Open Source · Python · GPU Accelerated

Cadenza

A video editor for the stage

A GPU-accelerated video editor built in Python. Designed for multi-camera concert and performance editing — with automatic audio sync, Lumetri color, and NVIDIA NVENC hardware export.

View on GitHub Quick Install

Capabilities

Built for performance.
Ready for the stage.

🎵

Audio Sync

FFT cross-correlation automatically aligns multi-camera recordings. Eyeball position first, then let the algorithm find the exact frame offset — accurate to within a single frame.

🎨

Lumetri Color

Full color grading with temperature, tint, exposure, contrast, highlights, shadows, and saturation. Applied in real-time via GPU compositing.

⚡

NVIDIA NVENC Export

Hardware-accelerated H.264 and H.265 export using NVENC. Exports 4K at ~28fps — roughly 3× faster than software encoding.

🎬

Multi-Track Timeline

Unlimited video and audio tracks. Drag, trim, razor cut with link-aware editing. Rubber band selection, multi-clip drag, and full undo/redo.

🔊

Audio Mixing

Per-clip volume in dB with a Premiere-style logarithmic slider, stereo pan, mute, and waveform display normalized to 0 dBFS.

💾

Project Save / Load

Full serialization of clips, in/out points, effects, envelopes, and track layout. Every razor cut and color grade is preserved.

Technology

Python.
GPU.
Open source.

Built entirely in Python using modern libraries. PyTorch drives the GPU compositor — every frame is composited on CUDA before being handed to NVENC for encoding. No C++ required.

PySide6UI framework (Qt6)

PyTorch + CUDAGPU compositing

PyAVVideo decode / encode

NVENCHardware H.264/H.265

scipyFFT audio sync

numpyAudio / waveform processing

Get Cadenza

Choose your path

        Just use it
      

Windows Installer

Download the pre-built Windows executable. No Python, no setup — extract the zip and run Cadenza.exe.

⬇ Download Cadenza.zip

Windows 10/11
NVIDIA GPU recommended
~2.5 GB download · ~4.5 GB extracted

        For developers
      

Run from Source

Clone the repo and run with Python. Requires conda and an NVIDIA GPU for full performance.

# Setup
conda create -n cadenza python=3.11
conda activate cadenza
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121
pip install PySide6 av numpy scipy

# Run
git clone https://github.com/rjansendc/Cadenza.git
cd Cadenza && python main.py

View on GitHub →

CPU-only mode works without an NVIDIA GPU — export will be slower.
Tested on Windows 11 · Python 3.11 · CUDA 12.1