When people keep asking you to repeat yourself

Being asked to say that again is not a volume problem, and the usual answer makes it worse. Past a measurable point, talking louder costs you words. Here is what carries them instead, and the four changes that put them back.

Louder stops helping, and then it hurts

That result is from the Journal of the Acoustical Society of America in 1956, and the shape of it is the useful part: pushing harder buys you almost nothing across the entire normal range, and past the top of that range it starts taking words away. Most people asked to repeat themselves push harder, which is the one move the measurement says will not work.

What actually carries the words

The same thing, in 46 seconds.

Your voice is made of vowels and consonants, and they do different jobs. Vowels carry most of the energy. Consonants, the t and s and k sounds, carry most of the meaning: they are what separates cat from cap from cab.

When you shout, nearly all the extra effort lands on the vowels. A study at Aalto University and the University of Helsinki measured exactly this across 22 speakers in an anechoic room, and reported that the vowel-to-consonant ratio rises sharply in shouted speech while articulation gets less precise. The loud part of your voice gets louder and the informative part does not keep up.

So you end up louder and less clear at once, which is a frustrating combination to be on either end of.

Fix 1: get closer instead of louder

Distance does what volume cannot. Every doubling of the distance between your mouth and the microphone drops your voice relative to the room, so a mic at arm's length is mostly recording your walls. Halving the distance is the same arithmetic in your favor, and it costs you nothing in articulation.

Aim for roughly 10 to 20 centimeters, slightly off to the side so your breath does not hit the capsule. If your level is still low after that, mic too quiet walks through the gain staging.

Fix 2: lift the range the consonants live in

Consonants sit higher than the body of your voice, roughly 2 to 5 kHz, and that range is also the one your own ear is most sensitive to. A modest EQ boost there makes you easier to follow at the level you already speak at.

A few decibels is enough. Push it further and you cross into harsh, and then the S sounds start to sting, which is a different complaint with its own page in harsh S sounds. If the top of your voice is missing entirely rather than just quiet, mic sounds muffled is the page for that.

Fix 3: even out your level so the quiet parts survive

A compressor narrows the gap between your loudest and quietest moments, then lifts the result. The consonants that were sitting underneath your vowels come up with everything else, and the vowels that were spiking stop arriving far above them.

There is a striking piece of evidence for this. Bell Telephone Laboratories studied shouted speech for NATO in 1969, because crash investigators kept trying to decipher pilots shouting over noisy radio. Shouted speech scored 85.2 percent intelligible in their test. The same shouted speech, peak clipped, scored 97 percent, against 98.25 for normal unclipped speech. Their explanation was that clipping restores the vowel-to-consonant ratio to something closer to normal.

A compressor is the civilised version of that: it pulls the vowels down rather than squaring them off, and it gets at the same ratio.

Fix 4: take the noise out from under you

If a listener is picking your words out of fan hum and keyboard clatter, the consonants are the first thing to get lost, because they are the quietest part of what you say. Noise suppression removes steady noise while you talk, and a gate closes the gaps between your sentences. The two work as a pair, in that order, which is covered in noise gate vs noise suppression and step by step in reduce background noise on your mic.

Putting it together

The four fixes are one chain, and the order is the one in effect chain order: clean the noise, lift the clarity range, then even the level. In Patchd each of those is a free node you drop onto your mic channel. The full broadcast version of it, with the reasoning for each position, is in the radio voice chain.

The takeaway

Talking louder buys you almost nothing between a moderately low voice and a very loud one, and past about 78 decibels at a meter it costs you words, because the extra effort goes into vowels while the consonants that carry meaning fall behind. Get closer, lift 2 to 5 kHz by a few decibels, even your level with a compressor, and take the noise out from under your voice. Four changes, none of them louder.

Patchd is a Windows audio mixer with that whole chain free on every channel. It is in open beta now, and if you clear the terms your invitation arrives straight away.

Stop fighting your audio.

Whether you are producing, podcasting, streaming, or gaming, Patchd is the audio mixer your setup deserves. It is in open beta now, and the waitlist is how you hear about the public release.