A voice isolator that runs on your computer
Most voice isolation services upload your file to a server, put it in a queue and charge per minute once a short free allowance is used. This voice isolator works differently. The AI model is downloaded to your browser once, and from then on every file is processed locally on your graphics card through WebGPU. That makes it private by design and free to use as often as you need.
Under the hood the voice isolator uses HT-Demucs v4, an open separation model from Meta AI Research. It splits the sound into four parts: voice, drums, bass and everything else. The voice part is your isolated voice; the other three together form the background. An optional second pass with DeepFilterNet, an open speech enhancement model, removes steady noise that is left on the voice track.
How to isolate a voice from audio or video
- Drop an audio file (MP3, WAV, M4A, FLAC, OGG) or a video (MP4, MOV, WebM) into the box above.
- Wait while the voice isolator works. The first run downloads the 170 MB model once; after that it starts in about a second.
- Press play. The background starts muted, so you hear the isolated voice on its own. Unmute any part to compare.
- Click "Download isolated voice" for a WAV file, or "Download background only" for everything except the voice.
What a voice isolator can remove
Background music
Music is where AI voice isolation is strongest, because the model was trained on millions of mixed songs. Music under a vlog, a radio interview or a wedding speech is pulled onto the background tracks, and the voice stays clear.
Crowd, room and street noise
Applause, chatter, traffic and wind are not voices, so most of them go to the background. Very busy scenes with other people talking are harder: a voice isolator cannot tell which voice you want, so nearby speakers stay on the voice track too.
Hum, hiss and fan noise
Steady noise sometimes stays mixed into the voice. For recordings like that, run the result through our background noise remover, which uses a model made for speech and steady noise.
Uses for voice isolation
- Video editing. Rescue dialogue from a clip with loud music, or swap the music for a licensed track without losing what people say.
- Podcasts and interviews. Clean up remote recordings that picked up music or a noisy café.
- Transcription and subtitles. Speech recognition and human transcribers both make fewer mistakes on an isolated voice.
- Music. Isolate the singer of a song to study a performance, make an acapella or sample a phrase.
- Archives and family recordings. Bring forward the voices in old tapes, home videos and voicemails.
Voice isolator tips for the cleanest result
- Start from the best source. A WAV or high-bitrate file isolates better than a heavily compressed copy downloaded from social media.
- Keep the original level. Do not normalize or boost the file before isolation; clipped peaks confuse the model.
- Check the background track. If words appear in the background download, the model was unsure. Mixing a little of the background back in often sounds more natural than a perfectly dry voice.
- Trim long files. Files up to 10 minutes are processed in one go. Split longer recordings into parts.
Voice isolator vs vocal remover vs noise remover
| Tool | Keeps | Removes | Best for |
|---|---|---|---|
| Voice isolator (this page) | Voice | Music and noise | Videos, interviews, songs |
| Background music remover | Voice and speech | Music | Videos with copyrighted music |
| Background noise remover | Speech | Steady noise, hiss, hum | Calls, voice notes, lectures |
| Vocal remover | Music | Voice | Instrumentals and karaoke |
Private, free and fast
Everything happens in your browser tab: decoding, voice isolation and writing the WAV files. With a recent Chrome or Edge on a computer with a graphics card, a three-minute file takes well under a minute. Browsers without WebGPU fall back to the CPU, which gives exactly the same result but takes several times longer. There are no accounts, no daily limits and nothing is watermarked.