5 min read
How to remove music from a video without destroying the dialogue
If the music is glued into a stereo mix, there is no clean undo. There are good workarounds, and one very good prevention.
The request sounds simple: take the music out of this video and keep the voice. Whether it is possible depends entirely on how the audio reached you.
Case 1: you still have the project
Then this is not a separation problem. Mute the music track, export the dialogue and effects stems, done. Always check whether the original edit exists before reaching for a repair tool.
Case 2: the video has multiple audio tracks
Broadcast masters and some MKV or MOV files carry several audio streams, often a full mix plus an M&E. Inspect the container before assuming there is one track. Extracting the right stream gives you a perfect result with no processing at all.
Case 3: one finished stereo mix
Here you are asking software to guess which parts of a shared spectrum belong to music. Modern source separation is impressively good on songs and much weaker on film mixes with reverb, effects and mastering compression. Expect artefacts on plosives, cymbals and anything wide.
Two things improve the result: work from the highest quality source available, and extract the audio losslessly instead of using a low bitrate copy. Every generation of MP3 or AAC compression makes separation harder.
The practical workflow
- Extract the audio from the video to uncompressed WAV, at the container's native sample rate.
- Check for multiple audio streams and for an existing M&E track.
- Only then attempt separation, and only on the best source you have.
- Rebuild the bed with your own atmospheres rather than trying to erase everything.
The audio extractor here pulls the audio out of MP4, MOV, MKV and WebM in the browser, at the true sample rate of the source, so you start from the cleanest possible material.