In this guide
If music is already a separate track, mute that track first. For a finished mix, target music and review Background audio to try keeping the surrounding recording; target speech and review Isolated sound when you only need the voice. Neither route guarantees undamaged speech or complete music removal.
To remove music from audio and keep the voice, first check whether you still have separate voice and music tracks. Muting the music track avoids estimating anything from a finished mix. If your only copy is a mixed MP3 or WAV, you can test source isolation: request the background music and keep the remaining audio, or request the speaker and keep the isolated voice.
Those two choices can produce different candidates. A podcast voiceover may need speech alone. An interview might need speech, a door closing, and the audience's reaction. Decide what should survive before choosing a prompt or judging the result.
Choose the method from the files you have
| Your available recording | First step | What this does not solve |
|---|---|---|
| An editing project with separate music and microphone tracks | Mute the music track and listen to the microphone recording. | Music captured by the microphone may still remain. |
| Stereo with the wanted voice already on one side | Listen to both channels and extract the useful one. | Any music mixed onto that same channel remains. |
| A finished mix; speech is all you need | Test a description of the speaking voice and inspect Isolated sound. | Other scene sounds may be excluded, and speech can change. |
| A finished mix; speech and surrounding sounds both matter | Test a description of the music and inspect Background audio. | Music may remain or wanted sounds may be removed with it. |
The free channel splitter can inspect existing left and right channels in supported short stereo MP3/WAV files. Two playback channels do not automatically mean separate voice and music recordings. If both sources are present on each side, selecting a side does not separate them.
Remove background music from an MP3 or WAV
- Keep the original and choose a useful test passage. Include a complete spoken sentence, a quiet ending, and a point where the music competes with the voice. Avoid testing only a pause with no speech: it cannot reveal damage to words. Record the excerpt's original start time if you will put it back into a longer edit.
- Open the background music removal tool. Upload MP3 or WAV audio up to 50 MiB (52.43 MB). MP4 and MOV video are also supported, with separated audio as the result. For a shorter audio test, the free trimmer supports short MP3/WAV files within its own limits.
- Describe one audible source. To reduce the music while trying to retain the scene, start with “The background music.” For speech alone, try “The narrator's spoken voice” or another description that matches what you hear. These are candidates to evaluate, not commands guaranteed to remove every instrument or preserve every word.
- Process, then compare the correct track. A music request puts the requested music in Isolated sound; review Background audio for the remainder. A voice request puts the wanted speech in Isolated sound. Listen to the other output too: wanted words there can explain why your chosen version sounds incomplete.
- Review before downloading. Check the same passages against the original at a similar listening level. If the candidate is useful, download its WAV. Downloads require an account and sufficient credits; the result screen shows the cost before downloading. See pricing for account options.
A larger WAV export is a working format, not evidence that an MP3's lost detail has returned. Keep your best available original rather than repeatedly converting or processing a previous result. The MP3 and WAV editing guide explains the source-versus-export distinction.
Which output keeps the voice?
| Example goal | Candidate prompt | Output to review | Check closely |
|---|---|---|---|
| Narration without its backing music | The narrator's spoken voice | Isolated sound | Quiet syllables, breaths, and music during speech |
| An interview scene without its music bed | The background music | Background audio | Both speakers, interruptions, applause, and room sound |
| Several people speaking over music | The people speaking | Isolated sound | Whether each wanted person survives; this is not a track per speaker |
For example, imagine a presenter talking over music followed by applause. A speech-targeted result might suit a voiceover edit but omit the applause you wanted for a scene. A music-targeted remainder is another candidate to inspect. Keep the applause interval in your review rather than choosing solely from the presenter's clearest sentence. This is an illustrative decision, not a reported separation result.
If you compare both strategies, base each request on the same original passage. Save the target description and output name with each candidate so “music removal” does not become an ambiguous file label. The source-description guide helps you change one audible detail at a time.
Listen for lost words as well as leftover music
Music can become quieter while the voice becomes less useful. Review these parts of the recording before treating a cleaner pause as a successful edit:
- A full spoken sentence: follow it without reading a transcript. Check consonants, quiet words, and the last syllable.
- The loudest overlap: replay where music and speech compete most strongly. Listen for a hollow voice, broken syllables, or musical fragments attached to the words.
- A pause and the next word: check whether background sound pumps or changes abruptly as the speaker resumes.
- Other wanted sounds: include a second speaker, laugh, impact, or audience reaction if it belongs in the final edit.
- The final placement: audition the candidate in the project at its intended level, with the corresponding original muted during a replacement comparison.
Keep a specific note, such as “the last word is incomplete at source 00:18.4,” rather than only rating how quiet the track feels. If the words are damaged, compare the alternative target strategy from the original, use an available separate microphone take, or retain the original passage. Repeated separation is not a guaranteed repair for missing speech.
Why a noise profile or a gate may not remove the music
Audacity's Noise Reduction documentation describes a noise-profile workflow for relatively constant background sounds and the risk of damaging wanted audio. A changing musical arrangement is a different problem from steady hiss. A music-only pause is not automatically a suitable profile for preserving speech throughout a song.
A gate can lower a passage between words, but during an open speech passage the mixed-in music is still part of the recording. For choosing among source isolation, profile-based noise reduction, and local edits, use the cleanup method comparison. Changing the treatment cannot recover words that were never captured clearly.
Common questions
Can I remove all the music without changing the voice?
That is not guaranteed from a finished mix. Music and speech overlap, and the separated voice can retain music or acquire artifacts. An available original microphone track is a better starting point than assuming a processed mix is equivalent to it.
Should I use a vocal remover to keep spoken dialogue?
A vocal-removal workflow usually aims to keep an instrumental after separating singing. For spoken dialogue, describe the speaking voice and review Isolated sound, or target the unwanted music and inspect Background audio. Choose by the wanted content, rather than the word “vocal” alone.
Can I do this with a mono recording?
You can test source isolation on a mixed mono recording. Mono has no separate left and right channels to extract, and it does not guarantee a clean voice. Listen to difficult speech and music overlaps before using a candidate.
Does this produce a finished video without music?
No. The isolation workflow produces WAV audio. To replace a video's soundtrack, import the chosen WAV into an editor, align it with the source, mute the corresponding original audio, and check sync. The video audio replacement guide covers full-length and excerpt placement.
Is background music removal free?
Saturalabs' AI audio downloads require an account and sufficient credits. The local preparation tools, such as the channel splitter and trimmer, have separate free workflows; using them does not make an AI download free.