Today, March 16, the new version of the most popular neural network Midjourney, now called Midjourney V5, was officially released. The developers said that their new development received a lot of improvements and upgrades – for example, the authors of the project fixed the main problem with working out the details of the fingers and limbs in the frame. Whereas previously AI-powered technology could not handle the adequate position of human fingers, now the technology can easily draw pictures of people without unnecessary artifacts. Also, the developers report a higher level of elaboration of the facial features and eyes of a person – there should be more details.
Moreover, Midjourney V5 allows you to create more elaborate and detailed images, although the time to generate each image, according to the developers, will be reduced. And, what is most interesting, now the system, when generating an image at the preliminary stage, will show two different images, by which the user can determine the general style and choose what artificial intelligence should use next. There was also an upgrade in terms of resolution – now Midjourney V5 allows you to create images with a resolution of 1024 × 1024 pixels, while in the previous version of the neural network the resolution was 512 × 512 pixels.
In the final version of the neural network, users will be able to generate images at a resolution of 2048×2048 pixels, but the developer himself is not yet ready to say when this version will be released. The authors of the neural network also added that their system now better understands user requests, which should have a positive impact on the quality of the final content, plus the output will be more diverse. Additionally, Midjourney V5 added support for seamless textures, a wide aspect ratio, a more detailed dynamic range, and expanded the set of styles. Accordingly, the already worthy images have become even better.