How I encode speech files using Opus

This saves about ~90% of space compared to Mono WAV files:

ffmpeg -i <input> -c:a libopus -vbr on -compression_level 10 -frame_duration 60 <output>

and will result in approximately 100kBit/s bitrate for many typical speech files.

In order to make it even smaller, use

ffmpeg -i <input> -c:a libopus -b:a 24K -frame_duration 60 <output>

This is still very high quality for typical speech (without too much music) and you will only notice differences in direct comparison. In a typical 1.1GB podcast file, it resulted in a filesize of 0.512% of the original WAV.