Midjourney has announced the release of version 5 of its commercial AI image-synthesis service. This new version is capable of producing photorealistic images of the highest quality, which some AI art enthusiasts have described as creepy and “too perfect”. The service is available as an alpha test for Midjourney subscribers, who can access it through Discord.

Midjourney v5 has already received positive feedback from graphic designers, such as Julie Wieland, who stated that it feels like “getting glasses after ignoring bad eyesight for a little bit too long”. This version is an improvement over its previous versions, as demonstrated by a comparison between Midjourney v3, v4, and v5 using the prompt “a muscular barbarian with weapons beside a CRT television set, cinematic, 8K, studio lighting.” The results show that each iteration adds more detail to the generated images.

Midjourney generates images based on text descriptions called “prompts” using an AI model trained on millions of works of human-made art, similar to image synthesizers such as Stable Diffusion and DALL-E. However, Midjourney was recently at the center of a copyright controversy involving a comic book that used earlier versions of the service.