Nightingale is a free app that lets you turn any music file into a karaoke track. It separates the vocals from the music, generates synchronized lyrics word by word, and allows for highlighting. It also supports videos.
Nightingale is an app that uses neural networks to generate karaoke tracks from your music library. It features lead vocal and instrumental separation, automatic lyric generation, and a scoring function, allowing you to enjoy your existing music files as high-performance karaoke tracks. It is developed as open source under the GPL-3.0 license and is available for free on Windows, macOS, and Linux.
Nightingale — Karaoke from your music library
Nightingale uses the Karaoke model and Democs of Ultimate Vocal Remover (UVR) to separate vocals, and WhisperX can automatically transcribe lyrics with word-by-word timestamps.
The instrumental tracks retain their choral components, allowing for natural-sounding vocals. It also features numerous practical functions, including real-time pitch detection for score display, score management per profile, gamepad control, and an adaptive UI supporting 4K resolution.
Furthermore, Nightingale not only processes the audio source but also handles the rendering and control of the background video during karaoke. The background can feature animations powered by GPU shaders, videos from Pixabay, or even the original video file, allowing for a wide variety of visual effects.
Furthermore, if the user loads a video file they have prepared, it is possible to extract only the audio from that video, perform vocal separation analysis, and then play the original video in sync as a background while singing.
Nightingale is distributed as a standalone executable file and is compatible with macOS, Linux, and Windows platforms. Setup is simplified, and necessary components such as Python, FFmpeg, and machine learning models are automatically downloaded and built upon first launch.
File analysis can be performed on a folder-by-folder basis. Analysis results are cached, so subsequent playback is faster. The analysis process supports CUDA acceleration using NVIDIA GPUs and CoreML on Apple Silicon, resulting in efficient processing.
Nightingale's recommended hardware environment is a system with an NVIDIA GPU or a macOS environment with Apple Silicon (M1/M2/M3 chips, etc.). It can run on a general PC (CPU only) without a specific GPU, but the processing speed will be significantly reduced. The analysis time per song is about 2 to 5 minutes with the recommended hardware environment, and about 10 to 20 minutes if processing is done with CPU only.
Nightingale is developed as open source under the GPL-3.0 license, and its source code is available on GitHub.
rzru/nightingale: Machine learning powered Karaoke app (with scores!)
https://github.com/rzru/nightingale
Related Posts:
