Glossary
LoRA (Low-Rank Adaptation)
A lightweight fine-tuning method that teaches a base image model a new face, style or object using a tiny add-on file.
LoRA stands for Low-Rank Adaptation. It is a technique introduced by Microsoft Research that lets you fine-tune a giant model (like Flux or Stable Diffusion XL) by only training a small set of low-rank matrices instead of the entire network. The result is a 10–200 MB file that, when loaded on top of the base model, gives the model new knowledge — your face, a clothing brand, a unique painting style.
On Synexa, LoRAs are how we train your AI influencer: you upload 15–30 reference photos and we produce a LoRA that preserves the subject's identity across thousands of new prompts.
Compared to full fine-tuning, LoRAs are roughly 100× cheaper, transferable between checkpoints, and stackable — you can combine an identity LoRA with a style LoRA in the same render.