Machine learning algorithms designed to generate uncanny or bizarrely surrealistic images are taking the tech world by storm, and things will only get more complicated in the future. Case in point: there’s a new version of Midjourney –a “totally new” AI architecture –which is just the beginning of a journey towards “something deep and unfathomable.” Like with DALL-E, Stable Diffusion, and other AI algorithms, Midjourney can take a simple or very complex text prompt and turn it into a composite image. Led by David Holz, who previously founded the Leap Motion motion control venture, the Midjourney team was recently able to stand out in the crowd of image-generating algorithms by winning the Colorado State Fair’s annual art exhibition.
In the official announcement posted to Midjourney’s Discord server, Holz said that version 4 had just entered its alpha stage. Midjourney v4’s model has “vastly more knowledge” of creatures, places, and everything. It has a much better grip on small details, can handle more complex textual prompts, and is better with multi-object and multi-character scenes. The updated software adds support for advanced functionalities like image prompting and multi-prompts. Midjourney v4 can generate much more realistic pictures than previous models, so much so that the team was forced to adapt stricter rules about images that can “cause drama.” The new Midjourney algorithm was trained on a new AI supercluster and has been in the works for over nine months. Users should expect quirks and inconsistencies during the alpha test while the team is pushing significant changes toward full launch.
Even though it’s still in the alpha stage, Midjourney v4 has already received a warm welcome from users and wannabe “creators.” The new AI model seems to be sensibly better at interpreting textual prompts. Some say it feels “too easy” to use with results that can appear premade and not just AI-generated. The improved AI could spell a bleak future for “Pro prompters who will lose their new job created one month ago,” a Reddit user joked. As for the Midjourney team, the new AI model will soon see many improvements, including increased resolution, controllable upscaling, custom aspect ratios (currently limited to 1:1), increased image sharpness, fewer text artifacts, and improved default styling. The journey into the unfathomable has just begun.
