Media

Video Audio Editor

Mute a video, swap its soundtrack, put music under it, or make a quiet recording louder. The picture is copied untouched in every mode, so only the sound is ever encoded and that is what keeps this quick. FFmpeg runs in your tab; nothing is uploaded.

  • Copies the video track packet for packet, so the picture that comes out is the picture that went in
  • Draws how an added track lines up with the video, and plays the balance back before you render

Drop your video here

It stays on your machine and is never uploaded

Reads MP4, MOV, MKV, WebM, AVI, WMV, FLV, MPEG-TS, MPG, 3GP, and most other formats FFmpeg understands.

Take the sound out — all of it, or one stretch of it. Removing the whole track decodes nothing at all, so it is the fastest thing here.

No video yet

Add a video file to begin. The engine is only downloaded once you do.

Audio

Overview

Changing the sound of a video is a decision about one of the two streams inside it. The other one has no reason to be decoded, and here it never is — which is what makes every mode below quick, and what keeps the picture exactly as you shot it.

  1. 01

    The picture is copied, never re-encoded

    Whatever you do to the sound, the video track is moved across packet for packet. It is not decoded, not scaled, and not compressed again, so a 4K recording comes out as the same 4K recording rather than a generation further from it. Sites that re-encode the whole file to mute it are charging a long wait and a visible quality loss for a job that touches one stream.

  2. 02

    Four jobs that people actually come here for

    Take the sound out, in full or over one stretch. Swap the soundtrack for a file of your own. Lay music under what is already there, with both levels under your control. Or raise a recording that came out too quiet. Each is a separate mode with only the settings that mode needs, rather than one panel of everything.

  3. 03

    A timeline that answers the question before you render

    Add a track and it is drawn against the video: where it starts, how far it reaches, how much of the end is left without it, and how much of the track will not fit. Music being fifteen seconds short of the video is the sort of thing people usually discover after downloading, and there is no reason it cannot be shown beforehand.

  4. 04

    You can hear it before you commit to it

    Press play and the browser mixes the video and the added track live, at the levels, fades, and starting offset you have set. Choosing how loud a music bed should sit under a voice is a listening decision, and guessing a number then waiting for a render to check it is the slow way to make one.

  5. 05

    Sometimes the sound is not touched either

    Drop a ready-made track in with no level change, no fade, and no offset, and if the container accepts that codec the audio is copied as well. Both streams move across untouched, the whole thing becomes a file rewrite, and the picture and sound in the result are the ones the two files arrived with. The plan says so before you press anything.

  6. 06

    Local processing, with no upload

    FFmpeg is compiled to WebAssembly and runs in the tab. Nothing is uploaded, queued, watermarked, or tied to an account, and no file is limited to a preview length. Interview footage, client recordings, and anything under an agreement stay on your machine, which you can confirm from the network panel while a render runs.

How to use

From a recording with the wrong sound to one with the right sound, in four steps.

  1. 01

    Add your video. The engine starts downloading at this point; it is about 31 MB and only happens once. The file is then read for its length, codecs, and whether it has an audio track at all.

  2. 02

    Choose what to do: mute it, replace the soundtrack, add music under the original, or change the level. Replace and Add music ask for a second file, which can be an MP3, M4A, WAV, FLAC, or most other audio formats.

  3. 03

    Set the levels, fades, and where the added track starts, then press play on the video and listen. The timeline underneath shows how far the added sound reaches and what is trimmed or left silent.

  4. 04

    Press Apply. The finished file is checked against the length of the video you started with, then appears on the right, ready to play before you download it.

Details

The details that decide whether the file you get back is the one you needed.

