How to isolate a voice in Audacity

"Isolate voice" means two different jobs. Pulling a singer out of a finished song is music separation. Pulling a speaker away from background noise, music or a noisy room is speech isolation. Audacity's built-in tools handle the first only in specific conditions and the second only partly, so it helps to know which you are doing.

Select the track, then Effect, Special, Vocal Reduction and Isolation, and set the action to Isolate Vocals. It works by keeping what is identical in the left and right channels, so it only works on stereo files where the voice is mixed dead centre. On a mono recording, or speech recorded on one mic, it does very little.

For a voice recording with noise behind it, stack three steps: a high-pass filter at 80 Hz and any hum removal, then Noise Reduction with a profile from a silent section, then a noise gate to quiet the gaps between sentences. This works on steady noise. It won't remove other people talking, music, or traffic that changes.

Intel's free OpenVINO plugins add AI music separation and noise suppression to Audacity. Music separation can split a song into vocals and instruments far better than the built-in effect. Installation is manual and support depends on your operating system and version, so check the plugin page before you start.

If the noise changes (cafe chatter, traffic, music in the room) or you just want it done, a speech-trained tool is faster than stacking filters. Hilite's voice isolation separates speech from background sound in one step, and echo removal handles room sound.