There are already several LoRA models for generating photos of this type on Civitai. My model focuses on high image quality and strong realism. When generating images, you should play around with the parameters, especially strength_model. Currently, the faces under the mask cannot be controlled via the prompt, but I am working on that... Have fun with the model!
Description
Improved model, more realistic faces, more details, better prompt following.
You have to play with strength and guidance. 2nd lora is possible now if compatible, see examples.
Have fun with this lora!