-
Alwayes use: "<lora:bh4d_bh4bie:1>, bh4d bh4bie," for activate Lora
-
Tested with models:
edgeOfRealism_eorV20Fp16BakedVAE (excelent results),
realisian_v40 (excelent results),
-
Training lora: max img 512 x 512
-
Use: x:512 x y:768 (excelent results)
-
Use low sampling steps, cfg scale low,
-
DPM++ SDE Karras
-
Use best negative prompts for a perfect result
-