Midjourney is currently testing an alpha version of its v7 image generation model.
According to the company, the new version is designed to improve both text and image prompt handling, with more coherent and detailed renderings of textures, bodies, hands, and other objects.
For the first time, model personalization is enabled by default in Midjourney. However, users must complete an onboarding process before they can even use v7. This involves selecting preferred images from 200 randomized pairs. Only after completing this step you can use the v7 model.
The personalization feature is intended to capture users’ individual aesthetic preferences and interpretations of visual content. However, the system does not consider prompt context during image evaluation, which may lead users to prefer different styles—such as illustrations or photographs—depending on the task.
Introducing draft mode for rapid prototyping
Midjourney is also introducing a new feature called "Draft Mode," which allows users to generate images roughly ten times faster than before at half the cost of a standard v6 job. While the resulting images are lower in quality, Midjourney says they are "consistent in behavior and appearance," making them suitable for early-stage concept development and brainstorming.
Users can upgrade draft images they like to higher quality or create variations with a single click. The mode can be activated using the "--draft" parameter command, though it only functions with v7.
Video: Midjourney
On the web interface, draft mode activates a "conversational mode" that lets users modify images through voice commands. For example, users can prompt changes such as replacing an animal or adjusting the time of day. Midjourney automatically interprets these spoken instructions and starts a new image generation job accordingly.
Two modes and future developments
The v7 model currently supports Turbo and Relax modes. Turbo jobs are priced at twice the cost of a standard v6 job, while Draft jobs are half the price. A standard mode for v7 is still in development and is expected to roll out soon.
While core features like mood boards and "--sref" already work with v7, several functions including upscaling, image editing, and retexturing still rely on v6 models. Updates for these features are planned.
Looking ahead, Midjourney plans biweekly feature releases over the next two months. A new reference system for characters and objects tops the priority list, and the company will involve its community in ranking future development priorities.
