MP3 to text

What MP3 does to your audio

MP3 is a lossy format. When a recording is saved as MP3, parts of the signal judged least audible to human ears are permanently removed to make the file smaller. For listening, this works extremely well — that is why the format took over. For transcription, it matters slightly more, because a speech model is not a human ear and does not necessarily ignore the same details.

In practice, the effect is small at normal bitrates. An MP3 at 128 kbps or above transcribes about as well as the uncompressed original. Below roughly 64 kbps, and especially at the very low bitrates sometimes used for voice recorders and phone systems, the loss becomes audible as a muffled or watery quality, and accuracy drops with it.

The important point is that this loss is permanent. Converting a low bitrate MP3 to WAV before uploading does not restore anything; it produces a large file containing the same degraded audio. If you have access to the original recording in a better format, use that instead. If the MP3 is all you have, upload it as it is.

Where MP3 files usually come from

Most MP3s that people want transcribed fall into a few groups, and each has a typical quality. Downloaded podcast episodes are usually encoded well and transcribe cleanly. Files exported from editing software depend entirely on the settings chosen at export. Recordings from older dictaphones and voice recorders are often at low bitrates and mono, which is where you are most likely to notice errors.

Audio pulled out of a video and saved as MP3 is a common case too. If you did that conversion yourself specifically to upload here, it was unnecessary — video files can be uploaded directly, and doing so avoids one round of lossy compression.

MP3 compared with the alternatives

WAV stores audio without compression. It is several times larger for the same recording and holds the original signal exactly. If you are choosing a format at the moment of recording and file size is not a concern, WAV is the safer choice.

FLAC compresses without discarding anything, giving you roughly half the size of WAV with none of the loss. It is the best compromise when you want smaller files but no quality cost, and it is supported here.

M4A, which is what most phones produce, is lossy like MP3 but generally more efficient at the same bitrate. If your phone offers it, there is no reason to convert to MP3 first.

None of this should worry you if you already have an MP3 recorded at a normal quality. The difference between formats is much smaller than the difference between a close microphone and a distant one.