Pretty much all communications-grade digital voice protocols sound awful, for the same reason that DMR sounds awful. Probably the least awful-sounding one, from the samples I’ve heard, is NXDN. Even that can’t hold a candle to plain old analog FM, however.
It all leaves me wondering if digital is just plain the wrong answer to cramming a voice signal into a smaller bandwidth. I’m inclined to thing using something like SSB plus an intelligent, software-defined receiver (with sophisticated noise filtering and precise, automatic carrier insertion) might be better.
Noise tends to be not a super-big issue at VHF and above, anyhow, so losing a degree of immunity to it might not be such a big deal, particularly if one uses digital signal processing techniques to remove as much of it as possible. SSB can sound as good as AM (which, unless noise gets in the way, blows digital out of the water when it comes to audio fidelity and overall intelligibility), provided you insert the carrier at just the right spot when receiving.
As a further plus, a voice-quality SSB signal uses half or less the bandwidth per channel of any of the digital protocols currently out there.
Sometimes the right answer to a question about employing a new technology is not to employ it at all (or, in this case, to employ it, but in a significantly different way than originally proposed).