IMPORTANT: since i'm transitioning model from hyperrealistic to stylized hyperrealistic - it might generate overcooked and oversaturated images using 84k vae, so using of anythingv3 vae or any other anime vae is reccommended for now.
-retrained on 50% of concepts with better captioning.
-trained further, refined and broadened knowledge base.
-fixed several places where it lacked style.
-trained out large quantity of cartoon and classical painting styles.
-further pushed stylized hyperrealism.
-slightly increased influence of RFKTR's SineTempore.