Your audio workspace
AUDIO WORKSPACE

Voice Remover

Separate vocals and accompaniment from a song.

Private upload
01Upload a song or recordingMP3 · WAV · FLAC · M4A · AAC · OGG
Preview your separated tracks before downloading
Sign in to process this file and keep your results in My files.
01 Upload →02 Preview →03 Download
ONLINE AUDIO TOOL · VOCREMOVER

Remove the voice from any song.

Upload a song and VocRemover splits it into two files: the isolated vocal and the instrumental underneath it. Use the instrumental as a karaoke or practice track, or keep the acapella for a remix, cover, or transcription.

Upload a file
Preview before download Private upload
HOW IT WORKS

From upload to separated tracks.

01

Add your file

Upload an MP3, WAV, FLAC, M4A, AAC, or OGG file up to 80 MB.

02

Separate the sound

The song is split into an isolated vocal and an instrumental.

03

Preview and download

Listen to both tracks in your browser, then download the ones you need.

How the voice remover works

VocRemover runs your file through htdemucs_ft, the fine-tuned Hybrid Transformer version of Demucs — an open-source music source separation model originally published by Meta AI researchers. Demucs looks at the waveform and the spectrogram of the song at the same time, which helps it tell a voice apart from instruments that share the same frequencies.

The model returns two stems: “vocals” and everything else (“no vocals”). Unlike a classic center-channel phase trick, it does not depend on the voice being panned to the middle, so it also works on live recordings, mono files, and songs with reverb-heavy mixes.

What people use it for

  • Karaoke and sing-along tracks from songs that have no official instrumental.
  • Practice tracks for singers, guitarists, and drummers who want to play along with the band.
  • Acapellas for remixes, mashups, and DJ edits.
  • Transcribing lyrics or melodies when the instruments make the vocal hard to hear.
  • Cleaning up a vocal demo recorded over a backing track.

Tips for the cleanest result

  • Start from the highest-quality source you have. A WAV or FLAC file separates more cleanly than a low-bitrate MP3, because compression artifacts get split along with the music.
  • Upload the full mix, not a file that has already been processed by another vocal remover — artifacts stack up.
  • If you only need part of the song, trim it before uploading; shorter files process faster.
  • Listen to both stems before downloading. The vocal stem is often more useful than people expect, even when they came for the instrumental.

Honest limitations

No separation model is perfect. Heavy reverb and delay on a vocal can leave a faint “ghost” of the voice in the instrumental. Backing vocals and harmonies are treated as vocals, so they are removed together with the lead. Instruments that sound very voice-like — a lead synth, a saxophone, a vocoder — can partly bleed into the vocal stem.

For most pop, rock, hip-hop, and acoustic songs the result is good enough for karaoke and practice straight away. For release-quality remixes, expect to do some light editing on top.

FAQ

Frequently asked questions

Which audio files can I upload?

MP3, WAV, FLAC, M4A, AAC, or OGG files up to 80 MB — enough for a full-length song in any of those formats.

What do I get back?

Two tracks: the isolated vocal and the instrumental. You can play both in the workspace before downloading either one.

Do I need to install anything?

No. Processing runs on cloud GPUs, so it works the same in any modern browser on desktop or mobile. You need a free account so your results are saved to your private file list.

Can it remove only the backing vocals and keep the lead?

Not yet. The model treats every voice in the song as one vocal stem, so lead and backing vocals are removed or kept together.

Can I use separated tracks from songs I do not own?

Separating a track does not change who owns it. Removing vocals for personal practice is different from publishing the result. Only upload files you have the right to process, and get permission before you publish or monetize anything based on someone else’s recording.

Try VocRemover →