Nvidia’s LorWeb editor learns styles from image triplets

Nvidia has introduced an image editing method that learns styles and transformations from just three images: a before shot, an after shot, and a new target. By seeing how one photo was transformed, the tool can apply the same look such as line-art, clay renders, or comic shading to any other image.

The system is built on top of the LorWeb framework, which treats edits as LoRA-like modules that can be mixed and matched. This allows creators to stack multiple transformations or reuse styles across different projects without retraining full models.

Nvidia has published full GitHub instructions, making it straightforward for users to experiment with custom style transfer in their own workflows.