What frequency range should speech be recorded in?

The sounds produced in speech contain frequencies that lie within the frequency band of 100–10,000 Hz. However, the main useful voice frequencies are contained within the band 300 Hz to 3,400 Hz. The frequencies outside this narrower range contain harmonics which lend character to the sound of a voice.

Source

If you just need the voice (e.g., for transcription purposes), you can clip it to the telephone frequencies:
300 to 3400 Hz

This provides really awful fidelity (sounds like someone on a phone), but gets you clear voice.

The best way to figure out what your white noise is in your recording environment is going to be to record in a silent room (only the white noise), and then filter based on that. The problem with actual white noise is that it’s generally over the entire frequency range. It’s also likely that it’s being produced by your computer, or within your recording chain.

Much of the reason that good quality sound gear costs so much is the engineering involved to get rid of noise sources.

Futhermore, on the low end, in the US, you’re going to see noise at 60Hz and 120Hz, so cutting out below 200Hz will get you somewhere.

The above are good responses specific to the speech frequency range. If you are recording a male with a deep voice, you can open the lower range to 200Hz, and if you want clearer sybilants (unvoiced consonant sounds), open the high end between 5000–8000Hz. Experiment to see which is the best compromise for you.

I just wanted to add that professional studios also use dampers on the microphones and set the mix sensitivity down and have the person speak fairly close to the mic to minimize the amount of ambient noise that is picked up. You may also try reducing the ambient noise in the room by shutting off any unneccesary electronics and putting space between you and your CPU (or even a physical barrier, just don’t block the fans).

Nowadays, processors exist that can diatomfuis speech from non-speech… they’re available in high-end hearing aids, and I’d guess that professional audio engineers have similar tools.

@hearkat—I had forgotten about that in close-micing. The other thing that close micing gets you is a boost in the lower frequencies. It sounds totally different. There’s actually a neat moment in the Soul Coughing song “How Many Cans?” where the vocals shift from being far-mic’d to being close mic’d.

(I just want to apologize for the iPhone errors in my previous response. That’s what happens when I hastily type a response because my next patient is waiting for me!)

@hearkat where can I find one of those “diatomfuis”-ing processors?~

Another (low-tech) way to cut down on unwanted ambient noise, particularly echo and room noise, is to acoustically deaden the walls, using something like the large paper egg flats they get in restaurants.