tschnz/Live-Video-Magnification

Real‑time Eulerian video magnification: amplify subtle motion & color from webcam or video. Linux/Windows/macOS

What it solves

LiViM amplifies subtle motion and color changes in video that are otherwise invisible to the human eye, such as blood flow (pulse) or microscopic vibrations. It allows users to detect these changes live from a webcam or within a recorded video file.

How it works

The project implements Eulerian video magnification techniques based on research from MIT CSAIL. It uses three distinct algorithms to process video frames:

  • Motion (Laplace): Uses a Laplacian pyramid and temporal IIR bandpass filters to amplify motion.

  • Motion (Phase): Uses a Riesz pyramid and Butterworth bandpass filters to reduce noise while amplifying motion.

  • Motion (Color): Uses a Gaussian pyramid and FFT bandpass filters over a rolling window to amplify color changes.

To maintain performance, the software provides both CPU and GPU pipelines (via WebGPU), with the GPU engine being over 200x faster on modern devices.

Who it’s for

Researchers, videographers, and hobbyists interested in signal processing and motion amplification for analyzing subtle physical changes in video footage.

Highlights

  • Real-time processing: Supports live webcam capture and video file playback with GPU acceleration.

  • Flexible comparison views: Offers processed, original, and side-by-side or top-bottom synced views.

  • Headless mode: A CLI tool for exporting processed video without the GUI.

  • Customizable parameters: Fine-grained control over amplification strength, frequency bands, and spatial cutoffs.

  • Multi-platform support: Prebuilt packages for Windows, Linux, and macOS.