Based on 3 capability dimensions
4 versions
Midjourney V8 is approaching launch, with the final round of image rating/ranking underway to calibrate personalization systems. The update also includes a new personalization interface on web allowing users to create profiles by scrolling through images, and Personalization and Moodboards features added to niji 7 anime model.
Most accurate prompt following yet. Coherent text in images. Better realism and fine details. Long prompt support.
Photorealistic images. Much better at hands and faces. More accurate prompt interpretation. Style consistency improved.
Major quality leap. More coherent images and better understanding of prompts. Became the gold standard for AI art.
3 versions
Multimodal Diffusion Transformer architecture. Best text in images of any SD model. Improved composition and detail.
Higher resolution images. Better image composition and faces. Two stage model pipeline for better quality.
Open source image generation released publicly. Runs locally on consumer hardware. Sparked the open source AI art revolution.