MidjourneyV7AlphaArriveswithPers
作者：大自然的搬运工 · 2026-07-11 20:48:46
Midjourney has officially launched the alpha version of its V7 image generation model, marking its first major architectural update in nearly a year. The release prioritizes deep personalization, requiring users to create a profile by rating around 200 images to unlock the model’s full potential. The standout feature is a new default "Draft Mode" that renders images ten times faster and at half the cost, though users can enhance these drafts with an upscale button. Alongside V7, the team introduced a voice conversation mode for the first time, allowing subscribers to speak prompts aloud and hear AI-generated audio descriptions of their results. This feature is currently limited to the web platform. Midjourney CEO David Holz described the model as "a totally different architecture" that is smarter with text prompts and significantly more beautiful, though he cautioned that it is an early alpha lacking features like inpainting and image reference. The update also debuts "Image Personalization," which learns individual aesthetic preferences to tailor outputs, moving the platform beyond generic results. Looking ahead, the company plans to release updates every one to two weeks, with V7’s advanced features rolling out quickly. This rapid iteration cycle signals Midjourney’s intent to compete aggressively in an increasingly crowded generative AI market.