If the song has heavy stereo reverb on the voice (common in shoegaze or 80s ballads), you are doomed. The reverb is spread to the sides, so when you cancel the center, you lose the voice but keep the echo. You end up with a ghost singing from a well.

Select your track → Effect > Special > Vocal Reduction and Isolation... → Choose "Remove Center" (or "Isolate Center" for the opposite effect).

For decades, this was impossible. A finished stereo mix was considered a "brick wall"—you couldn't pull the bricks out without breaking the wall.

But here’s the secret they don’t tell you in YouTube tutorials: The real art is in compromise . Let’s dive into the two main spells in Audacity’s grimoire, their strange side effects, and how to turn a messy extraction into something usable. Spell #1: The "Center Channel Cancel" (Vocal Reduction) This is the oldest trick in the book. It’s fast, free, and almost magical.

This creates the infamous "underwater" sound. The vocals become thin, phasey, and lose all low-end warmth. Why? Because drums are also center-panned. You’ve just made a trade: vocals for fidelity. Spell #2: The "Deep Learning" (OpenVINO) This is the modern, slightly terrifying approach. Audacity now supports AI-powered plugins (like OpenVINO or using external tools like UVR). This doesn't rely on stereo trickery. Instead, a neural network has been trained on thousands of songs to "learn" what a human voice sounds like vs. a guitar vs. a drum.