Convert MP3 to Opus

Everything else about this conversion is a cost. The audio is decoded and encoded again, the compatibility of the result is narrower than the MP3’s, and nothing about the recording improves. Whether it is a good idea reduces to one question: how much space does it actually save, and is that space worth anything to you.

That is a real question with a real answer for some archives and a firm no for others. A hundred hours of recorded lectures on a phone is a genuine problem that this solves. A tidy folder of albums that already fits is not a problem, and re-encoding it buys a second generation of compression in exchange for gigabytes nobody needed.

MP3 at 128 kbps costs about 0.96 MB a minute and at 192 kbps about 1.44. The three bands here resolve to 128, 64 and 38 kbps of Opus, which cost 0.96, 0.48 and 0.29. So the honest saving is a half or a little under a third, depending how far you are willing to go — not the tenfold reduction people remember from converting a WAV.

Across a large archive that still matters. Four hundred hours of recorded talks at 128 kbps MP3 is about 23 GB; the same material on the Small file band is about 6.8 GB. Across forty albums it does not matter at all, which is the distinction this page keeps coming back to.

Two lossy codecs in sequence compound, because the second encoder sees a waveform that looks complete and has no way to identify what the first one removed. It spends bits preserving artefacts and applies its own model on top of decisions already made.

How audible that is depends almost entirely on the ratio between the two bitrates. Encoding a 128 kbps MP3 on the High quality band, which gives 128 kbps of Opus, is close to inaudible; putting music through Small file at 38 kbps is obviously a compressed file. Opus degrades gracefully rather than harshly, which is a genuine advantage here — a heavily compressed Opus file sounds soft rather than metallic — but graceful is not free.

Recorded lectures, interviews, conference sessions, sermons, audiobooks, podcast back catalogues and voice notes are all a single speaker in a room, which is the easiest material any codec ever handles. Opus was designed with speech coding built into it and is exceptionally good at low rates on exactly this content.

The Small file band is the one to use, and on this material it costs less than the number suggests. A 128 kbps MP3 of a single speaker comes back at 38 kbps, under a third of the size, and a listener would struggle to identify which is which on headphones. The channel count is not something this page asks about — a stereo recording of one microphone stays stereo — so the saving comes from the band alone. For an archive measured in hundreds of hours, this is the conversion that makes it portable.

A music collection is the wrong shape for this. The files are already fairly small, the saving is a half at best, the material is what a codec finds hardest, and the destination — a car, a hi-fi, a portable player — is exactly where Opus support runs out.

There is one situation where it makes sense: a phone that has to hold a very large library and never plays anything anywhere else. Even then, converting from the original CDs or from FLAC, if either still exists, produces a better file at the same size, because it avoids the compounding entirely. Reach for the transcode when the MP3 is genuinely all there is.

MP3 works in fixed blocks and pads the start and end of every track to fill the last one, typically by a few dozen milliseconds. Players that read the encoder’s padding information can trim it; decoding for a conversion does not guarantee it is removed.

The Opus encoder then adds its own priming, and although the Ogg Opus header records that pre-skip so a well-behaved decoder can discard it, the MP3’s padding has by then become ordinary audio. The result on a live album, an opera or a mixed set is a short silence at every track boundary. Leave those albums as they are; they are also the ones you would least want re-encoded.

The failure mode of this migration is not audio quality, it is a device that will not play the result. Support is universal in browsers and desktop software, thin in car head units, older players, televisions and hi-fi streamers, and there is no way to determine it from a specification sheet.

So convert a handful, put them on the actual destination, and press play before doing anything irreversible. Delete the MP3s only after the archive has been used in anger for a while. This costs a week of holding both copies and prevents the version of this task where a 200 GB library is converted into an unplayable one.

An Ogg Opus file carries a block of named text fields, so title, artist, album and the ordinary values from the MP3’s ID3 tags have somewhere to land. The two systems are not identical, and anything unusual a particular tagger wrote is the first thing to go missing.

For a speech archive this rarely matters, since those collections are organised by filename and folder anyway. For anything you intend to browse in a player, convert one album, look at it there, and fix the fields across the batch in a tag editor if they need it.

Each drop takes up to a hundred files at 100 MB each, which for MP3 sources is over an hour of audio per file, and returns them together as a ZIP. Nothing is fetched to make it work: the Opus encoder ships with every browser, so the conversion starts the moment the files land.

Nothing is uploaded either, which for this material is the point. The archives people shrink in bulk are lectures, interviews, therapy sessions, legal recordings and personal voice notes, and a conversion that keeps them on the machine they already live on is the only kind worth using on them.