At a Glance:
Nightingale is a self-hosted karaoke application that automatically separates vocals, transcribes word-level lyrics, and scores pitch from songs in a local folder, Jellyfin, or Navidrome library, shipped as a single binary with automatic dependency bootstrapping.
Overview:
Nightingale is a desktop karaoke application that turns songs from a user's local music library into a full karaoke experience. It scans a music folder, Jellyfin server, Navidrome server, or a self-hosted web library, then uses neural networks to separate lead vocals from instrumentals and transcribe lyrics with word-level timestamps. The application provides synchronized lyric highlighting, real-time pitch scoring, and playback controls including key and tempo shifting. It is distributed as a single binary that automatically downloads and bootstraps all required dependencies, including ffmpeg, Python, PyTorch, and machine learning models, on first launch.
Key Decision Points:
Deployment model: Runs as a local desktop application, with an optional self-hosted web mode for access from phones, tablets, and other devices on a home network.
Music sources: Supports local folders, Jellyfin servers, Navidrome servers, and self-hosted web libraries, with songs cached locally after the first play from remote sources.
Hardware requirements: Analysis is GPU-accelerated via CUDA or Apple Silicon MPS, but falls back to CPU; a song typically takes 2–5 minutes on GPU and 10–20 minutes on CPU.
Lyrics sourcing: Can automatically transcribe and align lyrics, or fetch existing synced lyrics from LRCLIB when available, with a built-in editor and browser for selecting matches.
Platform support: Pre-built binaries are provided for Linux (x86_64, aarch64), macOS (ARM, Intel), and Windows (x86_64); in-app updates are supported on macOS and Windows only.
Core Features:
Stem separation: Isolates lead vocals from instrumentals using the UVR Karaoke model or Demucs, with adjustable guide vocal volume; backing vocals are preserved in the instrumental track.
Word-level transcription: Generates synchronized lyrics with word-level timestamps via WhisperX, with per-character forced alignment and romanization for CJK languages.
Pitch scoring: Provides real-time microphone-based pitch detection with star ratings and per-song scoreboards tracked across player profiles.
Key and tempo control: Allows post-analysis adjustment of song key and tempo, with playback variants cached for quick switching.
Audio-reactive visuals: Includes 10 GPU shader backgrounds that react to microphone input, themed Pixabay video loops, and support for using video files as synchronized backgrounds.
Gamepad support: Full navigation and playback control via gamepad, including D-pad, sticks, and face buttons.
Use Cases:
Setting up a karaoke station on a home network by running Nightingale in self-hosted web mode, allowing access from phones, laptops, and TVs.
Creating a karaoke library from a personal music collection by pointing the application at a local folder and queuing songs for automatic analysis.
Playing karaoke directly from a Jellyfin or Navidrome media server without needing to pre-process files or manage a separate song library.
Practicing vocal performance with real-time pitch feedback, guide vocal volume control, and low-latency microphone monitoring.
Open-Source Alternative Value:
Nightingale provides a self-contained karaoke system as a single binary, removing the need for manual setup of Python, ffmpeg, or machine learning frameworks. Its use of neural-network-based stem separation, automatic word-level lyric transcription, and support for self-hosted media servers offers a locally-run alternative for users who want karaoke capabilities from their existing music libraries without relying on streaming-based karaoke services. The application can also read pre-made UltraStar song folders, providing interoperability with existing community song collections.