These are beautifully realistic faces created with StyleCLIP, using as source small images generated with Stable Diffusion prompts. These small images where later processed with Real-ESRGAN using the face enhancing feature, then feeding the resulting image to StyleCLIP (sometimes a modifier was added for longer/shorter hair or changing the expression).
You will have noticed that there are few men, if any, but it has a simple explanation: first, as a man myself, I prefer feminine beauty; and second, StyleCLIP seems to almost always come up with ugly men. 🙂
Oh… and yes, some are a bit scary. Sorry!