golden ratio face

Training goal: Develop a LoRA capable of adjusting your facial features at will to achieve a series of golden proportions. Through extensive testing, various faces can be generated based on different base models, thus achieving the freedom to modify the golden ratio face.

After testing the best weight 0.6-0.8, above 0.8 will have a chance of collapse,

CLIP skip: 2

For a better drawing, please add the trigger word: JLSQ

model recommendation: AWPortrait_V1.4, majicMIX realistic

VAE: vae-ft-mse-840000-ema-pruned.safetensors

Sampling method: DPM + + 2M Karras

Drawing size: 512 * 512

The iterative steps 28  

cfg 7

The amplification algorithm suggests 4x-Ultrasharp

Welcome to return the picture, your return picture is the power of my creation!!