements |
The V4 Leap and Beyond
V4 marked a major technical shift: a new codebase and training on Google TPUs. It could handle more complex prompts, logos, and web-style graphics, broadening use cases beyond pure art. V5 and V5.1/V5.2 pushed photorealism and prompt following further, making the model suitable for concept art, marketing visuals, and social content.
V6: Retrained From Scratch
V6 was retrained over roughly nine months. It improved text rendering inside images and more literal interpretation of prompts—addressing long-standing limits of AI image models. V6.1 refined stability and consistency. The Niji line (anime-focused) was updated in step (Niji 5, 6, 7), giving illustrators and animators a dedicated style option.
Product and Workflow Evolution
Beyond model versions, Midjourney expanded its feature set:
- Upscale: 2× and 4× upscaling for higher-resolution outputs
- Vary: Region and full-image variation for iterative refinement
- Niji: Anime and illustration-oriented models (Niji 5, 6, 7)
- Web and API: Web interface and API access for teams and enterprises
- Consistency: Better coherence across multiple generations and characters
Pricing moved to a subscription model (Basic, Standard, Pro, Mega) with monthly limits and optional relax mode, making it predictable for both hobbyists and professionals.
Summary
Midjourney’s development from V1 to V7 reflects the broader rise of text-to-image AI: each version raised the bar for coherence, style control, and usability. Its Discord-first launch, strong community, and steady model updates have made it a default choice for many creatives. As V7 and future versions ship, the focus remains on quality, prompt fidelity, and workflow tools like upscale and vary.
Key Takeaways
- Midjourney open beta launched July 12, 2022; V1 had already shipped earlier that year.
- V4 introduced a new codebase and TPU training; V5/V5.2 improved realism and prompt following.
- V6 was retrained from scratch with better text-in-image and literal prompt handling.
- Niji models (5, 6, 7) cater to anime and illustration workflows.
- Upscale, Vary, web app, and API support professional and team use cases.
Try Midjourney on FuseAITools for text-to-image generation, upscale, and vary in one place.