Remove Background Music from Video

The music goes — speech and other sounds stay · video or audio · free · no upload

Removes background music and keeps speech and other sounds. Slower: about 2–6 minutes per minute of sound on a computer, and the model (about 54 MB) downloads when you start.

Drop a video or audio file

or

Choose file

MP4 · MOV · WEBM · MKV · MP3 · WAV · M4A · AAC · FLAC · OGG · OPUS

Music model (about 54 MB) loads when you start
No file handy?
  • No upload — files stay on your device
  • Free, no length limit
  • No sign-up, no watermark

Working on this device · nothing is uploaded

Cleaning your file…

0%Starting…
  1. Reading the sound from your file
  2. Removing background noise
  3. Saving the clean sound into your file

Keep this tab open until it finishes — closing it stops the work.

Done · cleaned on this device

Your file is clean

Download

Switch between Original and Clean while it plays — the picture keeps going, only the sound changes. Press S to switch.

0:00 / 0:00
Removed
Voice warmthVoice sounds thin? Add back some low tones.

Email me the download link

Open it on another device or save it for later. The link works for 24 hours.

Only if you ask: the finished file (never your original) is uploaded to our storage and deleted after 24 hours. Your address is used once to send the link and is not stored. Delivered by Resend.

How to remove background music from a video

01

Drop your file

Any video or audio file with speech over music. Music is already picked above the drop zone; drop the file and it starts on its own.

02

Wait while it separates

Your browser downloads the music model once (about 54 MB), then splits speech, music and other sounds on your device. This takes a few minutes per minute of sound; the progress bar shows the time left. Nothing is uploaded.

03

Listen, then download

Switch between Original and Clean while it plays, then save the same file with clean sound.

Speech, music and everything else, separated

Music mode uses Bandit v2, an open model trained to split a soundtrack into three parts: speech, music, and other sounds such as footsteps, traffic or a closing door. It leaves out the music and puts the other two back into your file. Voice only, the third button above the drop zone, keeps just the speech.

  • Speech stays as it was. In our tests, clean speech put through Music mode came back practically unchanged.
  • The music goes. With music 6 dB quieter than the voice, the music came out about 26 dB quieter; in stretches with only music, the result was close to silence.
  • Other sounds stay. With traffic behind the voice, the traffic was kept at almost its original level while the music dropped by about 17 dB.

How long it takes, and the download

Separating music is much heavier work than removing noise. The first time you use Music or Voice only, your browser downloads the model and its runtime (about 54 MB) and keeps them in its cache. After that, roughly:

DeviceTime per minute of sound
Computer with 8 or more coresabout 2¼ minutes
Computer with 4 coresabout 3 minutes
Browser that lets the page use only one coreabout 6 minutes
Phonelonger — above 5 minutes of sound we suggest a computer

Measured on an 8-core desktop (Intel Core i7-10700F) with the same model files and WebAssembly runtime this page uses. Your device may be faster or slower; once the first few seconds of sound are done, the progress bar shows the real time left.

Good to know

  • Singing counts as music. The vocals of a song are removed together with the music. This is not a tool for pulling vocals out of a song.
  • Every voice stays. Other people talking are kept, in every mode.
  • Mono sound. The model works on one channel; the result is put on both channels, so it plays normally everywhere.
  • Audio files too. MP3, WAV, M4A, FLAC and OGG work the same way and come back in the same format. In videos, the picture is copied without re-encoding.
  • Echo stays. Room echo on the voice is not removed.

Questions

Does it keep sounds that are not music?

Yes, in Music mode. Footsteps, traffic, doors, applause and similar sounds are kept together with the speech. If you want only the speech, pick Voice only.

What happens to singing in the music?

It is removed with the music: the model treats sung vocals as part of the song. It is not a tool for extracting vocals from a song.

How long does it take?

About 2–3 minutes per minute of sound on a computer with 4 to 8 cores, about 6 minutes per minute in browsers that let the page use only one core, and longer on phones. The progress bar shows the time left.

Is my video uploaded?

No. The model is downloaded to your browser (about 54 MB, once) and runs on your device. Your original never leaves it; only if you ask us to email you a download link is the finished file stored, for 24 hours.

Will my video lose quality?

The picture doesn’t: it is copied into the new file as it is, without re-encoding. The sound comes back in the same format at about the same bitrate (WAV as 16-bit). The new sound is mono (the same on both channels), so stereo effects in the original are not kept.

Next

Updated 2026-11-03