Tool 24AudioDemucs + DeepFilterNet

Vocal remover. Vocals, instrumental and clean voice

Pull the voice out of a song for karaoke or remixing, or clean the hum and hiss out of a voice recording. Listen to every part before you download.

Drop a song

MP3, WAV, M4A, FLAC or a video · up to 4 minutes

VocalsInstrumentalMP3 download
01

How it works

  1. 1

    Pick what you need

    Remove vocals from a song, or clean background noise from a voice.

  2. 2

    Add the file

    Audio or video. The length and price show before you start.

  3. 3

    Listen and download

    Play each track with its waveform, compare, and save the MP3s.

02

Studio-grade separation

Open models used by audio engineers.

Demucs v4

Meta's Hybrid Transformer Demucs, a top open model for splitting songs.

DeepFilterNet 3

Real-time speech enhancement that keeps the voice natural.

Hear it first

Waveform players, and an A/B switch for the cleaned voice.

Video in, audio out

Drop a video and the soundtrack is used.

03

Questions

Not answered here? Ask us →

How does the vocal remover work?

It uses Demucs (Hybrid Transformer), a music source separation model from Meta, to split a song into the voice and everything else. You get two MP3s: vocals only and instrumental (karaoke).

Can it remove background noise from a voice recording?

Yes. Switch to Clean up voice: DeepFilterNet removes hum, hiss, fans, traffic and room noise from speech, and you can switch between the original and the cleaned take while it plays.

Which files can I upload?

MP3, WAV, M4A, AAC, FLAC, OGG and OPUS, or a video (MP4, MOV, WebM), in which case the soundtrack is used. Up to 4 minutes without an account, 15 minutes signed in.

Can I use the instrumental for karaoke or covers?

For practice and private use, yes. Publishing a separated track of somebody else's song still needs the rights holder's permission.

How much does it cost?

Three free runs without an account, then credits per minute of audio.