Beltr is an AI-powered desktop karaoke application that transforms any personal music collection into a professional karaoke experience. Using Demucs v4 neural network technology from Meta Research, the software separates vocals, drums, bass, and melody from any MP3, FLAC, or WAV file in 30 seconds to 2 minutes. The system displays synced lyrics on a TV screen while allowing guests to control the experience from their phones via QR code connection—no app installation required. Key features include real-time pitch detection for live singing scores, customizable stem controls, and offline operation with a one-time purchase model. The application supports macOS, Windows, and Linux, with system requirements of 8GB RAM minimum (16GB recommended) and approximately 2GB disk space for the AI model. Future updates include native mobile remote apps, duet mode, and party statistics—all included with the lifetime license.
