Flux.1-dev LoRA trained on real photos of a person for realistic portraits and likeness capture.

Trigger word: xlyshx

Recommended strength: 0.8 - 1.1 (start at 1.0, lower if overtrained/stiff face)

Best used with: natural language prompts like "ultra realistic portrait, detailed skin texture, natural daylight, sharp focus"

Trained with ~20 high-res images, rank 32, ~1500 steps. Great for headshots, full body with some flexibility.