Blog / stem-splitter
How to Remove Background Music from a Video
· 21 min read
The 30-second answer: extract the audio from your video, run it through an AI stem splitter, keep the voice track, and lay it back over your footage. Two steps, under two minutes, no plugins and no paid software.
You're probably here for one of three reasons: your footage has music you can't clear, you're cleaning up an interview shot in a noisy café, or a client sent a video with a soundtrack that has to go. All three are the same technical problem — the voice and the music are mixed into one audio track — and in 2026, it's finally a solvable one. Here are the methods that work, ranked by output quality.
Why video audio is harder than music files
Removing music from a video feels like it should be an editing problem, but it's a separation problem — the same one behind vocal removal from songs. When footage is recorded, everything the microphone hears is summed into a single track: the speaker, the background music, the room. No slider in any video editor can "turn down the music" that's baked into the same signal. What modern AI tools do instead is reconstruct the two sources — voice and music — as separate stems, using models trained to recognize what each sounds like.
That's also why the order of operations matters: you separate first, then edit. Trying to fix music under a voice with EQ or noise reduction after the fact just degrades the voice along with it.
Method 1: AI stem separation (best quality — use this)
The workflow works on any video file — MP4, MOV, whatever your camera or phone produces — because the separation runs on the audio, and every editor can hand you the audio.
How to do it:
- Export the audio from your video. In Premiere, right-click the clip and choose Audio Gain → detach; in DaVinci Resolve, right-click → Detach Audio; in CapCut and most mobile editors it's a one-tap "extract audio" button. No editor at hand? A free converter can pull the MP4's audio track to MP3 or WAV in seconds.
- Run the audio through a stem splitter. Upload the extracted file, wait a few seconds, and the AI returns separate stems: voice, music, bass, drums. Free tier: 3 songs a day, no sign-up, and the upload is deleted right after processing.
- Lay the voice stem back into your video. In your editor, mute the original audio track and drop the voice-only stem in its place. The timing is identical — same length, same start — so it aligns with the footage automatically.
Output quality: on typical footage — vlogs, interviews, screen recordings with background tracks — the voice comes out sounding like it was recorded without the music. Expect artifacts when the music is loud, heavily compressed, or the voice and music sit in the same frequency range (a deep male voice under a bass-heavy track is the hard case).
Cost and limits: free for 3 videos per day, files up to 20MB of audio (about 20 minutes of MP3), no watermark. Works with music and with voice isolation for sung vocal content.
Method 2: Editor noise reduction (already in your software)
Every serious editor ships some form of noise reduction: Audacity's Noise Reduction, Premiere's DeNoise, Resolve's built-in voice isolation. These are designed for steady background noise — hum, hiss, traffic — not music.
How to do it: capture a noise print from a music-only moment in your clip, then apply reduction across the track.
Output quality: honest answer — poor for music. Noise reduction works by removing constant frequencies; music changes every second, so the tool either barely touches it or takes the voice with it. The exception: very quiet background music can be nudged down a few dB usefully, buying you a ducked bed under a voiceover rather than a removal.
Best for: footage where the music is subtle ambience rather than a real track, or as a light cleanup after stem separation.
Method 3: Re-record or replace the audio (the guaranteed fix)
Sometimes separation isn't the right battle. If the voice on your footage matters professionally — a paid course, a client deliverable — the cleanest result comes from re-recording the voiceover in a quiet room, or re-shooting. And if the music is the problem only because of rights, replacing it with a licensed or royalty-free track of a similar mood is faster than any cleanup.
Best for: content you control the production of, and situations where the original recording is replaceable.
Head-to-head comparison
| AI stem separation | Editor noise reduction | Re-record / replace | |
|---|---|---|---|
| Voice quality | ★★★★☆ | ★★☆☆☆ | ★★★★★ |
| Removes real music | ✅ | ❌ (steady noise only) | n/a |
| Keeps original voice | ✅ | mostly | ❌ (new take) |
| Time per video | ~2 minutes | minutes | hours |
| Cost | free (3/day) | free–included | studio time |
| Works on finished videos | ✅ | ✅ | ❌ |
Which method should you use?
- You have a finished video with music over a voice → AI stem separation. It's the only method that removes real music without touching the recording itself.
- The music is quiet ambience → try noise reduction first; it may be enough.
- The voice doesn't matter or can be re-recorded → replace the audio. Separation can't beat a clean take.
- You want to keep the music and lose the voice → same stem splitter, opposite output: keep the music stem for an instrumental version of your own footage.
- You need more than music removal → a free tool for every audio job, grouped by what you're trying to do, with the free-tier limits stated instead of buried.
Frequently asked questions
Can I remove background music from an MP4 directly? Separation tools process audio files, so the workflow is two steps: extract the audio from the MP4 first, then run it through the stem splitter. Most editors detach audio in one click, and the whole round trip takes under two minutes.
Will removing the music damage the voice quality? AI stem separation reconstructs the voice rather than filtering it out, so on typical footage the voice stays natural. Expect more artifacts when the music is loud, heavily compressed, or the voice and music share the same frequency range.
What if I want to keep the music and remove the voice instead? Same tool, opposite output: the stem splitter produces both a voice track and a music-only track. Keep the music stem and discard the voice — useful for cleaning interviews or creating instrumental versions of your own footage.
Is removing background music from a video legal? For your own footage — absolutely. For copyrighted material, removing the music doesn't remove the copyright: publishing or monetizing the result still requires rights to the music. Use royalty-free tracks or music you licensed.
Does this work on videos with live background noise instead of music? Partially. Stem separation is trained to split music from voice, not traffic from voice. Constant ambient noise is better handled by your editor's noise reduction (the "capture noise print" feature in Audacity or Premiere).