Best Audio Settings for Transcription: Format, Sample Rate and Channels

Last updated: 28 October 2026

Whether you upload a recording to an automatic transcription service or hand it to a human transcriber, the file settings affect accuracy and upload speed. The most important rule: the quality of the recording itself matters more than the settings. Converting cannot fix a recording made far from the microphone or full of noise. But the right settings make sure you do not lose anything needlessly.

Recommended settings

Setting	Recommendation	Why
Format	WAV or FLAC for the original. MP3 at 96 to 128 kbps is acceptable for delivery	Lossless formats keep every detail. Avoid bitrates below 64 kbps
Sample rate	16 kHz is enough for most transcription systems	Human speech does not need higher rates. 44.1 or 48 kHz do no harm but make the file bigger
Channels	Mono for a single speaker	Stereo doubles the size with no benefit for speech
Bit depth	16-bit	More than enough for speech
Level	No clipping, with peaks roughly between -6 and -3 dB	Clipping cannot be fixed later and confuses transcription

If the service you use specifies its own requirements, follow those first.

Example sizes (one hour of speech)

WAV, 44.1 kHz, stereo, 16-bit: about 635 MB.
WAV, 16 kHz, mono, 16-bit: about 115 MB.
MP3, mono, 96 kbps: about 43 MB.

How to prepare the file with Sayvoi

Open convert.sayvoi.com and drag in your recordings.
Choose WAV (or MP3 at 96 to 128 if you want a smaller file).
Set the sample rate to 16 kHz and channels to mono.
Press Convert all and download the result.

For details, read sample rate and mono or stereo. If you have a video, extract the audio first: how to extract audio from video.

Common mistakes that reduce accuracy

Recording far from the speaker. Place the microphone as close as you can.
Background noise or music. Read noise floor and how to record clean audio with your phone.
Distortion from a recording level that is too high. Read audio clipping.
Repeated conversion between compressed formats, since each pass adds loss.
Very aggressive noise removal, which can distort the speech itself and hurt accuracy. A clean recording from the start is better.

Check before uploading

Try the quality checker to verify level, silence and distortion before sending a recording. For professional projects, read voice data recording specs and check recording quality before delivery.

Frequently asked questions

Should I upload the video or just the audio?

Usually just the audio. It is smaller and uploads faster.

Does 16 kHz hurt accuracy?

For speech, no; it is enough for most transcription services. But if you have a reason to preserve the original, keep it at its original rate and convert a copy for upload.

Are my files kept private?

Conversion happens inside your browser and files are not uploaded to a server. Read why you should not upload private recordings to converters.