Method of making this Lora:
Downloaded ~500 of top results for https://www.pexels.com/search/portrait/
Resized them to max 1024px tall/width (I used powertoys bulk resize)
Manually went through them and discarded photos which were too abstract, had multiple people in it, or the faces were not fully visible
Used Kohya_ss to caption the photos with BLIP
Trained here on CivitAI (SD 1.5 Realistic default settings)
Intended to be used with face swap / custom face Lora.
Enjoy.