Nightingale: Turn Your Local Music Library Into an AI Karaoke Machine
Ever wished you could just sing along to your favorite tracks without hunting for karaoke versions? Or maybe you want to practice vocals on obscure albums that never had official instrumentals released. Nightingale is a clever open-source project that uses AI to separate vocals from any song in your own music collection, turning it into an instant karaoke experience.
No cloud dependency. No uploading your private files. Just your local library, processed on your own machine.
What It Does
Nightingale takes songs from your local music library and runs them through a vocal separation model (likely using something like Demucs or a similar AI source separation tool). It strips out the vocals, leaving you with a clean instrumental track. Then it gives you a simple interface to play that instrumental while displaying synchronized lyrics—if available—so you can sing along.
The whole pipeline lives on your machine. You point it at a folder of music, it processes each track, and you get a web UI or command-line tool to browse and play the karaoke versions.
Why It’s Cool
Privacy first. Since everything runs locally, your music never leaves your computer. No Google, no Spotify uploads, no sketchy third-party servers. If you care about keeping your collection private, this is the right approach.
Works with anything. Got a rare bootleg, a live recording, or an album from a Bandcamp artist who never released instrumentals? Nightingale can handle it. The AI separation is surprisingly good these days—it not only removes vocals cleanly on most pop/rock tracks, but also preserves the backing vocals and instrumentation well enough to sing along.
No monthly subscription. Unlike those online karaoke services that charge $10/month for a limited catalog, Nightingale is free and open source. You pay in compute time (and maybe a GPU if you want faster processing), not in recurring fees.
Customizable. Because it's a GitHub project, you can tweak the separation model, change the UI, or integrate it into your own music player setup. Devs can hook it into Plex, Jellyfin, or just run it as a standalone karaoke server.
How to Try It
Head over to the GitHub repo for instructions. Typically you'll:
- Clone the repo.
- Install dependencies (Python, PyTorch, and some audio libraries).
- Point it at your music directory.
- Run the processing script (may take a while depending on your CPU/GPU).
- Open the local web interface and start singing.
The README should have the exact commands. If you have an NVIDIA GPU with CUDA, the separation will be much faster. On CPU, expect to wait a few minutes per song.
Final Thoughts
Nightingale solves a real pain point: wanting to sing along to music that doesn't have official instrumental versions, without sacrificing privacy or paying for a subscription. The local-first approach is refreshing, and the AI separation quality these days is genuinely impressive.
If you're a developer who also happens to be a karaoke enthusiast or just likes tinkering with local AI tools, this is a fun weekend project. You could even extend it with your own improvements—like adding a queue system, better lyric syncing, or integrating it with a home media server.
For now, grab your collection, let the model do its thing, and finally belt out that deep cut you've always wanted to sing along to.
Follow us at @githubprojects for more cool open-source finds.
Repository: https://github.com/rzru/nightingale