Everything you need to know about Beltr.
Beltr is an AI-powered karaoke app for your desktop. It uses neural network vocal separation to turn any song you own into a karaoke track — vocals removed, lyrics on screen, phones as wireless remotes.
Drop any audio file into Beltr. The AI separates vocals from instruments in about 30 seconds. Synced lyrics appear on your TV or monitor. Guests connect their phones by scanning a QR code to queue songs and control playback.
8 GB RAM minimum (16 GB recommended), modern multi-core CPU, ~2 GB disk space for the AI model. Runs on macOS 12+, Windows 10+, and Linux.
Yes. After the AI model downloads on first run, everything runs locally. No internet connection required.
MP3, FLAC, and WAV files.
30 seconds to 2 minutes per song, depending on your hardware. The first song takes a bit longer while the AI model downloads.
Demucs v4 is an open-source neural network from Meta Research that separates audio into 4 stems: vocals, drums, bass, and melody. It's the engine behind Beltr's vocal separation.
Beltr works with audio files you own. It does not stream from or connect to any streaming service.
Studio recordings produce the best results. Live recordings, tracks with heavy effects, or low-bitrate files may produce lower quality separation.
Guests scan a QR code displayed on the TV screen. It opens a web page on their phone — no app download, no account required.
No hard limit. All connected phones can queue songs, browse the library, and vote on what plays next.
No. The remote works in any modern mobile browser. Nothing to download or sign up for.
$19.99 one-time purchase (launch price, normally $29.99). No subscription.
The full app with 5 song separations. No credit card required, no time limit.
The app still works — you can play any songs you've already processed. It won't process new songs until you buy a license.
Yes. Contact support@beltr.app within 14 days of purchase.
Each license key works on up to 2 devices.