Talking
I think the most cost effective way to make them feel more alive would be by letting them speak to each other, using speech bubbles with icons in them. There could be both emots and subject icons. To keep it looking nice and clear all the icons should probably be painted (so no pictures of specific voxel models) but the subject icons should be general enough so that there won’t have to be a lot of them. For example there just need to be one for monster (maybe like a black emot with red eyes and horns). Semi random combinations like one emot and two subject icons in the same speech bubble could quite easily make the player wonder what they’re actually saying (“Oh so you love eating monsters? That’s not creepy at all…”).
Another good think about talking is that they can do it all the time, not just when they’re idle, though idle hearthlings might walk over to someone to start a conversation.
As long as they’re close enough to another hearthling there is a chance that one of them will say something and once someone said something there is a higher chance that someone else will respond. Responding could be a “task” that is thrown around. Others might respond with emots at the same time, so a whole group could start laughing or get angry.
To make it look more like an ongoing conversation the response could repeat at least one of the former subject icons.
Shouting
Usually works the same as talking but with larger speech bubble and a smaller chance of happening.
It would be nice if they could shout warnings. If someone sees an enemy approaching town they shout a warning, which spreads from hearthling to hearthling and the guards won’t come running until they hear it. Could also be combined with warning bells.
Singing
Same as talking but they can do it on their own or at the same time as someone close by. Different songs could have different colored notes in the speech bubble. Hearing others sing makes them more likely to sing along and they might keep singing even if they walk away, so a song can spread through town.
Turning heads and hand gestures
On the animation side I think it would add a lot if they dynamically turned their heads (and maybe upper bodies a little bit) to look at points of interest. Not all the time and probably not at all while they’re working.
It would also be nice with animated hand gestures that match the emots in their speech bubbles. This would probably be limited to when they’re idle or walking, and only with hands that are not holding anything.