LoRA (Low-Rank Adaptation)
Image & Video AIA lightweight fine-tuning technique that trains a small adapter on top of a base model, commonly used to add specific styles or characters to AI image generators.
LoRA is a technique for customizing AI models without retraining the entire thing. Instead of updating billions of parameters (expensive), LoRA freezes the base model and trains a small set of new parameters that modify the model's behavior.
In the image generation world, LoRAs are incredibly popular. Civitai hosts thousands of LoRAs that add specific capabilities to Stable Diffusion: a particular art style, a specific character's likeness, a clothing style, or a pose. LoRA files are small (10-200 MB) compared to full checkpoints (2-7 GB), so users can collect and swap many of them.
LoRA is also used for text model fine-tuning (QLoRA is a memory-efficient variant), making it possible to customize large language models on consumer hardware.
Real-World Example
On Civitai you'll find thousands of LoRAs — small add-on files that give Stable Diffusion new abilities like generating specific art styles or character likenesses.
Related Terms
More in Image & Video AI
FAQ
What is LoRA (Low-Rank Adaptation)?
A lightweight fine-tuning technique that trains a small adapter on top of a base model, commonly used to add specific styles or characters to AI image generators.
How is LoRA (Low-Rank Adaptation) used in practice?
On Civitai you'll find thousands of LoRAs — small add-on files that give Stable Diffusion new abilities like generating specific art styles or character likenesses.
What concepts are related to LoRA (Low-Rank Adaptation)?
Key related concepts include Fine-tuning, Checkpoint, Stable Diffusion. Understanding these together gives a more complete picture of how LoRA (Low-Rank Adaptation) fits into the AI landscape.