Experimental LoRA on a real person trained with random images I found on the internet. A lot are not very HQ, 170+ images, but still good results.

I have erased all metadata on sample images to avoid legal issues, but I'll leave 1 sample settings for reference:

parameters

<lora:dilraba-epoch5:1> ultra realistic 8k raw photo, (photorealistic:1.4), masterpiece, dilraba, white nurse uniform, looking at viewer

Negative prompt: (worst quality, low quality:1.3), old, logo, watermark

Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 377146066, Size: 512x768, Denoising strength: 0.5, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+