Midjourney, one of the first AI-based image generators on the web, has released its first new AI image model in nearly a year.
The model, called the v7, began to unfold in Alpha around midnight on Thursday. It’s coming a week after Openai debuted a new image generator with ChatGpt.
But Midjourney’s is not a Gibrio Petitized model, but at least not official, but nevertheless, it can produce aesthetically pleasing pieces, at least in the eyes of this reporter’s Dilettante.
We are currently in the alpha test phase for our new V7 image models. This is our smartest, most beautiful, and most coherent model. Try it and look forward to updates every or two weeks for the next two months. pic.twitter.com/ogqt0fgiy7
– Midjourney (@midjourney) April 4, 2025
To use it, you must first evaluate around 200 images and still build around 200 “personalized” profiles. This profile adjusts the model to suit your individual visual preferences. The V7 is the first model of Midjourney to turn on personalization by default.
Once you’ve completed it, you can turn V7 on or off on the Midjourney website, and if you’re a member of Midjourney’s Discord Server, in the web app in Discord Chatbot, you can quickly select your model from the drop-down menu next to the “Version” label.
Midjourney CEO David Holz described V7 as “a totally different architecture” in an X post.
“V7 is […] Much smarter at text prompts,” Holtz continued in his announcement about Discord.[I]The magician’s prompts look great, the image quality is remarkably high with beautiful textures, and all kinds of bodies, hands and objects have very good consistency for all details. ”
The V7 comes in two flavors: Turbo (the cost of running) and Relax, driving a new tool called Draft Mode, which renders the speed and half of the cost 10 times. Draft images are of lower quality than standard mode images, but can be clicked to enhance and re-rendered.
According to Holz, many standard Midjourney features are still not available in the V7. He said they will arrive in the near future, perhaps within two months.
“This is an all-new model with unique strengths and perhaps some weaknesses,” Holtz writes incongruously. “[W]e wants to learn from you whether it’s good or bad, but you may definitely need a variety of styles of prompts. So play a little. ”
In my brief test, the V7 adhered reasonably well to my prompt, but I didn’t have time to actually pass the model through the ringer.
Midjourney is an unusual operation. It was started in 2022 by Holz, who co-founded the PC-neighboring company Leap Motion, but has not raised any external money.
In the second half of 2023, Midjourney reportedly expected to generate revenue of around $200 million. Recently, the San Francisco-based company has set up a hardware team to tackle projects that have not been detailed and continues to train previously in models of video and 3D object generation.
The company faces several lawsuits accusing millions of artists of infringing the rights of millions of artists by training AI tools for images that have been scraped from the web without the consent of image creators.
