Convert M4A to OPUS

M4A is Apple's audio container from 2004 and it turns up in two very different roles. One is a music file from a store or a library. The other, far more common on a phone, is a recording: a Voice Memo, a lecture captured on a desk, a meeting exported from Teams or Zoom, an interview, a call. Those are long, spoken and frequently mono.

That distinction decides everything on this page. Music at 38 kbps is a compromise; speech at 38 kbps is what messaging apps have sent for a decade and nobody complains. If your M4A is an album track, the Opus band should be High quality and the page you probably want is the one about fitting a library on a phone. If it is forty minutes of somebody talking, read on — the numbers are dramatic.

An iPhone Voice Memo recorded in its default quality is around 64 kbps of mono AAC, which is roughly 28 MB an hour. A lecture captured at a higher setting, or in stereo, can be two or three times that. Email attachment limits are commonly 25 MB, and a good many upload forms stop at 20.

Opus at the Small file band writes 38 kbps, putting an hour near 17 MB; Balanced writes 64 kbps for about 29 MB; High quality writes 128 kbps for about 58. So the low band alone is usually the difference between an attachment that sends and one that bounces, and a three-hour recording that was 84 MB becomes about 51 MB without becoming harder to follow.

Opus was standardised as RFC 6716 in 2012 out of two codecs designed for different jobs — one for speech, one for music — and it switches between their techniques according to what it hears, without being told. That is why it holds together at bitrates where MP3 produces a warbling mess and AAC starts to smear.

The practical comparison for this pair: Opus at 38 kbps is roughly what MP3 needs 96 kbps to match on spoken material, and messaging apps run voice notes at 16 to 32 kbps on the same codec. Nothing else offered on this site is close for speech. The trade is compatibility, which the section below deals with honestly rather than in a footnote.

A recording of one person talking into a phone carries almost identical information in both channels, and storing it twice is a straight waste of the bitrate budget. Converting to mono roughly halves the file again on top of everything above, with no perceptible effect on a lecture, an interview or a meeting.

There is no channel control on this page, so a stereo M4A produces a stereo Opus. Where a recording is genuinely single-source, the sequence that pays is to sum it to mono in a free editor first and convert afterwards. Where a recording has two people at opposite ends of a table on separate microphones, leave it alone — the separation is the thing that makes a transcript attributable.

Every current browser decodes Opus, Android has for years, VLC and the desktop players all read it, and most transcription tooling accepts it because it passes the file to a decoder rather than to a list of extensions. That covers the majority of destinations for a converted recording.

What does not: older upload forms with a fixed allow-list, some legal and medical intake systems, car head units, and the built-in preview on iOS in certain situations. When a file is refused for looking wrong rather than being wrong, renaming it to .ogg occasionally helps, since the same stream is legal under either extension. When the receiving system genuinely has no Opus decoder, this was the wrong conversion and MP3 is the file it wants.

The M4A holds AAC and the AAC already threw away what its model of hearing said would not be missed. Opus now discards a second helping using different rules, and spends part of its budget describing the first encoder's artefacts rather than the recording. That is the unavoidable cost of any lossy-to-lossy conversion and there is no setting that removes it.

On speech it is close to unmeasurable in practice. Consonants and sibilance are what degrade first, and at 38 kbps Opus they survive well enough that automatic transcription accuracy is not meaningfully affected. On music, applause, or a recording with music underneath narration, the low band starts to sound like a phone call — take High quality there, or keep the M4A.

Opus runs on a fixed 48,000 Hz clock and a decoder reports that rate regardless of what the encoder was handed. Phone recordings are usually 44,100 or 48,000 Hz already, so in most cases nothing visible happens; a 16,000 Hz voice recording will nonetheless present as 48 kHz afterwards.

For listening this changes nothing. It matters in one place: a speech pipeline that expects a specific input rate, or a tool that reads the technical rate out of the file and logs it, will see 48 kHz rather than the original figure. Anything that resamples internally — which is nearly all current transcription tooling — is unaffected, but a pipeline built around a fixed rate should be told.

The registry lists two codecs for M4A: AAC and Apple Lossless. Recordings and purchases are AAC; a CD ripped in iTunes with the lossless setting is ALAC, and ALAC cannot be decoded on this path at all. Such a file stops with a message saying nothing in it could be converted, rather than producing something broken.

Size tells you which you have before you start. Four minutes of AAC is three to five megabytes; four minutes of Apple Lossless is twenty to thirty. If it is ALAC and the goal is a small file, the sensible route is to re-export from the source rather than through a converter that cannot open it — and if the goal was archival, ALAC was already doing that job.

A large share of the recordings this page exists for contain other people: a lecturer, an interviewee, colleagues in a meeting, a patient, a source. Those people agreed to be recorded by you, not to have a copy pass through a conversion service that keeps files for an hour and says so in a footer.

Nothing here goes anywhere. The AAC is decoded by your own device, and Opus is the one audio target that needs no encoder download, so the entire operation happens in the tab with the network idle. Files are capped at 100 MB each — comfortably over three hours of a typical Voice Memo — and a hundred can be dropped at once and taken back as a ZIP.