"At the time, audio codecs were mostly divided into two categories: there were high-delay, high-fidelity transform codecs (like MP3, AAC, and Vorbis) that were unsuitable for real-time operation, and there were low-delay speech codecs (like AMR, Speex, and G.729) with limited audio quality."
I did not know this about audio codecs, but it makes a lot of sense, and this article is interesting, if not fascinating to anyone who is interested in audio codecs, sound engineering, and/or mathematical transforms...