ComfyUI-Fayens Brings Cinematic Polish to Face Swaps

A large smooth matte white face silhouette cutout resembling a clean face mask used for swapping.

ComfyUI-Fayens is a new collection of custom nodes for ComfyUI that streamlines face swap workflows from start to finish. It automatically extracts clean face crops, generates accurate masks, and prepares images for reference-based latent conditioning. The pack also bundles color correction, cinematic lighting, and an aspect ratio helper for quick output sizing.

Developer iFayens built the nodes to work directly with popular face models like InsightFace and InSwapper. The toolkit is designed for fast, consistent swaps—just load a reference image and a target, and the nodes handle the rest. iFayens shared the project on GitHub as an open-source resource, encouraging community feedback and use.

From crop to cinematic finish

Key Features
  • Fast face crop extraction with masks.
  • Color matching in LAB and HSV.
  • Sharpening, contrast, and gamma controls.
  • Cinematic lighting: teal, orange, vignette.
  • Film grain and noise effects.
  • Quick aspect ratio presets for any format.
  • Optimized for InsightFace and InSwapper models.

This toolkit suits creators who want faster face swaps without sacrificing visual quality. Because everything runs locally, privacy-minded users keep sensitive images completely off the cloud. Small studios and serious hobbyists will find the post-processing effects especially useful for adding a polished, professional look with very little extra effort.

What you need to know

Installation is straightforward: clone the repository into ComfyUI’s custom_nodes folder and install the requirements. The nodes depend on InsightFace models such as antelopev2, and they integrate smoothly with FLUX-based pipelines. iFayens emphasizes that while the tools work on medium-quality images, they really deliver top realism when fed high-resolution inputs.

"It works on medium quality images, but really shines on high quality inputs for the best and most realistic results." — Source: Reddit