How to test your microphone before recording an audiobook

The most common mic problem is not a bad mic, it is the wrong one. A laptop can silently default to its built-in microphone even when your good interface is plugged in, and you will not notice until you play back a chapter that sounds thin and roomy.

Open a mic test, look at which input device is selected, and speak. If the meter does not move, or it moves for the wrong device, switch inputs until the microphone you narrate with is the one lighting up the meter.

Your ears adjust to volume, so setting a level by feel is unreliable. Use a meter instead. Speak at the volume you actually narrate at, including your loudest emphasized lines, and set your gain so those peaks land with room to spare below the top of the meter.

Audiobook platforms have real targets worth knowing. ACX, for example, asks for peaks no higher than about -3 dB, an average level in a healthy middle range, and a quiet room noise floor. You do not need to hit those to the decimal here, but aim for a level that sits comfortably in range and leaves headroom, so a sudden loud line never clips.

If you record through a browser, or run a browser tool alongside your recorder, the browser applies automatic gain control (AGC) by default. AGC constantly nudges your input volume toward a target loudness. On a video call that is helpful. In a booth, where you set your level on purpose, it quietly undoes your work and makes your audio drift.

You can see it happen: with AGC on, go quiet and watch the level climb by itself. Chrome hides the switch under chrome://flags rather than its normal settings, so we wrote a short walkthrough: how to stop Chrome from adjusting your microphone volume. Turn it off, then run the test again and confirm your level holds still.

Noise suppression and echo cancellation exist to make calls intelligible, not to preserve a performance. They can pump, smear breaths, and thin out the low end of your voice, and once they have processed your take you cannot get the original back.

For narration, the better path is a raw signal into a quiet, treated space: switch both off, and fix the room instead of asking software to guess. If you must record somewhere noisy, do your cleanup later in your editor where you control it, not live where you cannot undo it.

Record maybe ten seconds of silence first. That is your room tone, and it tells you how quiet your space really is. A loud hum, a fridge, or traffic will sit under every line you record, so it is worth hearing before a full session, not after.

Then read a paragraph the way you will perform the book, including the loud beats. Listen back for anything harsh: clipping on peaks, popping p and b sounds hitting the mic, or a level that wanders. Fix those now, while it costs you two minutes instead of a re-record.

A good mic setup is wasted if your reading tool fights you: hand-scrolling a PDF pulls a hand off your performance and lets your place drift every time you stop to fix a line.

Narrate instead from a teleprompter that follows your voice and flags misreads live. That is what PromptVO was built for: it keeps your hands free and your place steady, so the level you carefully set at the top of the session is still doing its job at the end of it.