LoRA
What is lora?
Low-rank adaptation - a small add-on trained on top of an image model to teach it a specific face, product or style. LoRAs are one technique for keeping an AI model's identity consistent.
Also called low-rank adaptation, fine-tune, fine-tuning
Related terms
- Locked identityKeeping an AI model's face, body, skin and hair the same in every generated image, instead of producing a slightly different person each time.
- Digital twinAn AI replica of a real person, trained from their photos with their consent, so brands can generate new images of them.
- Diffusion modelThe class of AI model behind most modern image generators.


