This is a well written article about the human voice, and how microphones really work in terms of capturing it. Even if you read this and plan to forget everything—you’ll come out *way* ahead when recording your and your guests’ voices.
It’s full of insightful information, about the human voice:
…the voice changes spectrum in almost any other position than when we approach the speaking person with our ear – or microphone. Each position on the head or the chest has its’ own sound color – or timbre. For instance, the spectrum of speech recorded on the chest of a person normally lacks frequencies in the important range of 2-4 kHz. This results in reduced speech intelligibility. If the microphone does not compensate for this you should make corrections with an equalizer.
The important frequencies in non-tonal (Western) languages are illustrated by the diagram below. Here, the frequency band around 2 kHz is the most important frequency range regarding perceived intelligibility. Most consonants are found in this frequency band.
…and about what affects intelligibility in a reproduction of the voice:
A lot of research has been carried out in this area. In general, the results demonstrate that:
1. Optimum speech level is constant when background noise level is lower than 40 dB(A)
2. Optimum speech level appears to be the level that maintains around 15 dB(A) of S/N ratio when the background noise level is more than 40 dB(A)
3. Listening difficulty increases as speech level increases under the condition where S/N ratio is good enough to keep intelligibility near perfect
Furthermore, the 1-4 kHz frequency range should be “kept clear”. When, for instance, adding music as background for narration, a parametric equalizer cutting the music 5-10 dB in this frequency range will improve intelligibility.