Recording tips that give clearer transcripts

Clearer audio gives the software more to work with. Most of what helps is cheap: a closer microphone, a quieter room and the right file settings. Here's what makers' own guidance says.

Microphone and room

Noise, echo and the recording setup all alter how speech sounds to the software¹. So start with the place and the microphone.

Google advises you to "Position the microphone as close as possible to the person that is" speaking, especially when there's background noise². It also says to avoid audio clipping². Clipping is the harsh distortion that happens when a sound is louder than the recorder can handle.

Before a long recording, do what Google suggests: "Listen to some sample audio. It should sound clear, without distortion or" unexpected noise².

One microphone or several

If you record several people and each has a separate channel, Google says to "send each channel separately to get the best recognition results"². If everyone is mixed into one channel, it says to send the recording as it is².

Don't clean the audio first

This one surprises people. Google says that "applying noise-reduction signal" processing before sending audio to its service "typically reduces recognition accuracy"². The service is designed to handle noisy audio, and Google advises that all noise reduction be disabled².

That advice is from one maker, about its own service. Another tool may differ, so read your tool's documentation.

File format and quality

A codec is the method used to squeeze audio into a file. A lossy codec drops some detail to save space. A lossless one keeps it.

Google says that "Using mp3, mp4, m4a, mu-law, a-law or other lossy codecs during recording or transmission may reduce accuracy"². It recommends lossless FLAC or LINEAR16 and says "Lower sampling rates may reduce accuracy"². The sampling rate is how many times per second the sound is measured. Google suggests 16,000 Hz or higher, and says phone audio should be left at its native rate rather than converted².

Your tool may not take every format. OpenAI's file transcription page, checked 2026-10-26, lists the formats mp3, mp4, mpeg, mpga, m4a, wav and webm. It also sets a cap of 25 MB per file³. If a file is rejected, check the limits on the maker's page.

A short test run

A two-minute trial costs little. Record a sample in the real room with the real microphone, in the seat where people will sit. Play it back before the meeting starts. If you struggle to follow the words, the software will too.

Try your hardest case in the sample: a quiet speaker, a door that bangs, a fan. Fix what you can hear, then record for real.

Keep the original

Google's advice is to avoid re-sampling, which means changing the sampling rate of audio you already have². Keep the original recording and upload that.

Once the transcript arrives, remember the limits of accuracy. Good audio helps, but it doesn't remove the need to check.