What is a 'natural sound' in audio?
The other day, I found myself advising someone that their recording of speech was good but it didn't sound natural. I further advised that the sound quality they had achieved was commonly heard on the radio, but it would be tiring to listen to for a long period, if the recording was part of an audio book for example.
It's worth reflecting for a moment on what we would consider naturalness, in a recording, to be. Fortunately we have examples of natural sound around us all the time, so there is plenty of material for comparison.
Perhaps the most useful natural sound is the human voice. Our ears are very closely attuned to the sound of the voice and we hear it all the time, and - most importantly - pay close attention to it. Of course I do mean the human voice as produced from a human larynx, throat and mouth, traveling directly through the air to your ears, not via a loudspeaker.
Let's consider therefore how we can compare the natural human voice with the sound of the voice reproduced via a loudspeaker.
Firstly, the person we choose to provide our hypothetical example of human speech should have a reasonably normal quality of voice. Professor Stephen Hawking writes excellent books on cosmology, but he isn't going to make a good example. Neither would a 40-a-day smoker. But we don't have to be too choosy. Apart from a few wayward examples, almost anyone would do.
Now we have to consider context. Should we consider the example of a friend spotting you from the other side of a busy road and shouting you a greeting? Well we could, but it doesn't have a lot of commonality with anything we would be likely to do in audio.
What about a lover whispering sweet nothings into your ear? That might be a desirable scenario, but the sound of the voice at extremely close range is difficult to mimic accurately. It's an interesting challenge, but we need something simpler.
So what about someone talking to you in a normal voice from a distance of two meters? That's just over six US feet.
This is a good test because it is a commonplace situation with which we are all very familiar. Also, it is practical to simulate with audio equipment. Bear in mind that most loudspeakers have at least two drive units and a certain amount of distance is required to allow the sound to integrate. A distance of one meter wouldn't be enough as small changes in listening position produce significant changes in perceived sound quality (and that is something to consider when using near-field monitors).
So imagine this... There is a visually opaque but acoustically transparent curtain in front of you, behind which there is a person, ready to speak from a prepared script (or you could do it in the dark). And also there is a loudspeaker, mounted with its central axis at the same height as your volunteer's mouth and as close as possible to one side. Through this loudspeaker will be played a recording that this same person made earlier. An assistant has previously checked that everything is working and that the levels are very similar.
So now you hear a voice. Is it human or is it the loudspeaker? Now you hear another voice. Or is it another voice? Is it perhaps the same sound source? Or has the source changed but there is so little difference that you can't tell?
You could carry out this experiment for real. Or you could consider it to be a test of naturalness in audio, and have this thought in your mind next time you need a recording to sound natural. Listen to your recording and ask yourself whether you would be fooled.
Although the human voice is the supreme test of naturalness in audio, it is also worth considering whether your recordings of acoustic instruments, including drums, sound natural. And if they don't sound natural, should you be trying to get closer to a natural sound, or are you trying to improve on nature?
Of course, naturalness isn't always the requirement. But it is a very useful benchmark of audio quality. Listen to your recordings closely and ask yourself which aspects don't sound natural. And whatever doesn't sound natural, ask yourself whether it is a defect, or an improvement.