Nightingale

Transform your music library into a complete karaoke experience with AI-powered vocal separation and real-time pitch scoring. This innovative software uses neural networks to isolate vocals from any song, creating instant karaoke tracks from your existing music collection.

Key features include:

• Intelligent stem separation using UVR Karaoke model or Demucs to isolate vocals from instrumentals with adjustable guide vocal volume
• Word-level lyrics transcription via WhisperX that aligns every word to the audio timeline, with fallback to LRCLIB database
• Real-time pitch scoring that rates your singing performance with star ratings and maintains per-song scoreboards
• Multiple player profiles to track separate score histories for different singers
• Video file support for .mp4 and .mkv files with automatic vocal separation while displaying original video as background
• Dynamic visual backgrounds including GPU shader effects (plasma, aurora, nebula), Pixabay video loops, or source video content
• Full gamepad support for menu navigation, song selection, and playback control using D-pad, sticks, and face buttons

The software ships as a single self-contained binary that automatically bootstraps ffmpeg, Python, PyTorch, and ML models on first launch - no complex installation required. Works across with GPU acceleration via CUDA or Metal when available, plus CPU fallback for universal compatibility.