  • The video track copied packet for packet in every mode, so the picture is never decoded, scaled, or compressed again
  • Four modes: remove the sound, replace it, add a second track under it, or change how loud the original is
  • Silencing one stretch of a video rather than all of it, set by dragging a range across the length of the file
  • A live preview that mixes the video and the added track at the levels, fades, and offset you have chosen
  • Levels in both percent and decibels, from silence to a boost of about eight times, on a scale that matches how loudness is heard
  • A measurement of the audio that is already there — its loudest point and its average — and a one-press setting that lifts the peak to just under the ceiling
  • A timeline showing where the added track starts, how much of the video it covers, what is trimmed, and what is left silent
  • Looping for a track shorter than the video, with the number of passes shown, or plain silence for the remainder
  • Fades in and out on whichever track is being written, capped so a fade can never be longer than half of what it fades
  • Peak control for mixes and boosts, which holds the total under full scale and is inaudible when nothing reaches it
  • Output in the container the video arrived in, with the sound format chosen from what that container will actually accept
  • A length check on the finished file against the video that went into it, so a truncated picture is reported rather than handed over
  • HEVC, AV1, VP9, ProRes and anything else FFmpeg can demux, since none of them are ever decoded
  • Everything runs in the browser tab — no upload, no queue, no account, no watermark, and no preview-length cap

Use cases

The situations that send people looking for a way to change a video's sound.

  1. A recording with unusable background noise

    A clip filmed near traffic, in a windy street, or in a room with an air conditioner often has picture worth keeping and sound worth losing. Removing the track leaves the footage untouched and gives you something to put a voiceover or a music bed under, either here or in an editor.

  2. Music you do not have the rights to post

    A video shot where a radio or a playlist was on can be taken down or muted automatically by the platform. Taking the sound out yourself and adding a track you are allowed to use puts the decision back in your hands, and the picture is not re-encoded on the way.

  3. A voiceover recorded separately

    Recording narration on a phone or a proper microphone rather than through the camera gives far better sound. Replace mode drops it onto the video, and the offset control lines the start up with the picture without needing an editing timeline.

  4. A music bed under an existing voice

    Adding a soft track under a talking-head or product video is the most common reason to mix rather than replace. Both levels are yours to set, the preview lets you hear the balance rather than guess it, and peak control keeps the two from adding up into distortion.

  5. A recording that came out far too quiet

    A camera microphone at a low setting, or a lavalier that slipped, gives a file that is unusable at normal volume. Measure the level, read how far below the ceiling it actually is, and raise it by exactly that much rather than by a number picked from the air.

  6. One stretch that has to come out

    A phone ringing, a name that should not be in the file, or a stray comment during a recording only affects a few seconds. Silencing that stretch leaves the rest of the sound at its original level and the picture untouched, which is the whole edit in one pass.

  7. A screen recording with no sound at all

    Capture tools frequently export without a microphone attached. Adding narration or music to a silent recording works the same way as replacing a track, and the tool says plainly that there is nothing to mix with rather than quietly producing an odd file.

  8. Footage that has to stay on your machine

    Interviews, medical recordings, legal material, and anything under a client agreement have no business on someone else's server for the sake of removing an audio track. Everything here happens in the tab, and the network panel will show you that nothing left it.

See also

This tool writes videos, so pulling a track out of one and saving it as an MP3 or a WAV — or preparing a soundtrack taken from another recording — belongs to Audio Converter. Sound is changed across the whole file here and nothing is ever shortened, so cutting the video down to the part you want is a separate step and a job for Video Trimmer. And because the result stays in the container it arrived in, turning it into a different format afterwards is what Video Converter does.

Why the sound can change without the picture paying for it

Most of what looks surprising about this tool follows from one fact about how a video file is put together, and the rest follows from how loudness is measured.

  1. A video file holds two separate streams

    MP4, MOV, MKV and the rest are containers: they hold a compressed video stream and a compressed audio stream side by side, each with its own timestamps, and neither knows anything about the other. Replacing one of them is a matter of writing a different stream into the same wrapper, not of rebuilding the file.

  2. Which is why the picture is never decoded here

    Every other video job — resizing, converting, compressing — eventually has to decode frames and encode them again, and that is where both the waiting and the quality loss come from. Nothing about the sound requires it. So the video packets are read and written straight back out, and the picture in the result is the same data it was before.

  3. And why HEVC and AV1 are not a problem here

    Elsewhere on this site those two codecs come with warnings, because this build cannot write HEVC and cannot decode AV1 without hardware a browser tab does not have. Neither matters when the stream is only being copied: the packets move without anything looking inside them. An iPhone HEVC recording and an AV1 download are ordinary inputs here.

  4. Removing a track and silencing a stretch are different jobs

    Dropping the audio entirely means simply not writing it, so nothing is decoded on either side and the operation is a file copy. Silencing ten seconds in the middle means the track has to exist and be different, so it is decoded, changed, and encoded once. Both leave the picture alone; only one of them is free.

