Tipo
LORA
Fecha de lanzamiento
2025-12-12
Modelo básico
Flux.1 D
Model Parameters
clip_skip：1
Parámetros de entrenamiento
Epochs：1
Steps：0
Introducción de versión
NiwaStyle Asian girl II (Flux)
Strength : 0.6 - 1.4
>Advanced Config<
Sampling Steps 25-40
scale CFG : 3.5-8
_______________________________
Face only
Low-step LoRA
Trained with: 20 high-quality images
Training Steps: 1000 steps
Dataset: 3/4 Body, Half Body, Close-Up,
Lighting: Soft Tone, Daylight & Warm Light
Notes on Checkpoint Compatibility:
- Works best with portrait/realistic face checkpoints.
- For general checkpoints: use weight 0.7–1.0.
- For face-focused checkpoints: increase weight to 1.2–1.4 to enhance identity and details.
- With anime or stylized checkpoints: use lower weight (0.6–0.8) to prevent over-strength.
