Convert WebM to MP3

People rarely download a WebM deliberately, and when the file is only a few hundred kilobytes it is almost never a video. It is a voice message saved out of WhatsApp Web, an audio clip from a Discord channel, a recorded call from a meeting tool, or something a page captured through the browser’s own recorder. All of those produce a `.webm` file containing Opus audio and, frequently, no picture at all.

That explains the symptom that brought you here. The file plays perfectly in the browser it came from and refuses everywhere else: the car will not see it on a USB stick, the phone’s music app skips it, the transcription service rejects the upload, and the colleague you emailed it to replies asking what it is. Nothing is wrong with the file. It is simply in a format built for browsers.

The audio inside is Opus, published by Xiph.Org in 2012 and standardised as RFC 6716. It is genuinely one of the best audio codecs there is — it beats MP3 comfortably at every bitrate and handles speech at rates where MP3 falls apart, which is exactly why every voice-messaging product on the web uses it. Every browser can decode it, because WebRTC made that mandatory.

Outside the browser the picture inverts. Car head units, hi-fi streamers, dictation software, most Windows media players in their default state and a great deal of professional audio software have no Opus decoder at all. MP3 is nineteen years older and worse by every measurement, and it is the only format you can hand to a stranger with no questions attached.

This is the odd corner of the web platform that makes this pair different from the others. WebCodecs gives a page the operating system’s own encoders, and MP3 is the one format it uniformly refuses — the single most requested audio target on the internet is the one no browser will write.

So the first time you ask for an MP3 here, a compiled build of the LAME encoder is downloaded and registered. That is the brief pause before the progress bar moves. It happens once per session; convert ten recordings in a row and only the first waits. Nothing about the file is involved in that download, and nothing about the file goes back the other way.

The Opus in a voice note has already discarded detail. Encoding an MP3 from it discards a second helping, because the MP3 encoder is working from what survived rather than from the original sound. There is no way around this short of not converting — it is arithmetic, not a shortcoming of any particular tool.

Whether it matters depends entirely on what was recorded. Speech at conversational quality survives the round trip with nothing a human ear can detect, which covers essentially every voice note, meeting recording and interview. Music does not fare as well: cymbals and reverb tails are where the second pass shows. If the WebM is a rehearsal recording or a mix, use the high quality band and keep the original.

Roughly one megabyte per minute at the balanced setting, and that figure is almost independent of what the WebM weighed. A ninety-minute recorded call lands near ninety megabytes whether the source was six megabytes of tightly packed Opus or a two-gigabyte screen recording with the same audio inside it.

The quality control moves that up or down by about a factor of two in each direction. The small setting is right for speech that is going to be played through a phone speaker or emailed; the high setting roughly doubles the file and is worth it only when there is music in the recording. Opus is dramatically more efficient than MP3 at low bitrates, which is why the MP3 of a small voice note is so often several times the size of the WebM it came from.

Not every WebM here is audio-only. Screen recordings, recorded webinars and captured meetings all arrive as WebM with a video track, and often the picture is the part nobody needs — a slide deck that was emailed anyway, or an hour of somebody’s shared screen behind twenty minutes of useful discussion.

The video track is discarded outright rather than ignored, so an hour of 1080p capture becomes an hour of speech at a fraction of the size, which is a file you can put on a phone. If there is no audio track in the WebM at all the conversion stops and says so, which is the honest failure for a video somebody recorded with the microphone off.

A WebM can carry several audio tracks, and an MP3 holds exactly one. Whichever track the file lists first is the one that comes out. For a voice note this is a non-question; for a recorded meeting where the tool wrote the microphone and the system audio as separate tracks, it decides which half of the conversation you get.

There is no way to choose between them here, and pretending otherwise would be worse than saying so. If both tracks matter, export them separately from the software that made the recording, or open the WebM in Audacity, which reads the container and can mix or split the tracks properly.

Voice notes arrive in quantity. Drop up to a hundred WebM files at once and each converts independently with its own progress bar; once more than one has finished, a single button offers the lot as a ZIP. There is no queue, because there is no shared machine to queue on.

This is where doing the work locally stops being an argument about privacy and becomes one about time. Converting forty voice messages through a site that uploads them is forty uploads, a daily cap and a wait proportional to your connection. Here it is proportional to your processor, which is faster and does not run out.