M4A, AAC, OGG, WAV or FLAC to Text: Which Audio…

Most "how do I convert this to text" questions are really "will the tool open this file". Voice memos arrive as M4A, messaging apps export OGG, field recorders write WAV, archives keep FLAC, and video exports carry AAC inside an MP4. A local transcription app that reads all of them saves the conversion step, and the conversion step is where quality is lost.

This guide lists what ClapClip's audio to text reads directly on Windows, where each format usually comes from, and what to do with the few that still need converting.

Key takeaways

Six audio formats transcribe as they are: MP3, WAV, FLAC, AAC, M4A and OGG.
Video files do too: MP4, MOV, MKV and WEBM have their audio track extracted automatically.
Converting to MP3 first makes things worse, not better. Every lossy re-encode removes detail the recognizer uses.
A few formats still need a conversion: files such as .opus, .wma, .amr or .3gp are not on the list; convert them once, to WAV, and keep the original.
The format matters less than the recording. A clean M4A from a phone on the table beats a noisy WAV from across the room.

Which format is this file, and where did it come from?

Format	Typical source	Codec and container	Transcribe directly?
M4A	iPhone Voice Memos, many Android recorders, Zoom audio-only recordings, Apple Music exports	AAC audio in an MPEG-4 container	Yes
AAC	Standalone audio exports, downloaded tracks, the audio inside most MP4 videos	AAC, raw or in a container	Yes, standalone or inside MP4 and MOV
MP3	Podcast downloads, older recorders, exports from almost any editor	MPEG audio layer III	Yes
WAV	Field recorders, audio interfaces, dictation hardware, studio exports	Usually uncompressed PCM	Yes
FLAC	Archived recordings, audio enthusiasts, some long-form podcast masters	Lossless compression of PCM	Yes
OGG	Messaging-app voice notes and game or stream recordings, usually Vorbis or Opus inside an OGG container	Vorbis or Opus in OGG	Yes for .ogg files
MP4, MOV, MKV, WEBM	Screen recordings, meeting recordings, camera footage, downloaded video	Video containers; the audio track is what gets transcribed	Yes, audio extracted automatically

The extension is not always the truth. A messaging app may hand you a file ending in .opus that is an Opus stream in an OGG container, a renamed file keeps its real contents, and a "WAV" from some devices is compressed inside. When a file is refused, check what it actually is before converting it.

Use this rule: if the extension is on the list, add the file and start. Only convert when the tool refuses it.

When do you have to convert?

Formats outside the list need one conversion before transcription. The common ones:

.opus from messaging apps and some call recorders.
.wma from older Windows recorders and Windows Media Player rips.
.amr and .3gp from feature phones and some Android call recorders.
.aiff from older Mac audio software.

Convert once, to WAV, so that no further detail is lost, and keep the original file. With FFmpeg installed the command is the same for all of them:

ffmpeg -i voice-note.opus voice-note.wav

For a folder of files, a loop in PowerShell does the same job:

Get-ChildItem *.amr | ForEach-Object { ffmpeg -i $_.FullName ($_.BaseName + ".wav") }

Use this rule: convert to WAV, not to MP3. WAV is larger, but it is a one-time intermediate and it keeps everything the source had.

Does the format change accuracy?

Less than people expect. Accuracy depends on the recording: microphone distance, background noise, overlapping speakers, accent and how much of the speech is actually in the file. Within that, three things about the format do matter:

Lossy re-encoding compounds. A voice memo recorded as AAC, converted to MP3 to send, then converted again to upload has been compressed three times. Transcribe the earliest copy you have.
Very low bitrates hurt. Voice notes are often recorded at low bitrates to save data; they are still usable, but names and numbers are where the loss shows first.
Sample rate does not need to be high. Speech recognizers work from a 16 kHz representation internally, so a 48 kHz studio WAV has no accuracy advantage over a 16 kHz one; it is just larger.

Use this rule: when two copies of the same recording exist, transcribe the one that has been through the fewest encodes, regardless of its extension.

How do you transcribe each one locally?

The workflow is the same for every format: open ClapClip on Windows 10 or 11, add the file, choose the spoken language and the output (plain text, timed SRT or both), and start. The file is processed on your own PC, using the GPU where there is one, and is not uploaded. There is no length limit, so a two-hour FLAC is handled like a thirty-second voice note, and a mixed folder of M4A, OGG and MP4 files can be queued at once with batch transcription, one transcript per file.

The audio to text page covers every audio format above, and video to text covers MP4, MOV, MKV and WEBM.

Bottom line

If the file ends in MP3, WAV, FLAC, AAC, M4A, OGG, MP4, MOV, MKV or WEBM, transcribe it as it is. Convert only the handful of formats outside that list, once, to WAV. The recording's quality decides the transcript's quality far more than its container does, so spend the effort on the microphone, not the converter.

Related guides

Sources

FFmpeg documentation: the conversion tool used in the examples.
Robust Speech Recognition via Large-Scale Weak Supervision (Radford et al., 2022): the Whisper paper, including the 16 kHz input representation.
openai/whisper on GitHub: the open-source model family that local transcription tools build on.

M4A, AAC, OGG, WAV or FLAC to Text: Which Audio Formats Transcribe Directly