Midjourney releases V7, its first new AI image model in nearly a year
▼ Summary
– Midjourney has launched its new AI image model, V7, which began its alpha rollout recently.
– V7 requires users to create a “personalization” profile by rating around 200 images, enabling customization to individual visual preferences.
– The model offers two modes: Turbo (more expensive) and Relax, and introduces Draft Mode for faster, lower-cost image rendering.
– V7 features improved text prompt handling, higher image quality, and better coherence in details, but lacks some standard features like image upscaling, which will be added later.
– Midjourney, founded in 2022, is expected to generate $200 million in revenue in 2023 and faces lawsuits over alleged copyright infringement.
Midjourney, a pioneering name in AI-based image generation on the web, has launched its first new AI image model in nearly a year.
The model, called V7, began its alpha rollout around midnight EST on Thursday. This release comes just a week after OpenAI introduced an image generator in ChatGPT that went viral due to its ability to create Ghibli-style images.
While V7 is not specifically optimized for Ghibli-style art, it is still capable of producing visually appealing creations, at least to this writer’s novice eye.
To start using V7, you’ll first need to rate approximately 200 images to create a Midjourney “personalization” profile, if you haven’t already. This profile customizes the model to your individual visual preferences; notably, V7 is the first Midjourney model to have personalization enabled by default.
Once set up, you can enable or disable V7 on Midjourney’s website or, if you are a member of Midjourney’s Discord server, through its Discord chatbot. On the web app, you can easily select the model from the drop-down menu adjacent to the “version” label.
In a post on X, Midjourney CEO David Holz described V7 as having a “totally different architecture.”
“V7 is […] much smarter with text prompts,” Holz elaborated in a Discord announcement. “[I]mage prompts look fantastic, image quality is noticeably higher with beautiful textures, and bodies, hands, and objects of all kinds exhibit significantly better coherence in all details.”
V7 is available in two modes: Turbo (more expensive to run) and Relax. It also powers a new tool called Draft Mode, which renders images at 10 times the speed and half the cost of the standard mode. While Draft images are of lower quality compared to standard-mode images, they can be enhanced and re-rendered with a single click.
According to Holz, several standard Midjourney features are not yet available for V7, including image upscaling and retexturing. These features are expected to be added within the next two months.
“This is an entirely new model with unique strengths and probably a few weaknesses,” Holz wrote on Discord. “We want to learn from you what it’s good and bad at, but definitely keep in mind it may require different styles of prompting. So play around a bit.”
Image Credits: Midjourney
In my brief testing, V7 followed my prompts fairly well, though I didn’t have sufficient time to fully explore the model’s capabilities.
Midjourney is a unique venture. Founded in 2022 by David Holz, who also co-founded the PC peripheral company Leap Motion, it has not raised any external funding.
In late 2023, it was reported that Midjourney was on track to generate around $200 million in revenue. The San Francisco-based company recently announced the formation of a hardware team to work on undisclosed projects and continues to train models for video and 3D object generation.
However, the company faces multiple lawsuits alleging that it infringed on the rights of millions of artists by training its AI tools on images scraped from the web without the creators’ consent.
Source: TechCrunch
Topics
