Hacker News Reader

I'd imagine they down-sample the audio like crazy. They certainly don't need CD quality sound for it.

the standard is 8kHz-16kHz, 8-bit PCM. higher frequencies are of limited utility in recognizing human speech.