Midjourney launches V7 alpha, its new AI image generation model with faster rendering

Midjourney has launched V7 Alpha, its first new AI image model in nearly a year. To use V7, users must rate about 200 images to create a "personalization" profile unless they already have one, marking the first Midjourney model with default personalization features. Users can activate or deactivate V7 through Midjourney's website or Discord chatbot by choosing from the "Version" drop-down menu.

CEO David Holz said V7 features a "totally different architecture" with smarter text prompts, better image quality, and improved coherence in rendering bodies, hands, and objects. It comes in two versions, Turbo (faster and more expensive) and Relax, and introduces "Draft Mode," which renders images 10 times faster at half the cost, with lower initial quality that can be enhanced later. Image upscaling and retexturing are not yet available but are expected within two months.

This release follows OpenAI's new image generator in ChatGPT using GPT-4o, which quickly went viral for its high-fidelity Ghibli-style creations, while Holz highlighted that V7 brings unique strengths and possible weaknesses, encouraging users to adapt their prompting styles.