floyo logobeta logo
Powered by
ThinkDiffusion
floyo logobeta logo
Powered by
ThinkDiffusion

ComfyUI-LBM

50

Last updated
2025-05-27

A custom node for ComfyUI, this tool implements the Latent Bridge Matching (LBM) algorithm to facilitate rapid image relighting. It streamlines the relighting process through efficient image-to-image translation in a single step.

  • Supports quick relighting with minimal inference time.
  • Integrates depth and normal map generation capabilities.
  • Allows for selective processing using masks and offers multiple precision settings.

Context

This tool serves as a custom node within ComfyUI, specifically designed for image relighting tasks utilizing the LBM algorithm. Its primary purpose is to enhance the efficiency and quality of relighting images by transforming their lighting conditions in a streamlined manner.

Key Features & Benefits

The key features of this tool include fast image relighting achieved through a single inference step, which significantly reduces processing time compared to traditional methods. Additionally, it supports depth and normal map generation, expanding its utility in various image processing tasks, and provides a simplified workflow that enhances user experience.

Advanced Functionalities

The tool is capable of handling multiple precision options (fp32, bf16, fp16), allowing users to choose the best balance between performance and quality based on their system capabilities. Moreover, it includes mask support for selective image processing, enabling users to focus on specific areas of an image while applying relighting effects.

Practical Benefits

By utilizing this tool, users can significantly improve their workflow in ComfyUI, as it delivers high-quality relighting effects with reduced computational overhead. The automatic model downloading feature and optimized memory usage further enhance efficiency and ease of use, allowing for a more seamless experience in image processing tasks.

Credits/Acknowledgments

The LBM model is sourced from the Hugging Face Model repository, with its original implementation available on GitHub. The foundational paper detailing the LBM method is authored by Clément Chadebec, Onur Tasar, Sanjeev Sreetharan, and Benjamin Aubin. This tool was created by the 1038lab team and is released under the GNU General Public License v3.0 (GPL-3.0), while the LBM model is under the Creative Commons BY-NC 4.0 license.