New App Splits Songs Into Perfect Practice Tracks in Seconds

For decades, pulling a bass line out of a finished song was like trying to unscramble an egg. Now it takes 40 seconds.

AudioPrism, a new $80 Windows application, uses machine learning to split any recording into separate tracks for vocals, bass, drums, and other instruments. The transformation happens locally on your computer's graphics card, meaning no uploads, no waiting, and no cloud fees.

Musician Howard Armitage tested it with Fleetwood Mac's "Dreams" and watched his computer separate the entire song in the time it takes to say "Fast Fourier Transform" out loud. The isolated bass track was so clean that when he recombined all four stems and compared them to the original, he couldn't hear the difference.

The technology solves a real problem for musicians learning new songs. Want to hear just the bass part you're trying to master? AudioPrism extracts it. Ready to play along? You can hear everything except the bass, stepping into the original player's shoes with the full band backing you.

Even a deliberately difficult test with a dense live Peter Gabriel recording produced usable results. The extracted bass had some artifacts but was clear enough for practice, and a few minutes of cleanup would have made it pristine.

The Bright Side

This represents a massive shift in what's possible for home musicians. Technologies that big studios used for projects like remixing The Beatles' "Revolver" or restoring John Lennon's vocals on "Now and Then" now run on regular gaming computers in less than a minute.

The price point matters too. At $80 (currently on sale for $60), the tool costs less than a single private music lesson. Musicians can learn parts from any recording in their collection without buying expensive sheet music or relying on amateur transcriptions.

While AudioPrism has some quirks (it lacks basic mixing controls and its MIDI transcription feature underperforms), competing tools like the free, open-source StemKit are filling those gaps. The core technology works beautifully, and multiple developers are now racing to package it in ways that serve musicians' actual workflows.

The same neural networks powering this revolution are making stem separation increasingly accessible across different platforms and price points.

What seemed impossible to a 12-year-old learning bass is now routine, transforming bedroom practice sessions into collaborations with musical heroes.