Empfohlene Parameter

samplers

Euler, DPM++ 2M

steps

33 - 39

cfg

1 - 3.5

resolution

832x1216

other models

FS_v3 (5d4d457c3321), Flux1Compact_Dev_Fp8 (2eda627c8a)

Versions-Highlights

Updated and enlarged collection of training images, quality has been improved. I also completely changed the training settings. Now LoRA works more smoothly, affecting more gently. The hands are better than in previous versions.

Angular jaws that look like irons, huge square chins, Brezhnev-style eyebrows, razor-sharp cheekbones, pumped-up plump lips and bony collarbones - you won't find any of that here. Because this LoRA is dedicated to the ladies of the 16th-19th centuries and concepts of beauty of that time.

Zurück
Hassaku XL (Illustrious) - v2.2
Weiter
The Vanguard - V1

Modell-Details

Modelltyp

LORA

Basismodell

Flux.1 D

Modellversion

v3.0

Modell-Hash

8592ff9597

Trainierte Wörter

FS

Ersteller

Diskussion

Bitte log in um einen Kommentar zu hinterlassen.