Isolate a voice from any song or recording
Sung or spoken. Upload an MP3, a WAV or the audio of a video, and ChordSonic's pipeline takes the music out from behind the voice — leaving the words on their own, at the original length and timing.
Free · no signup · daily limit applies
Drop the recording in
MP3, WAV or the audio track of a video, up to 15 MB (25 MB with a free account). Uploads made without an account stay tied to your browser session and are cleared within a day.
The music steps back
ChordSonic's pipeline decides what each sound belongs to and lifts the voice away from the instruments behind it — not by filtering frequencies, which would take the voice with them.
Keep the voice, or the rest
The voice arrives ticked on its own. Flip the selection and the download becomes everything except the voice — same run, no second upload.
One run, two useful files
- As it stands, download gives you the voice by itself — the reason most people open this page.
- Untick the voice and tick the rest, and the same result panel builds the recording without it.
- Whatever is ticked is bounced to a single file, so you never have to choose before you hear both.
It separates music from a voice. It does not clean up a room.
The cases this handles well all have the same shape: a voice and MUSIC competing for the same file. A narrator over a bed. A podcast whose theme is still running under the first sentence. An interview recorded where something was playing. A song where you want the singer alone. In every one of those, the thing in the way is an instrument, and instruments are exactly what the pipeline knows how to pull apart.
What it will not do is denoise. Air conditioning, traffic, mains hum, mic handling, the reverberation of a hard room — none of that is an instrument, so none of it gets pulled away. It travels with the voice, because as far as the separation is concerned it IS part of the voice recording. Promising otherwise would only waste your upload.
One consequence worth knowing: the voice that comes back keeps its own effects. Reverb and delay that were printed during the mix are part of the performance and follow it out. You get the voice without the band, not a dry studio take.
Four pages, one separation — which one is yours?
Worth saying plainly: this page and the three below run the same engine on your file. They differ in which parts arrive ticked and in what the page helps you do afterwards. If your job is described better by one of these, follow it — the upload is the same.
The same job aimed at songs: you want the sung vocal of a record and nothing else. Shorter page, same result file.
For remixers. Adds the part this page skips — tempo, key and how to put the vocal over a new instrumental.
The opposite direction: keeps the music, takes the voice out. Use it when the voice is the thing you do not want.
When the voice is not the point and you want every part of the recording on its own track.
Voice isolator FAQ
Does this work on a speaking voice, or only on singing?
Both. ChordSonic's pipeline recognises a human voice rather than a melody, so a narrator over a music bed, an interview with a radio playing behind it, or a podcast intro with a theme under the talking all separate the same way a sung vocal does. Speech over music is in fact one of the easier cases, because a voice and a backing track sound very little alike.
Can it remove background noise, hiss or room tone?
No, and this is the honest limit of the tool. It splits a recording into the musical parts it was built from — voice, drums, bass and the rest. A fan, traffic, a hum from the mains or the reverb of the room are none of those things, so they stay attached to the voice and come out with it. If your problem is noise rather than music, this is the wrong tool and no setting here will change that.
What comes back if there is no music at all in the file?
Near enough the file you put in. With nothing to separate the voice from, the voice track is the recording, minus whatever the pipeline judged not to be a voice — which on a clean spoken recording is very little. That is a fine way to check a file, but it is not an improvement, so do not expect one.
Can I keep the music and drop the voice instead?
Yes, from the very same run. In the result panel untick the voice and tick everything else, then download: that mix is the recording without the talking or singing. If that inverted job is the one you actually came for, the vocal remover is the page written for it.
How clean is the isolated voice going to be?
Clean enough to understand every word in almost all cases, and clean enough to re-use in another edit in most of them. What follows the voice out is whatever was printed onto it: reverb, echo, and on a loud, heavily compressed master a faint trace of the mix. Quiet, well-recorded speech comes out best; a shouted vocal in a dense club record comes out worst.
What goes in, and what do I get back?
MP3, WAV, M4A, MP4, FLAC and AIFF go in — the audio of a video file is read straight out, so you do not have to convert first. What comes back is an MP3 of the same length, so the isolated voice still lines up with the original on a timeline.
Is it free, and is the paid version better at voices?
One file a day without an account, no card. And no, a paid plan does not isolate a voice any better: the voice gets its own track in the standard 4-part split everyone gets. The Supporter and Pro plans add a 6-part split, which puts guitar and piano on separate tracks — that is about the music behind the voice, not about the voice.
Create your own music
Generate unique new tracks of your own — built from the key, tempo and chords of any song you upload.
Wanted the music without the voice?
The vocal remover is this tool pointed the other way — it keeps the backing and takes the singing or the talking out, in one step.
Open the vocal removerWorking on a song rather than a recording? The chord finder reads the key, the tempo and the chords off the same file.