  5. The container decides what the new sound may be

    A muxer accepts a fixed set of codecs. MP4 takes AAC and MP3 among others, so an MP3 you drop in is copied straight into it; WebM takes neither, so a track going in there is encoded to Vorbis. That is why the output stays in the format you started with — it is the one wrapper guaranteed to accept the video stream as it stands.

  6. Two sounds added together can go over the ceiling

    Digital audio has a hard maximum, written as 0 dB. Mixing music under speech adds the two waveforms, and where their peaks land together the total can pass that maximum. Everything past it is flattened, which is heard as crackle. Holding the peaks just under the ceiling prevents it and does nothing at all when nothing reaches it.

Best practices

Habits that get one good file on the first run.

  • Play the preview before pressing Apply — the balance between a voice and a music bed is a listening decision, and hearing it costs seconds while rendering to check costs minutes
  • Start a music bed at about a quarter of the original's level, then adjust by ear rather than by arithmetic
  • Lower the original a little instead of raising the music when a bed is fighting a voice, since the total stays further from the ceiling that way
  • Measure before boosting a quiet recording, because the measurement tells you exactly how much room there is and a guess does not
  • Leave peak control on whenever two sounds are playing together, since it does nothing at all until something actually reaches the ceiling
  • Give an added track a second or two of fade at each end, which is the difference most people hear between an edit and a cut
  • Use Loop rather than a longer file when a short piece of music has to cover a long video, and check the timeline for where the last pass is cut
  • Trim the video first if you only want part of it, since this tool changes the sound over the whole file and does not shorten anything

Limitations

What this tool deliberately does not do, and where it will struggle.

  • It changes the sound and nothing else. The picture is never resized, converted, compressed, or trimmed, and the file always comes back the same length it went in.
  • It cannot separate one sound from another. Removing traffic noise from speech, taking a voice out of music, or cleaning up a hum are all separation problems, and nothing here attempts them — the choice is the whole track or a stretch of it.
  • There is one added track at a time. Two pieces of music one after another, or narration plus a bed plus an effect, need an editor.
  • One audio track goes in and one comes out. A file carrying several language tracks keeps only the first, and subtitles and chapters are not carried over either.
  • A silenced stretch begins and ends at the moment you set, with no ramp on either side. On a loud passage that edge can be audible as a click.
  • Encoded sound is lossy. AAC, MP3, Vorbis and the rest discard detail by design, so a track that is decoded and encoded here is a generation further from its source, however high the bitrate. Only the copy path avoids that, and the plan says which one you are on.
  • Peak control holds the total under the ceiling; it does not undo distortion that was already recorded into the file. Sound that clipped in the camera stays clipped.
  • The output stays in the container the video arrived in, because that is the wrapper certain to accept the video stream unchanged. Changing format is a separate job with a separate tool.
  • The added track has to be an audio file. Pulling the sound out of a second video first is a separate step, linked below.
  • The preview is the browser playing two files in step through a gain graph. The levels, fades, and offset are the ones the render will use, but it is a live mix rather than the finished file, and a slow machine can let the two drift briefly before they are pulled back.
  • Measuring the level reads the whole audio track, so it takes a moment on a long recording. It reports the loudest single point and the average, which is not the same as a broadcast loudness measurement.
  • The engine is about 31 MB on first use. It is cached afterwards, but the first run in a fresh browser profile carries that download.
  • A file the browser itself cannot decode cannot be previewed or listened to here, going in or coming out. Which files those are depends on the browser: AVI and WMV are refused everywhere, HEVC plays in Safari and usually not elsewhere, and MKV is accepted by some browsers and not others. The engine can still process many of these files; download the result and check it in a player.
  • The whole file is held in memory alongside the result, and WebAssembly is 32-bit, so a recording beyond roughly a quarter of a gigabyte becomes unreliable — particularly on a phone.

FAQ

Questions that come up when a video has the wrong sound on it.

How do I remove the audio from a video?

Add the video, leave the mode on Mute with the whole video selected, and press Apply. The audio track is simply left out of the output while the video track is copied across packet for packet, so nothing is decoded on either side and the picture is untouched. It is usually much faster than re-encoding a video, though the wait still depends on how much data the browser has to read and write. The result stays in the same format you started with.

