Midjourney Releases V7 Image Generation Model in Alpha, Comes With a Faster and Cheaper Draft Mode

Midjourney released its latest image-generation artificial intelligence (AI) model, V7, on Friday. The V7 AI model is currently available in an alpha version, and the company is currently letting its user base test the model's capability. Arriving nearly a year after its predecessor V6, the V7 Alpha model is also the first in the AI firm's lineup to have personalisation turned on by default. The foundation model also introduces a new “Draft Mode,” which is said to cost half as much as the standard mode while offering 10 times the speed.

Midjourney's V7 Brings New Modes and Features

In a post on X (formerly known as Twitter), the official handle of Midjourney announced the release of V7. Compared to V6, the image generation model has a higher text prompt adherence, and both the image quality and textures are better, the company added. The AI firm also claimed that the coherence among different elements is higher than its predecessor.

Midjourney highlighted that V7 is its first model to have model personalisation turned on by default. It means that before getting access to the AI model, users will have to rate a large number of images to build their personalisation profile. This profile enables the Midjourney models to understand the user's preference, and it can adjust the outputs accordingly. However, once a user gets access to V7, they can turn personalisation on or off.

The company is also introducing a new feature dubbed Draft Mode with this model. Draft Mode is said to be half the cost of the standard mode and renders images at 10 times its speed. It is essentially an iteration mode where users can give the model multiple prompts to get the desired output and to make inline edits to images. The mode supports voice input for a conversational experience.

Users can either click “Draft mode” on the web client or add “--draft” at the end of the prompt to activate the mode. Midjourney highlighted that draft mode images are lower quality than the standard mode, but the model behaviour and aesthetic generation are consistent.

The V7 image generation model is available in Turbo and Relax variants. Notably, Turbo mode costs twice as much as standard output generation. Additionally, the V7 model currently does not support upscaling, editing, and retexture-related tasks, and it redirects to V6 to complete these tasks.

Midjourney also plans to release new features for the AI model every week or two for the next 60 days. One feature in the roadmap includes character and object reference.