Midjourney V7 is a glow-up you’ll actually notice
Matt V
April 07, 2025
MIDJOURNEY
Midjourney has just released V7, its first new AI image model in almost a year.
The alpha version began rolling out on 4 April, not long after OpenAI’s latest image tool made headlines for its Ghibli-style results.
Midjourney’s model isn’t built specifically for that style, but it still delivers strong visual quality.
What makes V7 different is its new architecture and built-in personalisation.
Before using it, you’ll need to rate around 200 images to create your visual profile: once done, you can switch the model on through Midjourney’s website or Discord server.
CEO David Holz says V7 offers sharper textures, better object details (including those tricky hands), and more accurate results from both text and image prompts.
There’s also a new “Draft Mode” that speeds up image generation by 10x and cuts costs in half, though the initial images are lower in quality and meant to be refined later.
Here’s what to keep in mind:
V7 needs a quick setup to personalise results
Draft Mode is cheaper and faster, with images you can refine
Some key features are still on the way
Built different
Some features like upscaling and retexturing aren’t live yet but should arrive in the next couple of months.
Holz also pointed out that V7 might respond best to slightly different prompts than earlier models, so it’s worth experimenting.
Midjourney continues to run independently, without outside funding.
It was expected to bring in around $200 million in revenue last year and has started work on new tools for hardware, video, and 3D content.
The company is also dealing with ongoing lawsuits over the use of images scraped from the web during training.
Watch me rate 200 images like it’s a BuzzFeed quiz.