How were you recording the voices? Do they know they were being recorded? Were you using your phone?
Someone else may post with solutions, but I think that’s hopeless. Noise Reduction only works on noises that don’t change, so those popping and scraping noises are permanent. Also, you need background noise or Room Tone by itself as a sample. If the voices are so bad you can’t tell when they speak, those tools won’t work.
The sound sample has compression distortion and environment suppression distortion (wine glass voice). Both of them are permanent.
I know nobody’s going to do it his way any more, but when I did it, I made a wooden backing for high quality, all-direction microphone.
I painted it black and put it on a heavy black blanket in the middle of the meeting table. I plugged it into a tiny sound mixer and into a Linux machine with Audacity. I set the sound mixer for moderate volume and to reject low-pitch sound and rumble. We had to share meetings with all the divisions of the company in all the time zones and it nearly always worked.