LORA

2K Models

14K Posts

LoRA (Low-Rank Adaptation) is a compact fine-tuning format that teaches a base model a new style, character, or concept without retraining it. A LoRA file is usually only a few megabytes, loads on top of a checkpoint like FLUX, Illustrious, Pony XL or SDXL, and can be stacked with other LoRAs in the same generation. This tag is the largest LoRA collection on Tensor.Art: style LoRAs that reshape the whole look of an image, character LoRAs trained on a specific person or OC, concept LoRAs for poses, outfits and compositions, and utility LoRAs such as detail tweakers. Every LoRA page lists its trigger words and recommended weight — paste the trigger into your prompt, set the weight (0.6–1.0 works for most), and generate online or download the file for local use in ComfyUI and A1111. Not sure where to start? Filter by base model to match your checkpoint, or sort by run count to see what the community actually uses.