M4A is the format Apple devices use when they record sound. Voice Memos on iPhone and iPad, QuickTime audio recordings on a Mac, the audio exported from GarageBand and the local recording that Zoom saves under the name audio_only.m4a are all this type. Technically it is AAC audio stored in an MP4 container with the video part left empty, which is why some programs treat it as a video and others refuse it altogether.
That refusal is the usual reason people land on a page like this one. A transcription tool asks for MP3 or WAV, the voice memo is M4A, and the search for a converter begins. Transcrify accepts the M4A directly. The container is opened on our servers, the AAC stream is decoded and the words are transcribed, without a conversion step on your side.
Quality is not an issue. Voice Memos records at a bitrate meant for speech, and the recorder sits close to the person talking, so these files often transcribe better than a video shot from across a room. The weak spots are the same as for any phone recording: wind, a pocket, or a table that picks up every tap.
Voice Memos does not expose files the way a computer does, so the step people get stuck on is not the transcription but the export. Open the recording in Voice Memos, tap the three dots or the Share button, and choose Save to Files. Pick a folder in iCloud Drive or On My iPhone. Then come back to this page in Safari, tap the upload box and select the file from Files.
If you are at a Mac, it is simpler: Voice Memos syncs through iCloud, and you can drag a memo straight from the app onto this page. Memos recorded on an Apple Watch appear in the same list once they have synced.
Recordings longer than a few minutes can be several tens of megabytes. The upload is sent in small chunks, so a long memo over a mobile connection still completes, and you can watch the progress on the page.
When Zoom records a meeting to your computer, it writes two files: a video MP4 and an audio_only.m4a next to it. For a transcript, the audio file is the better choice. It is a fraction of the size, uploads faster and contains exactly the same speech. Drop that file here and you get the meeting as text with each participant labelled, as far as the voices can be told apart.
The same applies to lecture capture apps on iPad and to the many note-taking apps that export audio in M4A. Whatever produced the file, the transcript comes back with timestamps per segment, in any of more than fifty languages, detected automatically.
Voice memos tend to be personal: a meeting with a doctor, an idea dictated in the car, a conversation someone agreed to have recorded. The speech models run on servers we operate, and the recording is not sent to a third-party speech API. Transcripts you make here as a guest are kept the same way and move into your account if you create one later.
A guest can transcribe 10 recordings of up to 3 hours each and read or copy the result. A free account adds the editor, your transcription history and downloads in SRT, VTT, ASS and TXT, with 5 hours across 10 files. AI chat with the recording, summaries and export to PDF or Word are part of the paid plans.