For the best fidelity, likeness, and realism:

Don't use this LoRA for Text-to-Image.

Instead, go to Edit section (Not Image-to-Image). Set the base model to Qwen-Image-Edit 2509, apply this LoRA, and use a blank image (1328 x 1328 px recommended; available in the showcase above) as the input image

• Strength: 1.0 - 1.2

• If using Lightning LoRA, set CFG scale to 1 (necessary).

Pro Tip: Results are much better in ComfyUI, where you can combine this with regular Qwen image LoRAs.