How do I replace the audio in a video with my own?

Choose Replace and add an audio file — MP3, M4A, WAV, FLAC and most other formats are read. The original sound is dropped and yours is written in its place. You control where it starts, whether it loops if it is shorter than the video, its level, and its fades, and the timeline shows how far it reaches. If your track needs no changes and the container accepts its codec, it is copied in without being encoded, so both streams come through untouched.

Does muting a video reduce its quality?

No. The video track is copied rather than decoded, so the picture in the result is the same compressed data it was before — not a re-compression of it. That is worth knowing because most sites that mute videos re-encode the whole file to do it, which costs a long wait and a visible generation of quality for an operation that touches one stream. The only thing lost is the sound, which is what you asked for.

Are my videos uploaded anywhere?

No. The engine is FFmpeg compiled to WebAssembly and it runs inside your browser tab. Both files are read from disk into the tab's memory and the finished video is written back out as a download. Nothing about either of them crosses the network, which you can confirm by watching the network panel while a render runs.

How do I add background music to a video?

Choose Add music and drop in the track. The original sound is kept and the music is mixed underneath it at whatever level you set — around a quarter of the original is the usual starting point for speech. Press play on the video first: the browser plays both at your settings, so you hear the balance before rendering rather than after. If the music is shorter than the video, Loop repeats it to the end; if it is longer, the extra is cut off, and the timeline shows how much.

The audio in my video is too quiet. How do I make it louder?

Choose Volume and press Measure the level. That reads the whole track and reports its loudest point and its average as decibels below full scale — a recording at minus eighteen has eighteen decibels of unused room. Setting the peak to just under the ceiling then uses exactly that room, which is more reliable than picking a number and rendering to see. Bear in mind that raising a quiet recording raises its noise with it, and that nothing can undo distortion that was recorded in.

What happens if my music is shorter than the video?

You choose. Leave silence plays the track once and lets the rest of the video run without added sound, or Loop starts it again each time it ends until the video finishes. Either way the video keeps its full length — this is worth stating, because the usual command-line approach to putting audio on a video shortens the picture to match a shorter track, which is exactly the failure this tool is built to avoid. The finished file is checked against the original length before you are offered it.

Can I mute just one part of a video?

Yes. Choose Mute, switch from the whole video to one stretch, and drag the range across the length of the file. Everything outside it keeps its original level. Changing part of a track means the sound is decoded and encoded once, unlike removing it outright, though the picture is still copied untouched. The cut is sharp at both ends, so on a loud passage the edge can be audible.

What format do I get back?

The same container you put in — an MP4 comes back as an MP4, a MOV as a MOV, a WebM as a WebM. That is deliberate rather than a limitation: the source container is the one wrapper certain to accept the video stream exactly as it stands, and choosing anything else could force the picture to be re-encoded, which is the one thing this tool never does. If you also need a different format, convert the result afterwards with the video converter.

Can I extract the audio from a video instead?

Not here — this tool writes videos, not audio files. Pulling a track out of a recording and saving it as an MP3, M4A, WAV or FLAC is what the audio converter linked below is for, and it is also the way to prepare a soundtrack taken from one video for use on another: extract it there, then bring the audio file back here.

Why does my added track stop before the end of the video?

Because it is shorter than the video and it was set to play once. The timeline says so directly — it names how much of the video the added sound covers and how long the silent remainder is — and the fix is either Loop, moving the start earlier, or a longer file. The reverse case is shown the same way: a track longer than the video has its overhang cut off, and the amount is stated before you render rather than discovered afterwards.

Does this work with iPhone, 4K, or HEVC video?

Yes, and better than the other video tools here do. HEVC recordings from an iPhone and AV1 downloads are difficult elsewhere on this site because this build cannot write HEVC and cannot decode AV1 without hardware a browser tab does not have. Neither applies when the stream is only copied — the packets move without anything looking inside them. Resolution is equally irrelevant for the same reason: a 4K file costs no more than a 720p one, since neither is decoded. The one thing that does change is the preview: a browser that cannot play HEVC cannot play it here either, so on Chrome or Firefox you would be setting levels for an iPhone recording without hearing them.

Related tools

More local, browser-only tools for working with video and audio.