danielravina/stemkit

What it solves

StemKit solves the problem of isolating individual instruments or vocals from YouTube videos. It allows you to take a full song and break it down into separate tracks—like drums, bass, guitar, piano, and vocals—so you can create your own karaoke tracks, acapellas, or instrumentals without needing professional studio software or cloud-based subscriptions.

How it works

StemKit acts as a local audio-processing hub. You provide a YouTube link, and the app uses built-in tools to download the audio and run it through specialized AI separation engines (such as Mel-Band Roformer and Demucs) directly on your computer. It then displays the results in a synchronized interface where you can play the video and adjust the volume, mute, or solo each individual instrument track in real-time.

Who it’s for

This tool is for musicians, producers, DJs, and music enthusiasts who want to remix, study, or practice specific parts of a song. Because it runs entirely offline and requires no accounts or API keys, it is ideal for users who prioritize privacy and want to keep their audio processing local.

Highlights

• Fully Local Processing: All audio splitting happens on your machine; your files and searches are never uploaded.
• YouTube Integration: Search for songs or paste links directly into the app.
• Flexible Controls: Mute, solo, and adjust volume for individual stems with a synchronized video player.
• One-Click Presets: Quickly switch between configurations like "Karaoke," "Acapella," or "Drums + Bass."
• Export Options: Save any or all isolated stems as high-quality WAV files.
• Hardware Acceleration: Supports GPU acceleration via NVIDIA (CUDA), Apple Silicon (MPS), and experimental AMD (ROCm) for faster processing.