LoRA Training

Train your own custom AI model using your images. Set a style, name it, and use it across your boards.

What is LoRA Training? 🧠

LoRA stands for Low Rank Adaptation. It is a way of fine tuning an AI model on a specific set of images so it learns and replicates a particular style, aesthetic, or visual language.

Instead of training an entire AI model from scratch, LoRA takes an existing model and adapts it to your inputs. Faster, lighter, and just as powerful.

Why it matters for architects and designers

Your work has a voice. The way you sketch, render, and present is uniquely yours. LoRA lets you teach xFigura that voice so every generation feels like it came from you, not a generic AI.

Here is what you can train it on:

• Your drawing style — upload your hand sketches and train a model that generates in your exact line weight, style, and composition

• Your rendering aesthetic — moody, photorealistic, watercolor, graphic. Train on your best renders and replicate that look instantly

• A project style — train on reference images for a specific project to keep every output visually consistent

• A firm identity — studios can train a shared LoRA that reflects the firm's signature aesthetic across the whole team

Examples 💡