To get the best audio quality in your final lossy-eccoded audio file output at the smallest possible file size, you have to do 3 steps:
sample rate convert from 96kHz to 44.1kHz, because the lossy formats do not support 96kHz
dither the 24-bit file to 16-bits because the lossy formats do not support 24-bits, and if you don’t dither, all you are doing is chopping off the top 8 bits, which sounds terrible
encode the audio into MP4 AAC, not the much lower-quality MP3 that MP4 replaced — MP4 AAC playback is built into the hardware of every single audio-video player made since about 2002, so there is literally no reason to use MP3 — and a 160 kbits MP4 AAC will have equivalent sound to a 320 kbits MP3 that has double the file size so you get “the very best MP3 sound in half the file size,” or you can do MP4 AAC 320 kbits to get the ultimate lossy-encoded audio quality.
So you will go from 24/96 WAV to 16/44.1 MP4 AAC 160/320 kbits. That will be a dramatic file size reduction but a relatively small loss of audio quality for voice material.
The software you use will have to have this feature checklist:
- 96 kHz audio support
- sample rate conversion
- MP4 AAC encoding.
I do these processes in Logic Pro X, but that also includes many features like multitracking and software instruments that you probably don’t want, and it may not be available on your platform, so I can’t really give you a software recommendation. But generally speaking, you will want pro-oriented software to get the above features.
If you are not going to do the sample rate conversion and dithering, you will actually get better quality in your final 16-bit 44.1kHz file by simply recording in 16-bit 44.1kHz in the first place.