LoRA (Low-Rank Adaptation) is a mathematical technique used to reduce the number of parameters that need to be trained when fine-tuning Stable Diffusion models. It works by creating a diff of model instead of saving the whole thing.
To use LoRA, one must place the LoRA file in stablediffusion-ui\models\Lora and use Pivotal Tuning Inversion CLI provided with lora. LoRA models, checkpoints, hypernetworks, textual inversions, embeddings, Aesthetic Gradients, and LORAs can be found on Civitai.com.
There are several resources available to learn and use LoRA in image generation. Some of them are: