Midjourney Timeline: Version Release Dates and Major Updates

Midjourney is a generative AI service that creates images from text and visual prompts. The San Francisco company Midjourney, Inc., founded by David Holz, opened the product to public beta in July 2022 after testing it through a Discord bot.

Midjourney is both the name of the service and the name of its main image-model family. That family uses version numbers such as V6, V7, and V8.1. Niji covers anime and illustration, and Video V1 turns images into short video clips.

V8.1 is the current default image model. Midjourney now has a full web interface alongside Discord, plus editing tools, personal style profiles, visual references, and video. The timeline below traces how those parts arrived.

Last updated: July 16, 2026.

Latest Midjourney Update

Midjourney released Draft Mode for V8.1 on June 16, 2026. A draft job creates 24 lower-resolution concepts at half the Fast-hour cost of a standard-definition V8.1 job. Members can choose a draft and use Vary to render it at full quality and resolution.

Draft Mode arrived five days after V8.1 became the default model for ordinary image jobs. V8.1 renders faster than V7, follows prompts more closely, handles text better, and can create 2K images in HD mode.

Midjourney Timeline at a Glance

Midjourney’s early development moved quickly. Three model generations appeared within months, and the public beta put the service in front of a much wider audience before the end of its first year.

2022: Early Models and the Public Beta

February to July 2022: Versions 1, 2, and 3 Establish the Model Line

Version 1 served as Midjourney’s default model from February through April 2022. Its images were abstract and painterly, with limited control over coherent subjects. Version 2 followed from April through July. It was colorful and artistic, with a more refined generation system underneath.

Version 3 became the default in July. Its images were more coherent and better composed, although prompts could produce unpredictable results. Versions 1 through 3 also established the numbering system used by every main Midjourney image model since then.

July 2022: Midjourney Opens Its Public Beta

Midjourney opened its beta to the public in July 2022 after earlier limited testing. New users joined the Midjourney Discord server and submitted prompts through the /imagine command. The bot returned image grids inside Discord, which also served as the product interface, community space, and public gallery.

The public beta defined the early Midjourney experience. Image generation happened in shared channels, where members could watch other people write prompts and choose results. Direct web creation reached the full membership in 2024.

November 2022: Version 4 Changes Image Quality and Prompt Handling

Version 4 used a new codebase and model architecture trained on Midjourney’s AI supercluster. It handled objects, places, creatures, and scenes with several subjects more reliably than the first three generations. Fine detail and image-prompt performance also improved.

V4 became the default on December 20, 2022. Midjourney later offered several V4 style variants. Across those variants, prompts produced more recognizable subjects and handled complex arrangements with greater consistency.

December 20, 2022: Niji 4 Starts a Separate Model Line

Niji 4 was the first anime and illustration model developed by Midjourney and Spellbrush. It used the Midjourney creation environment and began its own version sequence, with visual priorities shaped around drawing and anime styles.

Niji runs alongside the main Midjourney model family. Members can choose either line according to the look they want, so Niji 4 and the main V4 model represent separate choices inside the same service.

2023: V5, Web Creation, and Better Prompt Accuracy

March 2023: Version 5 Moves Toward Photographic Output

Version 5 produced more photographic images than its predecessors and followed prompts more closely. The model generated 1024-pixel images and supported a wider range of aspect ratios. Detailed prompts often worked better than short instructions because V5 used less automatic styling than the version that followed.

V5 became the default on March 30. Compared with the early models, it produced more controlled scenes, more realistic materials, and a closer reading of the prompt.

May 4, 2023: Version 5.1 Strengthens Default Aesthetics

V5.1 worked better with natural language and created coherent results from simpler prompts. It reduced unwanted borders and artifacts, sharpened images, and restored a stronger built-in aesthetic. Short prompts became more practical because the model filled in more visual detail on its own.

June 2023: Version 5.2 Refines Detail and Composition

V5.2 sharpened images and improved color, contrast, composition, and prompt understanding. It also responded across a wider range of the Stylize setting. Midjourney made it the default on June 22 and used it as the standard model until February 2024.

The V5.2 period included tools such as Vary Region and the Style Tuner. Vary Region let members regenerate a selected part of an image. The Style Tuner later became a precursor to broader personalization, Moodboards, and style exploration.

December 13, 2023: Image Creation Reaches the Web Alpha

Midjourney began limited testing of direct image creation on its website. Early access focused on experienced members with large generation histories. The Alpha introduced a conventional prompt bar, visual settings, an archive, and web controls for reference images.

The website Alpha gave members visual controls and direct access to their image history. It became a second creation interface, with routine settings available outside the Discord command system.

December 20, 2023: Version 6 Improves Long Prompts and Text

V6 understood longer prompts more accurately and handled complex scenes with greater coherence. Image prompting and Remix became more precise. Words requested inside quotation marks also appeared more clearly in generated images.

