These models are essential for generating realistic images. Extensive training is required for the LoRA to capture the nuances and imperfections of human appearance.

A base model cannot be trained this way, as facial features and body characteristics inevitably become averaged out against the model's other weights.

Choose the option you like best.