Convert M4A to WAV
Get an Apple recording into the format editors and transcribers ask for.
How it works
Most requests for this conversion come from a form rather than from a listener. Transcription services, forced-alignment and speech-training pipelines, phonetics software, voice-cloning tools and a good deal of academic analysis code all specify uncompressed WAV at a stated sample rate, frequently 16 kHz mono. They do that to remove a variable: two AAC decoders can differ in the last bit, and a pipeline that measures the waveform would rather not inherit that argument.
The other common source is editing. Dropping an M4A into a multitrack session works, but the program decodes it on import anyway and each program does so slightly differently. Converting first makes the decode deliberate and repeatable, and gives every collaborator the identical file.
Two codecs, one extension
An M4A holds either AAC, which is lossy, or Apple Lossless, which is not. Both decode to the same kind of PCM and both are handled here without being told apart in advance. The distinction matters only in what the numbers mean: from Apple Lossless the samples are the recording, and from AAC they are the decoder's reconstruction of it.
Neither case improves by writing more bits per sample. Decoding a 64 kbps voice memo at 24-bit produces a file half again as large describing precisely the same reconstructed signal.
Picking the settings the receiving tool wants
Read the requirements before converting rather than after. Speech pipelines built on standard toolkits usually want 16 kHz mono at 16-bit, because that is what their acoustic models were trained on and anything else gets resampled on their side with results you cannot see. Music and general editing want the source rate left alone at 16-bit or 24-bit.
Downmixing to mono is the setting with real consequences. It is right for one speaker on one microphone and wrong for an interview recorded in stereo with a person on each side, where the separation is what makes overlapping speech legible. When in doubt, keep the channels and let the receiving tool decide.
Size follows directly from those two choices. 16 kHz mono at 16-bit runs about 1.8 MB per minute, so an hour of speech is roughly 110 MB. The same hour at 44.1 kHz stereo is around 600 MB, which is enough to fail an upload limit that the M4A passed comfortably.
Questions
Is my recording uploaded?
No. The M4A is parsed and decoded on your machine by the browser's own audio decoder, and the WAV is written in a Web Worker. Nothing is sent. That matters more here than on most pages, because the recordings people convert for transcription are interviews, consultations and meetings, and the usual way to transcribe one involves handing it to a company outright.
What sample rate does a transcription service want?
16 kHz mono is the common answer, because speech recognition models are typically trained at that rate and telephone-band audio carries almost all the intelligibility of a voice. Some services accept anything and resample internally. Check first: sending 44.1 kHz stereo to a system that wants 16 kHz mono is not an error, but it uploads twenty times the bytes for no gain.
Should I choose 24-bit?
Only if further processing is coming. From a lossy AAC source the extra depth stores no additional detail, since the decoder's output does not have any. It does prevent rounding noise from building up across a long chain of gain and filter operations, so it is a reasonable choice for an editing master and a waste for a delivery file.
How large will the result be?
Multiply minutes by 10.1 MB for 44.1 kHz 16-bit stereo, 5.3 MB for mono, or 1.8 MB for 16 kHz mono. A one-hour meeting recording that arrived as a 30 MB M4A becomes about 600 MB at full rate in stereo. The tool writes RF64 above 4 GB so long recordings do not hit the original WAVE format's ceiling.