How to Remove Background Music from a Video and Keep the Voice

You recorded a great interview, but there’s a café playlist running underneath it. Or your vlog got a copyright claim because of a song playing in the background. Or you just want to replace the music in a clip. In all of these cases you want the same thing: remove the music, keep the voice.

Why the usual tricks don’t work

In a finished video the voice, the music, and every other sound are mixed into one audio track. That makes the obvious approaches fail:

• Lowering the volume lowers everything.
• Cutting frequencies with EQ muffles the voice and still leaves the music audible.
• Noise reduction is designed for steady noise like hiss or hum. Music changes constantly, so noise reduction either does nothing or chews up the speech.
• Re-recording or dubbing works, but it costs far more time than the clip is worth.

To remove music cleanly you have to separate the soundtrack into its parts first, then rebuild it without the part you don’t want.

How dialogue–music–effects separation works

Film and TV studios have always kept dialogue, music, and effects on separate “stems” so a film can be dubbed into other languages. Researchers used that structure to build the Divide and Remaster (DnR) dataset — thousands of soundtracks where all three layers are known — and trained models to pull a mixed soundtrack back apart.

Our video music remover uses DBS-TIGER, a model trained on DnR. It returns three layers:

• Dialogue — speech from everyone on screen.
• Background music — score, licensed tracks, and singing.
• Other sounds — footsteps, traffic, applause, room tone.

Keeping “other sounds” is what makes the result feel natural. A video with only dialogue can sound oddly dead, as if it were recorded in a booth.

Step by step

1. Open remove music from video but keep voice and sign in.
2. Upload your video — MP4, MOV, WEBM, or M4V, up to 90 MB.
3. Under Keep in the finished video, leave Dialogue and Other sounds checked and uncheck Background music.
4. Start processing. Processing time grows with the length of the video.
5. Download the finished MP4 — or grab the individual WAV layers if you want to fine-tune the mix in your own editor.

Your picture is copied into the new file without re-encoding, so it looks exactly like the original. Only the audio track changes.

Getting the best result

• Clear speech separates best. Interviews, talking-head videos, tutorials, and narration work very well. Several people talking over each other, shouting, or a loud crowd are harder.
• Singing counts as music. If someone on screen sings, their voice will usually land in the music layer.
• Check stereo-heavy edits. The model works on a mono version of the soundtrack, so the finished audio is centered. That’s invisible for most dialogue videos but noticeable on cinematic edits with wide stereo effects.
• Keep the music layer. It’s included with every result, so you can bring a little of it back later if the scene feels empty.

Will this get rid of a copyright claim?

Sometimes — but don’t count on it. Content ID and similar systems mostly match audio, so removing the matched song often clears the claim. But some platforms also match the picture, and removing the music doesn’t change who owns the rest of the footage. Use music removal to fix your own recordings and to swap in music you’ve licensed, not to repost other people’s content.

---

Try it on your own clip: remove background music from a video.