Midjourney has released its new V7 image generation model in alpha, offering major improvements in understanding text prompts, rendering image details, and overall output quality. The model now interprets user instructions with greater accuracy, producing visually richer and more coherent results. Notably, the updated system displays significant advancements in generating realistic hands, bodies, textures, and objects.
One of the standout features in V7 is built-in personalization. Though users must initially unlock it—a process taking about five minutes—it can be turned on or off at will. This feature enables the AI to better recognize and replicate individual preferences, making it more attuned to user intent. Midjourney sees this as a breakthrough in tailoring generative AI experiences.
In tandem with V7, Midjourney is introducing ‘Draft Mode’—a new image generation option that’s ten times faster and costs half as much as standard generations. While images produced in this mode are of lower quality, they retain the style and behavior of higher-resolution outputs. Users can instantly enhance or vary draft images, allowing for rapid iteration during the creative process.
Draft Mode also brings in a “conversational mode” and voice input functionality. Users can now speak commands or request changes to existing images directly via the web interface. For instance, asking the system to replace a cat with an owl or change the scene to nighttime will automatically adjust the prompt and generate a revised image.
Initially, V7 will be available in two-speed settings: Turbo and Relax. Turbo mode delivers faster results at double the standard cost, while Draft Mode remains a more budget-friendly alternative. The regular speed mode is undergoing further refinement and will be rolled out soon.
Other features like upscaling, editing, and retexturing will temporarily revert to the V6 model until updated. Meanwhile, mood board creation and SREF tools are active and set to improve with upcoming updates. Midjourney has promised frequent enhancements over the next two months, including a new character and object reference system.
Users are encouraged to experiment with V7’s new capabilities, as it may require different prompting approaches compared to earlier versions.
