On Wednesday, Midjourney announced version 5 of its commercial AI image compositing service, which can generate lifelike images at the quality level some AI art aficionados call it. is scary and “too perfect”. Midjourney v5 is currently available as an alpha test for customers who subscribe to the Midjourney service, available via Discord.
Julie Wieland, a graphic designer who often shares her Midjourney designs on Twitter, said: “MJ v5 now makes me feel like I finally have glasses after ignoring poor eyesight for a while now. too long time”. “Suddenly you see everything in 4k, it’s strangely overwhelming but also amazing.”
Wieland has shared some of his Midjourney v5 builds with Ars Technica (see below in the gallery and in the main image above) and they certainly show an advancement in visual detail since Midjourney came out. Released in March 2022. Version 3 is out in August and Version 4 is out in November.
Midjourney works similar to image compositing tools like Stable Diffusion and DALL-E in that it generates images based on text descriptions called “reminders” using a trained AI model on millions of man-made works of art. Recently, Midjourney has been at the center of a copyright controversy over comics using earlier versions of the service.
After testing with v5 for a day, Wieland noted improvements including skin texture and “extremely realistic” facial features; more realistic or cinematic lighting; better highlights, reflections and shadows; more expressive angles or the overall view of a scene and “more refined, near-perfect eyes”.
Over the past year, the idea that AI art creators can’t properly draw hands has become something of a cultural thing. Notably, Midjourney v5 can produce realistic human hands quite well. “Most hands are fine, with five fingers instead of 7-10 on one hand,” says Wieland.
In the service’s Discord release notes, Midjourney also notes that v5 now responds with a “much wider range of styles” than v4, and is also more responsive to prompts, producing less unwanted text want more and provide the image resolution multiplied by 2.
If the Midjourney upgrade has a visual drawback for AI art fans, it’s probably because the visuals can be so realistic and “perfect” that a model’s accuracy will be lost. take away some of the thrill of continuously generating AI images to find a match, what can be called the “slot machine effect”. Though one Twitter user by the name of Philipp Lenssen noted, “If you think of a particular visual theme, it’s still like the lottery. But with a higher hit rate than v4.”