Midjourney launches its new V7 AI image model that can process text prompts better

Midjourney has released the alpha version of V7, which it says is an “entirely new” AI image generation model and is much smarter at processing your text prompts. The image quality of its output is noticeably higher, the Midjourney team says, and can create better textures, bodies and hands. AI image generators typically struggle with creating accurate depictions of hands, but based on photos posted by some users on their socials, V7 is capable of spitting out some photorealistic images of human hands.

The new model comes with a feature called “Draft Mode,” which can render images at half the speed the program usually takes. Its results are rougher and less detailed, but it will cost half of what a standard generation costs. Midjourney says Draft Mode is the best way to iterate on ideas. You can use it, say, if you’re collaborating with someone and just spitballing ideas with them — it works with voice, so you don’t even have to type out each other’s suggestions — or if you’re unsure what kind of vibe you’re going for. If you like an image Midjourney creates in Draft Mode, you can click “enhance” or “vary” on it to re-render it at full quality.

At launch, V7 will have Turbo and Relax modes for standard rendering, with the former costing twice as much as a normal speed job on the V6 model. Midjourney needs more time to optimize its standard speed mode for V7, but it will make the option available in the future. The new mode is missing more capabilities, as well, including upscaling, editing and retexturing, which will fall back to the program’s V6 model for now. Midjourney promises to roll out new updates for the model every week or two over the next two months.

To be able to test the alpha version of V7, you’ll have to unlock your personalization profile first. Midjourney describes personalization as “a style assistant for your image creations,” since it teaches the AI your visual preferences. You’ll have to rank at least 200 pairs of images to create a V7 Global Personalization Profile and test the model. Personalization is switched on by default for the V7, but you’ll be able to switch it off if you want.