Midjourney releases V7, its first new AI image model in nearly a year

Midjourney introduces V7, a new AI image model offering enhanced features and personalization.

: Midjourney has launched version 7 or V7, marking its first new AI image model release in almost a year. This updated model includes advanced features like personalization profiles requiring users to rate 200 images, improving image coherency and text prompt efficiency. V7 offers two options, Turbo and Relax, with the introduction of Draft Mode for faster, budget-friendly rendering, despite lacking some features like image upscaling. Midjourney, founded by David Holz, operates independently, recently projected $200 million revenue, and faces legal challenges over alleged copyright infringements.

Midjourney has introduced its latest version, V7, representing the company's first new AI image model release in approximately a year, announced around midnight EST on Thursday. This unveiling comes shortly after OpenAI's new image generator in ChatGPT gained rapid attention for its capability to produce Ghibli-style photos, although V7 does not officially specialize in such styles. However, it promises aesthetically pleasing outputs, gaining praise from various industry observers despite its different positioning from OpenAI's offering.

The new model necessitates users to establish a Midjourney “personalization” profile by rating about 200 images, a step aimed at tailoring the model to individual visual preferences. Personalization is a key feature of V7, switched on by default, significantly improving its responsiveness to text prompts. Users can activate or deactivate V7 via Midjourney’s website or its Discord server, making access and utility streamlined for the community. This level of customization marks a notable advancement from previous versions and highlights Midjourney's ongoing commitment to innovation and user-centric enhancements.

David Holz, Midjourney CEO, shared insights into the development of V7, describing it as a “totally different architecture.” This new AI model demonstrates improved image quality, flawless coherence across image details, and extraordinary responsiveness to text and image prompts. V7 elevates texture and fine detail reproduction, enhancing the overall output quality of generated images. Holz also announced the model's availability in two versions: Turbo, which incurs higher operational costs, and Relax. A novel Draft Mode is also introduced, providing image rendering at tenfold the standard speed and at half the cost, yet at slightly reduced quality, allowing users to opt for efficiency.

While some standard features like image upscaling and retexturing are currently unavailable in V7, Midjourney plans to introduce these in upcoming updates expected within two months. Holz encourages users to experiment with the model, acknowledging that its revolutionary features come with distinct strengths and potential weaknesses. This openness to user feedback underscores Midjourney's strategy of iterating its technology based on genuine user experiences and demands.

Operating without external funding since its inception in 2022 by Holz, who has a background in PC peripherals through Leap Motion, Midjourney is a rarity in the rapidly consolidating tech space. Despite facing lawsuits accusing the company of rights infringement for utilizing web-sourced images without creator consent, the firm is reported to be nearing $200 million in revenue as of late 2023. Besides focusing on image generation, Midjourney is venturing into hardware developments and expanding its model training to include video and 3D object generation, positioning itself at the forefront of generative AI innovation.

Sources: TechCrunch, The Verge, Wired

Generative AItechnologyinnovation