Glossary

LoRA (low-rank adaptation)

A LoRA, short for low-rank adaptation, is a small add-on trained on top of a large AI model to teach it one specific thing, such as a face, a product or a look.

See our faces and cast

How a LoRA works A large base AI model plus a small LoRA trained on one face gives the same face in every shot. Base AI model knows people in general + LoRA one face Same face, every shot Switch the LoRA off and the base model is unchanged
A LoRA sits on top of a large model and teaches it one face, so the same person appears in every shot.

Retraining a whole image model is slow and costly. A LoRA instead learns a thin layer of adjustments from a focused set of reference images and sits on top of the base model. Switch it on and the model reliably reproduces that face or style; switch it off and the model is unchanged.

In fashion it is the usual way to keep one person consistent. Without it, an AI fashion model drifts from shot to shot: the jaw softens, the eyes change colour, the hairline moves. With it, the same face can appear across a catalogue and the next season's too.

How we handle LoRA

Our team trains and maintains these adaptations for brands that want faces or a look of their own, so every shoot uses the same people. You never handle the files: you approve the faces and we keep them identical, through our exclusive faces and cast service.

Related terms: diffusion models, AI fashion models and prompt engineering

Every term in the glossary

Chat avatar
Book a free creative call Talk to our creative team
Call
WA
Email