MidJourney V4 — Ups and Downs
With prompts and examples
Evolution (or even: a revolution?) of generative AI goes on. 2021 OpenAI published CLIP as OpenSource. From my perspective, it was the beginning of the AI art revolution: people began to create, explore, and combine CLIP with VQGAN, StyleGAN, etc.
In later 2021, DiscoDiffusion, Pytti, and other mesmerizing AI models followed.
Then, in 2022 — Discord-based MidJourney was published in March 2022 — DALL·E 2, later #StableDiffusion.
And we are still at the beginning.
MidJourney caught my attention after seeing more impressive visuals popping up everywhere. This Discord-bot-based model was published in three versions, which of them had its unique look if prompted without many modifiers.
I mean, I liked them, but after a while, the generic generations became boring: too perfect, too illustrative, and yet nothing new after a while. Most of the faces looked in this way (generically beautiful):