Midjourney released V6 on December 20, 2023, then made it the default on February 14, 2024. Later model generations followed a similar pattern, with public testing before the default changed.

2024: References, Personalization, and Broad Web Access

March 12, 2024: Character Reference Enters Testing

Character Reference introduced a way to carry recognizable character traits from a reference image into new V6 or Niji scenes. Members supplied a reference through the web interface or the --cref parameter in Discord and adjusted how strongly the model preserved the face, hair, and clothing.

Character Reference worked best with characters created in Midjourney, though exact identity matching was limited. It gave members a practical way to develop the same visual character across several scenes.

June 7, 2024: Niji 6 Improves Anime Detail

Niji 6 improved the structure of anime eyes, small background details, and other features that could break in earlier images. It also handled short Japanese kana and some simple Chinese characters better than Niji 5. Its illustration-focused style followed the separate Niji model line.

June 12, 2024: Personalization Learns Individual Preferences

Midjourney opened an early test of model personalization. The system learned aesthetic preferences from pair rankings and images liked on the Explore page. Members with enough preference data could activate the result with --p and adjust its strength through Stylize values.

Personalization used an individual profile to handle visual choices left open by the prompt. A member’s own preferences now worked alongside Midjourney’s default aesthetic.

July 30, 2024: Version 6.1 Becomes the Default

V6.1 targeted common flaws in bodies, hands, small faces, plants, animals, textures, and distant details. Standard image jobs ran about 25 percent faster than V6. Text became more accurate, new 2x upscalers handled enlarged images, and Personalization received a revised model.

Midjourney made V6.1 the default on release day. Editing actions such as Reframe and Repaint fell back to V6 at first because V6.1 lacked its own inpainting and outpainting model.

August 23, 2024: The Web Platform Opens to All Members

Midjourney opened web creation to its full membership. New users could sign in with Google or Discord. Returning Discord members could connect their earlier image history. Creation, browsing, organization, settings, and account management now shared one interface.

Members could now choose between typed Discord commands and visual web controls. Subscription settings synchronized across both interfaces, so moving between them required little setup.

October 23, 2024: The External Image Editor Begins Testing

The External Image Editor accepted images uploaded from a computer. Members could crop or expand the canvas, repaint selected regions, add new content, and modify a scene through text prompts. A Retexture mode estimated the underlying structure and changed materials, surfaces, and lighting.

The Editor gave members a second workflow: upload an existing image, select the area that needed work, and revise it inside the browser.

2025: V7 and the First Video Model

April 3, 2025: Version 7 Is Released

Version 7 handled text prompts and visual references more accurately. Images showed richer textures and better-formed bodies, hands, objects, and fine details. Draft Mode sped up concept work, and Midjourney began preparing a new system for consistent subjects.

During testing, Midjourney updated V7’s image quality, speed, editor interface, and Style Reference behavior. The model became the default after those pieces were in place.

May 1, 2025: Omni Reference Replaces the Narrower Character System

Omni Reference entered testing as V7’s system for placing a specific subject into a new image. Its scope covered people, objects, vehicles, nonhuman creatures, and compositions with more than one subject.

Members supplied the reference through the web image tray or the --oref parameter. Omni Weight controlled how closely the result followed it. Higher Stylize and aesthetics values could weaken that connection.

June 17, 2025: V7 Becomes the Default Model

Midjourney changed the default from V6.1 to V7 after several months of testing. By this point, Omni Reference and the V7 Style Reference system were available, Draft Mode offered faster exploration, and revised Moodboards and personalization profiles worked with the new generation.

Members could select V6.1 when they preferred its look or needed an older workflow. Midjourney used V7 as the default until June 2026, when V8.1 took its place.

June 18, 2025: Video V1 Adds Image-to-Video Creation

Midjourney launched its first video model one day after V7 became the default image model. Video V1 starts with a Midjourney image or an uploaded picture. The Animate action produces short video variations, and members can extend a result through additional generations.

Video V1 started a separate model sequence for motion. Image models such as V7 supplied the starting frames, and the video model animated them inside the paid web service.

2026: Niji 7 and the V8 Generation

January 9, 2026: Niji 7 Refines Illustration and Character Work

Niji 7 improved coherence, prompt adherence, line work, eyes, reflections, and small background elements. The model adopted a cleaner, flatter visual tendency that emphasized its line quality. Its more literal response also supported specific designs and repeatable characters.

March 17, 2026: V8.0 Alpha Opens for Testing

Midjourney opened an early V8 model on a separate Alpha site. V8.0 followed detailed directions more closely, improved coherence and text rendering, and generated images about five times faster than the previous generation under the tested settings.

The Alpha also introduced native 2K output through HD mode and a higher-compute quality option. Midjourney kept V8.0 in temporary Alpha testing and moved on to V8.1 the following month.

