Every Song on Your Computer Just Became Karaoke — Open-Source AI App Nightingale Is Here

Pick any MP3 file saved on your computer. Nightingale uses AI to strip out the vocals, display lyrics on screen automatically, and — if you plug in a microphone — score your pitch in real time. No karaoke machine, no subscription. Your own computer becomes the karaoke room.

How AI Builds a Karaoke Track in 3 Steps

The way Nightingale works is surprisingly straightforward.

▲ Nightingale playback screen. Song lyrics are highlighted in real time while a beautiful background video plays.

Why It Beats a Traditional Karaoke Machine

Compared to existing karaoke apps or machines, Nightingale stands out for several reasons.

Any song works. Traditional karaoke limits you to a pre-loaded catalog, but Nightingale turns every music file on your computer into a karaoke track. It supports nearly every audio format — MP3, FLAC, WAV, M4A — and even video files like MP4 and MKV, meaning you can sing along to a music video playing in the background.

Adjustable guide vocals. You can freely adjust the volume of the original vocals (separated by AI) from 0 to 100 %. Crank it up when you're learning a new song, then dial it down as you gain confidence.

Stunning visuals. The app ships with five GPU-shader background themes (plasma, aurora, waves, nebula, and starlight) plus free background videos from Pixabay (underwater, space, nature, city, abstract) that it downloads automatically.

Gamepad support. Hook it up to a TV and control everything with an Xbox or PlayStation controller. Perfect for living-room karaoke nights with friends. You can create multiple player profiles and track per-song scores on a leaderboard.

Installation & Usage — One File Is All You Need

One of Nightingale's biggest strengths is that setup is dead simple. Just download a single executable.

The first launch downloads AI models (a few hundred MB), so an internet connection is required. After that, it runs completely offline. Your song files are never uploaded anywhere.

A GPU makes it faster. With an NVIDIA GPU, CUDA acceleration processes a song in 2–5 minutes; without one, CPU-only processing takes about 10–20 minutes. Once a song is processed, it's cached, so there's no wait the next time. On Mac, Apple Silicon's MPS acceleration (Metal Performance Shaders) is used automatically.

What Open Source Makes Possible

Nightingale is an open-source project licensed under GPL-3.0. It's written in Rust for speed and built on the Bevy game engine for silky-smooth visuals.

Unlike paid karaoke apps, there's no monthly fee, no song-count limit, and you can turn your entire music library into a karaoke collection. The project was first released on March 15, 2026, and has already hit the front page of Hacker News.

Every AI technology it uses is also proven open source:

• Demucs — A source-separation AI built by Meta (formerly Facebook). It can isolate vocals, drums, bass, guitar, and more into individual tracks.
• WhisperX — An improved version of OpenAI's Whisper speech recognition, capable of word-level timing alignment.
• UVR Karaoke Model — A vocal-separation model optimized specifically for karaoke. Runs efficiently via ONNX Runtime (a high-performance inference engine).

Who Is This For?

If you need singing practice, the guide vocals and pitch scoring let you train systematically on your own. If you love hosting house parties, connect it to a TV, grab a gamepad, and turn your living room into a karaoke bar. If you're a music teacher, you can instantly turn any song a student brings in into a karaoke track for class.

Genre doesn't matter — Western pop, K-pop, indie, you name it. If the song is on your computer — even one you recorded yourself — it can be turned into karaoke.

Related Content — Get Started with AI the Easy Way | Free Learning Guide | More AI News