We use LoRA models with stable diffusion models. They can be trained quickly with minimal 30-50 images of your subject. (https://civitai.com/articles/2099/lora-models-and-how-to-use...)