How to Record Better Phone Audio for a Cleaner Transcript

No transcription tool recovers words that were never captured. The recording sets a ceiling, and everything afterwards works underneath it.

The good news is that most of what improves a phone recording is free and takes no equipment.

Distance beats everything

If you change one thing, change this. Halving the distance between the microphone and the speaker's mouth does more than any app, setting or tool.

A phone flat on the table between two people is good. A phone in a shirt pocket is poor. A phone across a room is usually unrecoverable for anything but the loudest speaker.

The reason is that speech falls off sharply with distance while room noise does not, so every extra metre worsens the ratio twice over.

Kill the constant noise

Steady background noise is more damaging than occasional loud noise, because it sits underneath every word rather than covering a few.

The usual culprits, in order of how often they ruin a recording here: generators, air conditioning, ceiling fans, traffic, and wind. Moving a metre away from a fan or closing a window is worth more than any amount of post-processing.

Outdoors, wind is the one to take seriously — it overwhelms a bare phone microphone completely. Shelter the phone with your body if you cannot move.

Handle the phone as little as possible

Fabric and fingers against the microphone produce a rumble that lands directly on top of speech. Put the phone down on a stable surface, screen up, and leave it there.

Avoid resting it on something that transmits knocks — a table people lean on, or a surface where a laptop fan is running.

Silence notifications first

A message tone lands straight on the microphone and, reliably, mid-answer. Put the phone in do-not-disturb before you start rather than after the first interruption.

Ask for one voice at a time

Overlapping speech is the hardest problem in transcription, in every language and with every tool. A short word at the start of a meeting or a focus group — asking people to let each other finish — improves the transcript more than any technical measure.

For groups, it also helps enormously if the facilitator names people as they speak.

Say the metadata aloud

Date, place, who is present. Eight seconds at the top of the recording. It gives you a provenance record, a filename, and a clean reference for names and speaker labels later.

Record one continuous file

Starting and stopping produces several files that have to be reassembled and breaks the consistency of speaker labelling across the conversation. If you can, record straight through and trim afterwards with something like the audio cutter.

Pick the better format if offered

Some Android recorders default to AMR, built for phone calls at tiny file sizes, and it audibly softens consonants. If your recorder lets you choose, use M4A or WAV.

And never re-record: forwarding an original voice note always beats playing it aloud and capturing it on a second phone.

If it already went wrong

Some of it is fixable. Cleaning up muddy or noisy audio covers the repairs that genuinely help, and which problems cannot be undone.

A free account includes transcription minutes each month with no card. Sign up and compare a carefully recorded conversation with a careless one — the difference is usually larger than people expect.