Take thousands of hours of different people's voices and add them up over a span of one minute. Normalize if necessary; otherwise just clip/compress.

or would it just sound like white noise? i'm not sure.