Generally any combination of a constant-bitrate low-latency voice codec with a cipher in XOR stream mode should work. If the codec does "comfort noise" you should disable that to keep the bitrate constant during silence.
I'm fairly certain Cell networks are not encrypted at all, by default. Or at least it's disabled completely by the towers in Afghanistan. :whistling:
You could look for resources that cover digital HAM radio operation. They should have some stuff about the basics of voice encryption. Most of it is not secure until you get to high-end stuff like Motorola AES 256. Some of this 'encryption' is just privacy codes (cell networks are not encrypted but use digital privacy codes I think).
Once you digitize the voice, then it should be pretty much regular encryption.