Tipo
Checkpoint
Data di pubblicazione
2023-05-18
Modello Base
SD 1.5
Introduzione alla versione
All previews made using Schizo Negative. No embeddings!
Improved backgrounds.
Improved hand/feet construction.
More of the original training shows through.
Generally a more flush skin tone than v2.
More vibrant eyes.
Better recognition of hair/eye color.
More consistently matches ear color to hair color if present.
A fair bit less consistent for good outputs, but when it gets it good, it gets it really good.
Baked in ClearVAE NanLess.
Kittenchow based LoRAs have not been tested in this version.
