Convert OPUS to AIFF

Opus runs on a fixed 48,000 Hz internal clock. A decoder reports that rate regardless of what the encoder was originally handed, so an Android recording made at 44,100 Hz, a voice note captured at 16 kHz and a stem pulled from web video all arrive here as 48 kHz audio, and the AIFF records that rate in its COMM chunk.

It is not a defect and no listener will hear it. It is a fact your session has to be told. A project running at 44,100 Hz will either convert the import on the fly — costing a processing step and a small quality decision nobody made deliberately — or, in older software and in hardware samplers, play it at the wrong speed and pitch. Check the session rate before importing a set, and resample deliberately in an editor if 44.1 is required.

Uncompressed 16-bit stereo at 48,000 Hz costs 192,000 bytes a second — about 11.5 MB a minute — whether the minute is a performance or silence, because every sample is written out as a plain number. Opus at 64 kbps costs about 0.5 MB a minute, and messaging voice notes are frequently encoded lower still.

So the multiple lands near twenty, and higher from a voice note. A ten-minute recording that arrived as a 5 MB attachment becomes roughly 115 MB as an AIFF, and an hour becomes close to 700. That is the number to check against your session folder and your backup before converting a batch, because it is easy to turn a directory of small attachments into something that no longer fits on a working drive.

Opus compression removed parts of the audio permanently, according to a model of what a listener would not miss, and those parts are not stored compactly somewhere in the file. AIFF then does the opposite job perfectly: it stores every sample it is given, without compression, exactly as handed over.

The result is a very large file containing precisely what the small one contained. That is useful when something downstream demands PCM and it is not an improvement by any measure. Where the person who sent you the Opus still has the recording it came from — a WAV off a field recorder, a session bounce, an uncompressed export — one message asking for that file is worth more than any conversion on this page.

Opus is a Xiph codec from 2012 and Apple never adopted it for its media path, so Mac audio tooling that predates a bundled decoder has no way to read it at all. Hardware is the harder case: samplers, groove boxes and older studio units implement PCM and no decoder at all, because a decoder is firmware somebody would have to write and maintain.

Current Logic, Pro Tools and Reaper will import a good many formats directly, so if the refusal came from software rather than hardware it is worth trying the import once before converting. Where the requirement comes from a device or from a written ingest specification, converting is the entire answer and there is no argument to be had with it.

The PCM written here is signed 16-bit. That would be worth arguing about from a 24-bit master, where the extra depth exists so that gain changes, summing and limiting have room to work without the noise floor rising into audibility. From an Opus file there is no such depth to preserve — a lossy codec discarded that information long before the decoder ran.

A 16-bit AIFF sits inside a 24-bit or 32-bit session without any further loss, because the session depth governs the mix rather than every source dropped into it. Where a specification asks for 24-bit uncompressed audio, this converter cannot produce it, and it is worth asking what those bits would be holding given where the audio came from.

An AIFF is a sequence of chunks and the one written here holds exactly two. COMM states the channel count, the frame count, the bit depth and the sample rate. SSND holds the samples. Nothing else is written — no name, no annotation, no author, no artwork.

Opus files carry Vorbis comments, an open list of uppercase keys with text values, and a client's voice note or a documented field recording quite often has something useful in there: a date, a location, a take number, a description of what is being recorded. All of it ends at this boundary. Copy what matters into the file name or a session note before converting, because the AIFF cannot be asked afterwards.

A great deal of what arrives as `.opus` is genuinely mono: a phone recording, a messaging voice note, a single-microphone capture. Nothing here promotes it to stereo, and that is the right outcome — two identical channels double the file size to convey the same information, which at 11.5 MB a minute is not a trivial waste.

In a session, a mono region placed on a stereo track behaves normally and is panned by the desk rather than by the file. If something in your import chain is converting mono sources to stereo automatically, that is worth turning off before a batch, since it doubles a set of files that were already twenty times their original size.

Files are capped at 100 MB each, and for an Opus source that is a great deal of audio — several hours of speech at typical voice-note bitrates. So the input limit will almost never stop you. What will is the output: a two-hour recording that comfortably passes the cap as a 60 MB Opus becomes well over a gigabyte of PCM inside a browser tab.

For voice notes, takes, stems and cues this is comfortable and a hundred files can be dropped in one pass and taken back as a ZIP. For a long-form recording, split it in an editor first or convert it with a desktop tool. Saying so in advance is more useful than letting a long file exhaust the tab's memory after it has been read.

Writing AIFF costs no dependency at all. The decoded samples are muxed as a WAV and rewrapped by arithmetic — the header rebuilt big-endian, the sample rate written as the 80-bit extended float the format insists on, and each sample's bytes reversed. Decoding Opus needs nothing either, because every browser implemented it for calls.

So this conversion moves nothing across the network in either direction, and you can watch that in developer tools while it runs. For a client's unreleased material, a recorded conversation or anything under a confidentiality clause, that is the property that decides whether the job can be done in a browser at all rather than a convenience on top of it.