The launch of a new version of the Midjourney neural network, which was called Midjourney V5
Starting today, our community can test Midjourney V5. This version has much higher image quality, more varied output, wider stylistic range, support for seamless textures, wider aspect ratio, wider dynamic range, and more.
The new version will show two preview images, prompting you to choose which style you prefer more.
Of the obvious noticeable changes, improved faces and eyes with a lot of small details are promised, as well as (finally!) More accurately worked out fingers and limbs. Recall that previous versions often generated six fingers, and their shape was often strange. Images in general will be more detailed, and the generation time will be reduced.
Midjourney V5 will create images up to 1024 x 1024 pixels (up from 512 x 512 pixels in the previous version). The final version of the neural network should create images of 2048 x 2048 – although we do not yet know when this will happen.