File formats and audio

Audio format

The file format the audio is stored in, for instance m4a, mp3 or wav.

The audio format is the way the recording is stored. For transcription the choice matters less than people think, with one exception: how hard the audio has been compressed.

The common ones

FormatTypeWhere you meet it
m4aCompressed (AAC)Standard on iPhone and Mac
mp3CompressedEverywhere, older standard
wavUncompressedRecording equipment, editing
flacCompressed without lossArchives
aacCompressedStreaming, video
ogg / opusCompressedBrowser recording, video meetings
mp4 / movVideo with an audio trackScreen recording, video meetings

Nearly all transcription tools accept all of these. Video is unpacked to the audio track automatically.

Lossy and lossless compression

Lossless (wav, flac) keeps all the information from the recording.

Lossy (mp3, m4a, aac, opus) throws information away to save space. At ordinary settings this is not noticeable to the human ear, nor to the transcription.

The problem arises with hard compression. An mp3 at 32 kbit/s, or a speech recording that has been compressed twice because it passed through several stages, loses details the model needs. Consonants go first, and consonants are precisely what distinguishes words from one another.

Practical rule

The default setting on your phone or recorder is fine. Do not pick “low quality” to save space, and do not convert a file more times than necessary. One minute of speech in m4a at the default setting is under a megabyte.

What actually decides the result is not the format, but how close the microphone was to whoever was speaking.

See also

Sample rate, word error rate.

Speech, written out

Sayable transcribes speech with dialect, tells the speakers apart and drafts the summary. Free to get started.