How to Record Clean Audio: Level Meter, Interruptions and Saving Takes

The problem

You record a whole lecture and discover afterwards that the microphone was covered, or the phone went to sleep, or you can’t tell whether the quiet bit was silence or a failure.

What you see while recording

• An elapsed time and a live level meter. It reads “No sound”, “Picking up sound, quietly” or “Picking up sound clearly”.
• The live transcript, if you’ve left it on. Tap a line’s timestamp to jump playback to that sentence later.
• A Jump to latest button if you’ve scrolled back in the live text.

Keep, discard, save

• The recording saves itself when you stop, with a generated title built from your title format, and the last category you used.
• Discard removes the audio and transcript of the take after confirmation.
• Starting a new recording while one is unsaved asks whether to keep or discard it.

Format and size

SpeakSense records mono 16 kHz, 16-bit audio, which is what speech recognition wants. That is about 2 MB per minute, or about 115 MB per hour. Use Settings → Storage to see what recordings take up.

Tips

• Check the meter during the first ten seconds, not after the meeting.
• Put the phone on the table with the microphone toward the speaker.
• Mute notifications. Calls and alerts can interrupt a recording.
• Allow microphone access in Settings if the app says recording is off.

FAQ

What audio format is recorded? Mono, 16 kHz, 16-bit audio in a .caf file, which is ideal for speech recognition. It works out at roughly 2 MB per minute.

Does it keep recording with the screen off? Yes. The app has background audio.

Can I throw away a take? Yes. Discard asks you to confirm that the audio and transcript will be deleted.

Can I change language mid-recording? No. The transcription language is fixed while recording.