Midjourney, a pioneer in AI image generation on the web, released its first AI image model since nearly a year.
Dubbed V7is the model that began rolling out as an alpha on Thursday at midnight ET. This comes a week following OpenAI’s debut of a new image creator in ChatGPT, which quickly became viral for its ability create Ghibli style photos.
Midjourney is not a Ghibli optimized model — at least, not officially — but can still generate aesthetically pleasing work, at least according to this reporter’s amateur eye.
The alpha-test phase for our new V7 Image Model is now underway. It’s the smartest, most beautiful and most coherent model we have ever created. Expect updates every two weeks or so for the next two month. pic.twitter.com/Ogqt0fgiY7
— Midjourney (@midjourney) April 4, 2025
If you haven’t done so already, you will need to create a Midjourney profile of “personalization”. This profile allows you to customize the model according to your personal preferences. V7 is Midjourney’s first model with personalization turned on by default. Once you have done this, you can turn V7 on and off on Midjourney’s website, or, if you are a member of Midjourney’s Discord server chatbot, if you are logged in. You can select the model in the web app by using the drop-down next to “Version”.
Midjourney’s CEO David Holz described V7 in a report as a “totally new architecture”. Post on X
Holz continued, “V7 is… much smarter with its text prompts,” in an announcement made on Discord. “[I]mage’s prompts look fantastic. Image quality is noticeably better with beautiful textures. And bodies, hands, objects, and all other kinds of details have significantly improved coherence.” Draft images are lower quality than standard mode images, but can be enhanced with a single click.
According to Holz, a number of Midjourney standard features are not yet available for V7. These include image upscaling, retexturing, and more. He said that they will be available in the near future. They could arrive within two months.
Holz wrote in Discord that “This is a completely new model with unique strength and probably a couple of weaknesses”. “[W]I want to know what you think about it. But keep in mind that it may need different prompting styles. Play around with it a little.”
In my brief testing, V7 adhered reasonably well to my prompts, though I didn’t have time to really put the model through the ringer.
Midjourney is an unusual operation. Started in 2022 by Holz, who co-founded PC peripheral company Leap Motion, it hasn’t raised a dime of outside money.
In late 2023, Midjourney was According to they are expecting around $200 million in revenues. The San Francisco-based company recently announced that it was establishing hardware teams to work on projects it did not disclose, and that it continued to train models previously announced for video and 3D objects generation.
This company is facing Several Lawsuits accuse the company of violating the rights of millions artists by training AI on images scraped off the web without the consent from the images’ creators.
Kyle Wiggers, TechCrunch AI Editor. His writings have appeared in VentureBeat, Digital Trends and a variety of gadget blogs, including Android Police and Android Authority, Droid-Life and XDA-Developers. He lives in Manhattan, with his music therapist partner.
View Bio
