Why a voice recording does not need to be large
A WAV file stores every sample in full: an hour of CD-quality stereo is over 600 MB. Dictaphones, podcast recorders and some phone apps save exactly that, and even an MP3 exported at 320 kbps is about 140 MB per hour. For music that headroom can make sense. For one person speaking it is almost all waste, because speech uses a narrow band of frequencies and one channel carries it completely.
The Speech setting reflects that: 64 kbps, mono. An hour of lecture becomes roughly 29 MB and remains perfectly clear, small enough for email, messengers and course platforms with strict upload limits. Balanced keeps stereo at 96 kbps for interviews recorded with two microphones or podcasts with a music bed; Music at 128 kbps is the floor for songs, where lower rates start to sound thin.