LoRA (Low-Rank Adaptation)
A LoRA is a small add-on that fine-tunes an AI image or video model to reproduce a specific style, character, product, or face without retraining the whole model.
LoRA (Low-Rank Adaptation) is a lightweight way to teach a base model something new — a brand's product, a consistent character, or an art style — using a small trained file layered on top of the model instead of a full, expensive retrain.
It's the standard technique for character and brand consistency: train a LoRA on a handful of reference images, then generate endless on-brand variations. LoRAs are small, shareable, and can be combined.