Midjourney announced the V6.1 of its Image Models, which brings a host of enhancements and new features.
In V6.1, users will notice more coherent images, with improved rendering of complex subjects like arms, legs, hands, bodies, plants, and animals. According to the announcement, the image quality has been significantly enhanced, with reduced pixel artifacts and better textures, skin tones, and retro 8-bit aesthetics.
Small features in images, such as eyes, small faces, and distant hands, are now more detailed and lifelike. New upscales have been added with improved image and texture quality. Standard image jobs are also around 25% faster, so creating is faster than ever.
Text accuracy has also been improved when drawing words via “quotations” in prompts. A new model of personalization introduces improved nuance, surprise, and accuracy to the personalization model and data from any job; this comes with personalization code versioning, where users can use any personalization code from old jobs.
A new –-q 2 mode has been added, which takes 25% longer to process but can sometimes add more texture, albeit at the cost of reduced image coherence. Overall, images should look generally more beautiful across the board.
However, V6.1 currently does not include a new inpainting/outpainting model. If you use zoom, reframe, repaint, or vary region, it will fall back to the V6.0 model. Also, if you prefer to switch back to V6, simply go to /settings and click V6, type—v 6 after your job, or click the settings panel on the website.
Looking ahead, the Midjourney team plans to release a V6.2 model in the next month or so, which will feature further improvements.
