camfiltergpu-dev/custom-filter-library

This repository is a library of example Python scripts for building custom video filters that plug into a separate proprietary application called CamFilterGpu. CamFilterGpu itself is described as a live video enhancer with foreground and background separation, color grading, GPU-accelerated effects, and AI-based tracking, meant to work smoothly with tools like OBS, Zoom, and Discord as a virtual camera source. The main CamFilterGpu application is not open source and is not part of this repository. It is a paid product with a free trial, downloaded separately from its own website, and this repository only supplies the scripting layer that lets users write their own effects for it. The scripts here hook into CamFilterGpu's GPU-accelerated video pipeline using standard Python along with libraries like OpenCV, numpy, and onnxruntime, which come built into the main application. The repository includes seven baseline example filters meant as starting points and teaching examples: a newsprint-style halftone effect, a moving black and white or color noise overlay, a live typing effect where letters float upward on screen, an effect that turns a person's outline into ASCII text with adjustable size and color, a face and hand tracking effect that overlays a polygon mask and hand skeleton, an extreme sharpening effect, and a planned lightning-between-fingers effect that is not finished yet. To use a script, you download the desired Python file, place it in CamFilterGpu's custom filters folder, launch the CamFilterGpu application, and select the filter from its built-in Node Pipeline Editor to apply it to your live camera feed. Because CamFilterGpu itself handles the GPU acceleration and virtual camera routing, anyone editing these scripts can focus purely on writing the video processing or tracking logic in Python, using any text editor or IDE. The example scripts themselves are released under the MIT license, meaning they can be freely copied, modified, and reused in personal or commercial projects.