Convert WebM to FLAC

A WebM carries Opus or, on older files, Vorbis. Both are lossy codecs: they analyse the sound, decide which parts a listener will not notice, and throw those parts away permanently. That happened when the file was recorded, and no conversion performed afterwards can reverse it.

FLAC is genuinely lossless, and here that means it losslessly preserves an approximation. The FLAC will be several times larger than the WebM and will sound exactly the same — not slightly better, not imperceptibly better, identical. If you came here believing this conversion upgrades a recording, the useful thing this page can do is tell you it does not, before you spend the disk space.

FLAC was released by Josh Coalson in 2001, taken over by Xiph.Org in 2003 and standardised as RFC 9639, and it was designed for one job: storing audio you intend to still have in twenty years. It compresses PCM by roughly half with no loss, which is the difference between an archive that fits on a disk and one that does not.

It also carries a checksum on every frame. That sounds like a footnote and is the most important property in the list. A WAV that has been silently corrupted by a failing drive or a bad copy plays with a click in it and looks fine; a FLAC with the same damage reports a mismatch when it is verified. For material you cannot re-record — an interview, a lecture, a call, a broadcast capture — being able to check a file is intact is worth more than any amount of bitrate.

Both preserve the decoded samples exactly, so the choice has nothing to do with fidelity. FLAC is about half the size, holds tags, and verifies itself. WAV, written by Microsoft in 1991, is read by literally everything and needs no decoder at all, which is why it remains the input every audio tool documents.

The split follows the verb. If you are going to *do* something with the file today — transcribe it, edit it, feed it to a script — take the WAV, because it is what the tool wants and its size does not matter for an afternoon. If you are going to *keep* it, take the FLAC, because in five years the size and the checksum will both matter and no editing session will.

There is a version of this conversion that is entirely correct, and it turns on the word "master". If the WebM is the only copy of something — a recorded deposition, a community radio capture, a family voice message, a conference talk nobody else filmed — then it is a master regardless of how it was compressed, and mastering practice is to stop compressing it any further.

From the FLAC onward there is no generational loss. Every clip, every edit, every export starts from the same samples. That is the correct reason to make one, and it holds no matter how lossy the original recording was. What does not hold is the idea that the FLAC has more sound in it than the WebM did.

FLAC typically lands between 40 and 60 per cent of the equivalent uncompressed audio, with speech at the efficient end because it is quiet and simple. Against the WebM it came from, expect a multiple rather than a fraction: an hour of recorded conversation might be forty megabytes as Opus and two hundred as FLAC.

That five-fold jump is a fair description of how much better Opus is at speech than any lossless scheme can be, rather than a sign that something has gone wrong. Plan the disk before converting a year of recordings, and remember that the FLAC is a filing decision — it is meant to sit still, not to be carried around on a phone.

FLAC encoding is not something every browser exposes. Where the platform provides an encoder, it is used directly; where it does not, a compiled build of the reference encoder is downloaded the first time you ask for a FLAC and reused for the rest of the session. That is the brief pause before the progress bar moves on the first file.

It is fetched once and it is a static file like any other on the page. Nothing about your recording is involved in that download, and the audio itself never goes anywhere — the encoding happens in the tab, whichever of the two encoders ends up doing it.

Only the first audio track survives, and any video track is discarded outright. A recorded meeting that captured the microphone and the shared system audio as two tracks yields one of them, and there is no way to pick. For archival work that is worth checking before the original is filed away or deleted.

Tags do come across, to the extent the WebM has any. Title, description, artist and date are read out of the Matroska tag element and written into the FLAC as Vorbis comments — but a browser recorder or a meeting tool fills in almost none of them, so in practice the FLAC arrives with audio and a blank sheet. For an archive that is arguably the right starting point: a recording filed under whatever a meeting tool called it is a recording nobody will find. Set the date, participants and description deliberately in a tagger; FLAC has a proper tagging system and an archive is the one context where using it pays off.

Archiving is a bulk activity. Up to a hundred WebM files can be converted in one go, each with its own progress, and the results download together as a ZIP. The work happens on your machine, so a large batch is limited by your processor rather than by an upload allowance or a daily cap.

That also settles the question that makes this pair awkward elsewhere. The recordings people convert for archiving are the ones with the longest expected life and the most sensitive contents, and running a decade’s worth of them through an unfamiliar server is a poor first step in a preservation plan. Nothing carrying the audio is sent anywhere here.