Midjourney V7 Released: A Major AI Image Model Update

Midjourney Launches its Newest AI Image Model: V7

Midjourney, a pioneering AI image generation platform, has recently unveiled its first significant model update in almost a year.

The new model, designated V7, began its initial rollout in alpha around midnight ET on Thursday. This release follows closely after OpenAI’s debut of a novel image generator within ChatGPT, which rapidly gained attention for its capacity to produce images in the distinctive style of Studio Ghibli.

Key Features of Midjourney V7

While not specifically optimized for Ghibli aesthetics – at least not in an official capacity – V7 is capable of generating visually appealing imagery. Initial impressions suggest a high level of artistic quality.

To begin utilizing V7, users will initially be asked to evaluate approximately 200 images. This process establishes a Midjourney “personalization” profile. This profile customizes the model’s output to align with individual visual preferences. V7 marks the first Midjourney model to have personalization enabled by default.

Access to V7 can be enabled or disabled through the Midjourney website. Discord server members can also activate it within the platform’s chatbot interface. The web application provides a simple drop-down menu for model selection next to the “Version” label.

Architectural Improvements and Performance

Midjourney CEO David Holz characterized V7 as representing a “totally different architecture” in a post on X.

Holz further elaborated on Discord, stating that V7 demonstrates “much smarter” handling of text prompts. He also noted that image prompts yield exceptional results, with noticeably enhanced image quality and improved textures. Furthermore, the model exhibits significantly better coherence in rendering bodies, hands, and various objects.

V7 is offered in two variants: Turbo (which incurs higher operational costs) and Relax. It also introduces a new tool called Draft Mode. Draft Mode renders images ten times faster and at half the cost of the standard mode. While Draft images have a lower resolution, they can be refined and re-rendered with a single click.

Future Updates and Considerations

Certain standard Midjourney features are not yet available for V7. These include image upscaling and retexturing. Holz indicated that these functionalities will be implemented in the near future, potentially within two months.

“This is an entirely new model with unique strengths and probably a few weaknesses,” Holz explained on Discord. He encouraged users to provide feedback on the model’s performance, while also cautioning that it may necessitate adjustments to prompting styles. Experimentation is encouraged.

Initial testing suggests that V7 responds effectively to prompts, although a more comprehensive evaluation was not possible at the time of this report.

Company Background and Legal Challenges

Midjourney operates under a unique business model. Founded in 2022 by David Holz, previously co-founder of Leap Motion, the company has not sought external funding.

In late 2023, Midjourney projected revenues of approximately $200 million. The San Francisco-based company recently announced the formation of a hardware team to pursue undisclosed projects. It continues to develop models for video and 3D object generation.

The company is currently involved in several legal disputes. These lawsuits allege copyright infringement, claiming that Midjourney’s AI tools were trained on images scraped from the web without obtaining consent from the original artists.