A laptop's built-in microphone can produce a perfectly usable voice recording: a lecture note, a podcast draft, an answer for a job application or a message for a colleague. Most bad recordings are not the microphone's fault. They are too far away, too quiet, recorded in an echoing room or squashed by processing meant for video calls. This guide goes through the few decisions that matter, with numbers you can check on the level meter of the voice recorder before you record anything important.
Distance matters more than the microphone
Sound from your mouth spreads out as it travels. For a small source in open space, every doubling of distance lowers its level by about 6 dB (the inverse-square law). The noise of the room does not drop the same way, because it comes from everywhere. So moving closer is the cheapest way to make your voice louder compared with the fan, the traffic and the echo.
A laptop on a desk typically puts its microphones half a metre or more from your mouth. Leaning in to half that distance: about 6 dB more voice for the same noise. An inexpensive headset or a clip-on microphone at 5 to 10 cm gains far more than any software can. If you use one, keep it slightly to the side of your mouth so that the bursts of air from "p" and "b" do not hit it straight on.
The room: soft and small beats big and bare
Echo is what makes a recording sound like it was made in a bathroom. Hard, parallel surfaces (bare walls, a tiled floor, a glass window) bounce sound back; soft, irregular things (curtains, a sofa, a bed, a full bookshelf, clothes in an open wardrobe) absorb or scatter it. Recording in a bedroom usually beats a large empty living room. Facing into the room, with soft furniture behind the laptop, beats facing a bare wall from close up.
Before a long take, record ten seconds of silence and watch the meter. If the room alone keeps the level above about -50 dBFS, that hiss or hum will be audible under your voice. Switch off fans and air conditioning for the take, close the window, and silence notifications. A laptop's own fan is often the loudest thing in the room, and it spins up when many browser tabs are open.
Reading the level meter
Digital audio is measured in dBFS, decibels relative to full scale. 0 dBFS is the loudest value a file can hold; anything louder is clipped, and clipping sounds like crackle that cannot be removed later. Every halving of the amplitude is about -6 dB, so a peak at half of full scale is -6 dBFS and a peak at a tenth is -20 dBFS.
The voice recorder shows the peak of the last fraction of a second in dBFS and in words. Its thresholds are rules of thumb, not a standard:
| Peak level | What the meter says | What to do |
|---|---|---|
| below -50 dBFS | no sound | check the microphone choice and permission |
| -50 to -30 dBFS | too quiet | move closer or speak up |
| -30 to -1 dBFS | good | aim for loud words peaking around -12 to -6 |
| above -1 dBFS | too loud, may clip | move back a little |
A quiet recording can be made louder afterwards, but its noise comes up with it. A clipped one cannot be repaired. When in doubt, err on the quiet side of "good".
Browser clean-up: when to switch it off
Browsers can apply three kinds of processing to a microphone, and in Chrome all three are on unless a page asks otherwise, because they were built for calls:
- Echo cancellation removes sound that comes out of your own speakers. Useful if anything plays during the take; harmless otherwise.
- Noise suppression removes steady background noise such as a fan. It works well on speech, but it can make music, singing or laughter sound watery and clipped at the edges.
- Automatic gain constantly adjusts the volume. It rescues someone who drifts away from the microphone, but it also pumps the room noise up in every pause.
For a spoken memo with a laptop microphone, leave all three on. For music, or for a good external microphone in a quiet room, switch noise suppression and automatic gain off and set your distance carefully instead. The recorder lets you change all three before you start.
Which file to keep: a worked example
Say you record a 10-minute interview. The recorder asks the browser for Opus (or AAC on Safari) at 128 kbit/s:
- Compressed: 128,000 bits × 600 seconds ÷ 8 = 9,600,000 bytes, about 9.6 MB.
- WAV, 16-bit, 48 kHz, mono: 48,000 × 2 bytes × 600 seconds + a 44-byte header = 57,600,044 bytes, about 57.6 MB, six times larger.
- The same WAV in stereo is twice that again, about 115 MB.
Opus at 128 kbit/s is far beyond what speech needs; at that rate you will not hear a difference from the WAV. Keep the original for listening, sharing and transcription apps that accept WebM, Ogg or M4A. Convert to WAV only when an editor, a submission form or an older program refuses the original. Converting cannot add back detail that compression removed; it only makes the file easier to open.
Common mistakes
- Recording from the wrong microphone. A webcam, a monitor or a disconnected headset may be the system default. Pick the device by name; browsers only show names after you have allowed the microphone once.
- Judging by ear, not by the meter. Your headphones' volume tells you nothing about the file. Watch the dBFS reading.
- Leaving call processing on for music. Noise suppression treats sustained notes as noise.
- Typing during the take. A laptop microphone sits next to the keyboard and picks up every key press loudly.
- Closing the tab before downloading. The recording lives in the page's memory; it is gone when the tab closes.
- Sending a 100 MB WAV by e-mail. Send the compressed original unless WAV was asked for.
After the recording
Trim the silence and the mouse clicks at the start and end with the audio trimmer, which exports WAV and can fade in and out. If you are recording a script, hearing it read aloud first with text to speech is a quick way to catch awkward sentences before you say them. And when your audio is part of a tutorial, the screen recorder can add your microphone over a tab or window.