April 14, 2026: V8.1 Alpha Revises the V8 Model

V8.1 replaced the first Alpha and returned to a more familiar V7-like aesthetic. Style References and Moodboards behaved more consistently. HD generation used less time and GPU capacity. Standard-resolution jobs also ran faster at a lower cost.

V8.1 restored image prompts and image weights. Midjourney also added a prompt shortener for inputs beyond the model’s length limit and updated Describe to produce prompts suited to V8 behavior.

April 30, 2026: V8.1 Reaches the Main Website and Discord

V8.1 moved from the separate Alpha environment to midjourney.com and Discord. Midjourney also updated sharpness and image quality, with the clearest changes in HD output, Style References, and Moodboards. The default changed six weeks later.

June 11, 2026: V8.1 Becomes the Default

Midjourney announced V8.1 as the automatic image model. Standard-definition images could appear in about four seconds. HD jobs produced images at twice the dimensions and four times the pixel count of V7 output.

Prompt adherence, coherence, and text rendering also improved. Members used V7 Omni Reference during development of a V8 replacement. Midjourney also scheduled the temporary V8.0 Alpha for retirement.

June 16, 2026: Draft Mode Arrives for V8.1

V8.1 received a new form of Draft Mode built around larger concept batches. Each job creates 24 images at reduced resolution and quality. A selected concept can then run through Vary at full resolution and quality.

Draft jobs use half as many Fast hours as standard-definition V8.1 jobs. On the same day, Midjourney released --preview for temporary tests of unfinished model features. Preview output can change between tests and has no stable version status.

How Access to Midjourney Changed

The 2022 public beta depended on Discord. Members joined the server, entered commands in channels or direct messages, and received image grids in the conversation. The Discord Bot supports image creation, settings, reference parameters, and many editing actions.

Direct web creation began with a limited Alpha in December 2023 and opened to all members in August 2024. The website now has visual settings, an image tray, creation and organization feeds, folders, personalization profiles, Moodboards, Editor, and video controls. Some options have different names or workflows on Discord.

Both interfaces require a Midjourney subscription for normal creation. A Discord account is only required when a member wants to work through Discord. Web users can sign in with Google, although returning members should use their original Discord account first if they want their earlier image history.

How Midjourney’s Focus Changed

Early Midjourney prompts often produced surprising, heavily stylized images. The first model generations concentrated on basic coherence, subject knowledge, resolution, and closer control over how literally the system followed a prompt.

V6 and V7 put more control in the member’s hands. References carried a character or style across images, Personalization learned individual taste, and region editing made local revisions within an existing image.

The web platform brought generation, organization, editing, and animation into one workspace. V8.1 concentrated on everyday production concerns such as speed, prompt accuracy, readable text, high-resolution output, and quick batches of draft ideas.

Midjourney FAQs

Is Midjourney free to use?

A paid subscription is required for image and video creation on midjourney.com and through the Midjourney Discord Bot. The separate niji journey mobile app may offer a limited trial. That trial applies only to the app.

Do you need Discord to use Midjourney?

New members can create on midjourney.com with a Google or Discord sign-in. A Discord account is required for the Midjourney Bot. Returning members with an older Discord history should sign in through that account before connecting another login method.

Does Midjourney have an official API?

Midjourney offers no general public API. Its rules prohibit unauthorized automation and third-party services that automate interaction with Midjourney. Rare exceptions require explicit authorization from the company.

Is Midjourney open source?

Midjourney’s image and video models are proprietary. The company keeps the model weights and training code private, and subscribers create through Midjourney’s web or Discord infrastructure.

Can I use older Midjourney versions?

Several legacy image models are selectable through the version setting or a version parameter. Feature support varies by generation, especially for editing, references, resolution, and personalization. The current model selector shows which versions are available.

What is the difference between Midjourney on the web and Discord?

Both interfaces use the same subscription and support core image generation. The website has visual settings, image organization, personalization management, Editor, and direct video controls. Discord emphasizes typed commands, community channels, custom option sets, and bot-based workflows. Some features exist on only one interface or use different controls.

Are Midjourney images private?

Midjourney is open by default, so creations can appear on the Explore page even when they were made in a private Discord server or direct message. Pro and Mega members can use Stealth Mode to hide them on the website. Images posted in a shared Discord channel are visible to people in that channel.

Related AI Timelines

• ChatGPT Timeline: Follow OpenAI model releases, ChatGPT product changes, image generation, and video milestones.
• Anthropic Claude Timeline: Track Claude model generations, product access, Claude Code, and Anthropic platform updates.
• Google AI Timeline: Review the development of Google’s major AI research, models, and consumer products.
• Meta AI Timeline: See how FAIR, Llama, Meta AI, and Meta’s generative media work developed over time.