Lora trained on dataset from Nano Banana Pro. She is not a real person.

main preset: strength 0.7

more natural editorial/lifestyle: 0.6

stronger face lock: 0.75

Thanks to @Huslyo123 for the Idea and prompt.