Midjourney Debuts V7: First New AI Image Model in a Year

Midjourney, one of the earliest and most popular AI-based image generators on the web, has just launched its first major model update in nearly a year. Dubbed V7, the new model began rolling out in alpha around midnight EST on Thursday, marking a significant milestone for the San Francisco-based company.

The release comes on the heels of OpenAI’s latest viral image generator inside ChatGPT, which captivated users for its ability to render Ghibli-style illustrations. While Midjourney’s V7 isn’t officially tuned for Ghibli-like aesthetics, it’s already generating visually striking images with remarkable detail and coherence.

A Smarter, More Personalized Midjourney

V7 isn’t just an incremental update—it’s a complete architectural overhaul. According to Midjourney CEO David Holz, the new model is “much smarter with text prompts” and delivers higher image fidelity, improved textures, and more anatomically accurate renderings of hands, bodies, and objects. Users will also notice better visual consistency, a common challenge in earlier versions.

One of the standout features of V7 is personalization by default. Before diving in, users are prompted to rate around 200 images, helping the system fine-tune itself to individual visual tastes. This personalization profile informs how the AI interprets prompts, making outputs more aligned with each user’s unique aesthetic preferences.

Once personalization is set, users can activate V7 directly on Midjourney’s web app or via its Discord bot. The web interface offers a simple drop-down selector to toggle between versions.

New Tools and Modes: Turbo, Relax, and Draft

V7 introduces two core operating modes: Turbo and Relax. Turbo delivers faster results at a higher computational cost, while Relax offers a slower but more cost-efficient rendering experience.

A new Draft Mode has also been added, enabling users to generate images at 10x the speed and half the cost of standard mode. While draft-quality images are lower resolution, they serve as a fast prototyping tool and can be enhanced later with a single click.

However, some familiar Midjourney features—such as image upscaling and retexturing—aren’t yet available in V7. Holz notes that these features will return in future updates, likely within two months.

Experimental, Yet Promising

“This is an entirely new model with unique strengths and probably a few weaknesses,” Holz shared on Discord. He encouraged users to explore and adapt their prompting style, as V7 may respond differently compared to its predecessors.

Early tests suggest the model is fairly robust even with challenging prompts. Sample images based on inputs like “a dragon with a spiky tail” or “a Ghibli-style character” display notable improvements in rendering complexity, color balance, and character design.

The Bigger Picture

Midjourney remains an anomaly in the AI space. Launched in 2022 without any venture capital, the company reportedly approached $200 million in revenue in 2023—all without external funding. It's also expanding into hardware development and training models for video and 3D object generation, signaling broader ambitions beyond static imagery.

That said, Midjourney—like other generative AI platforms—is navigating legal challenges. The company faces lawsuits over alleged copyright infringement, as its models were trained on vast collections of online images without explicit permission from creators.

What's Next

V7 represents a bold step forward for Midjourney. With smarter prompt handling, enhanced personalization, and rapid prototyping tools, the new model is a major leap in AI image generation. While still in its early stages, V7 signals where AI creativity is headed—and how fast it's evolving. As Holz put it, now’s the time to play, explore, and help shape what comes next.