Midjourney launches its first AI image model with personalization named V7

Midjourney, one of the pioneering platforms in AI-generated imagery, has rolled out its first new image model in nearly a year. The new model, dubbed V7, entered alpha testing around midnight ET on Thursday and brings with it a suite of upgrades aimed at both personalization and performance.

Announced by Midjourney CEO David Holz on Discord and X (formerly Twitter), V7 is described as being built on a “totally different architecture” compared to its predecessors. The release comes shortly after OpenAI’s latest image generator made waves online for its ability to produce Ghibli-style illustrations. While V7 does not explicitly aim to replicate that aesthetic, it is garnering early praise for its visual quality and responsiveness to prompts.

One of V7’s standout features is personalization, a first for Midjourney. Users are required to rate approximately 200 images to establish a personalization profile that tailors output to their individual visual preferences. Unlike previous models, this feature is enabled by default in V7, allowing for more consistent results aligned with user taste.

Users can activate the model on Midjourney’s website or via the company’s Discord server. Within the web app, V7 is accessible through a version selector in the interface. Holz noted that V7 offers significant improvements in text prompt interpretation, image coherence, and the rendering of complex elements such as hands, bodies, and intricate textures.

In addition to the standard rendering modes, V7 introduces a new feature called Draft Mode. This mode is designed to produce images up to 10 times faster and at half the computational cost of standard rendering, albeit at a slightly reduced quality. Draft Mode is ideal for rapid prototyping and can later be enhanced with a single click.

V7 is available in two configurations: Turbo, which consumes more resources but delivers higher responsiveness, and Relax, which is more cost-effective. Some features common to previous Midjourney versions—such as upscaling and retexturing—are not yet available for V7 but are expected to be rolled out in the coming months.

“This is an entirely new model with unique strengths and probably a few weaknesses,” Holz wrote. He encouraged users to experiment with different prompt styles to fully explore the model’s capabilities, emphasizing that community feedback will shape further development.

With V7, Midjourney aims to solidify its standing in an increasingly competitive AI art landscape.