In this guide
First check whether the voice is already on a separate track or channel. For steady hiss, try a noise profile; for speech overlapping music or changing sounds, compare a voice-isolation result and listen for missing words.
To isolate a spoken voice in Audacity, first check for a separate microphone track or a cleaner left or right channel. If the voice and a steady noise are mixed together, use a noise profile to reduce that noise. When speech overlaps music, traffic or other changing sounds, noise reduction alone may not separate them: test a voice-isolation tool and compare its result with the original.
These are different problems. A stereo channel is part of the recording as captured; an isolated voice produced from a finished mix is an estimate. Keep the original so you can recover any words that processing changes.
Choose the method for your recording
- Separate microphone tracks available: mute the other tracks and work from the desired microphone. Check for bleed from other voices.
- Voice louder on one stereo channel: split the channels, listen to each, and keep the more useful one if it contains the whole passage.
- Steady fan or hiss under speech: capture a noise profile from a pause and apply a restrained reduction.
- Speech overlaps music or moving background sound: try voice-focused separation, then inspect the words and remaining background. Results vary with the source.
1. Check tracks and stereo channels
Open a copy of the audio in Audacity. If the project has individual microphone tracks, use each track's Solo button to find the voice you need. Listen through the entire passage: another speaker may still be audible through that microphone.
If you have one stereo track, open its track dropdown and choose Split Stereo to Mono. Solo each resulting track in turn. If one has a much clearer voice, keep that channel and mute or remove the other before export. A mono track plays through both speakers. If both channels contain the same mixture, splitting them will not magically separate the voice from the background; mixing them to mono will combine their contents instead.
For example, an interview recorded with one microphone on each channel can often be cleaned up by choosing the relevant channel. A phone recording with traffic and speech in both channels needs a different approach. The Audacity manual explains channel splitting.
2. Reduce consistent noise in Audacity
- Find a short pause containing the same steady background noise, without words, breaths or music tails. Select it.
- Open Effect → Noise Reduction and Repair → Noise Reduction and choose Get Noise Profile. Menu placement can vary by version.
- Select the dialogue to process, reopen Noise Reduction and preview a modest setting. Compare it with the original, especially quiet syllables and word endings.
- Use the effect's residue option to hear what is being removed. If you hear recognizable speech, back off or undo and adjust the settings.
Noise Reduction works best on consistent noise and needs a representative noise-only sample. A voice and music playing together do not provide one; aggressive settings can make dialogue watery or distorted. See our Audacity noise reduction guide for a closer look at the controls, or consult the Audacity manual.
3. Try voice isolation when sounds overlap
If music or changing noise competes with the words, try a voice-focused separator on a short, difficult excerpt first. With Saturalabs Voice Isolator, upload your audio or video, process it for voice, and compare the result with the original. Downloads require credits. Listen to soft words, consonants, pauses and transitions as well as the loudest sentence.
If the background remains obvious or words sound clipped, compare another source recording or adjust your editing approach. Separation cannot guarantee a clean voice or reconstruct words that were never captured clearly. It may also retain multiple speakers: isolating speech from background sound is not the same as separating two overlapping voices. For that problem, see how to separate two voices in one recording.
For singing over an instrumental, use the separate Audacity vocal-removal guide. Music-separation models and older center-channel effects solve a different problem from spoken dialogue and should not be assumed to isolate speech cleanly.
Export and check the result
Before exporting from Audacity, mute or remove any unwanted original track so it does not play over your cleaned version. Export a WAV working copy, reopen it and listen from beginning to end. If the dialogue is going back into a video, preserve the original timing and check sync at both the start and end. Keep the source and project until the final edit is approved.
Can Audacity isolate one person's voice?
It can select a person's separate track or channel, and editing can remove another person's speech during non-overlapping turns. When people speak at the same time in one mixed recording, Audacity's standard noise reduction is not a reliable speaker separator. A specialist tool may help, but test an excerpt and listen for missing or blended words.