There has been plenty of discussion about the advantages and disadvantages of having the in-game chat be text-only. I would like to propose a feature to go along with the text chat: vocal chattering from characters when a message is sent in proximity-based character chat.

In essence, every time a character says something in a local chat, they would make vocal sounds. These could be synthesized based on actual voice samples and modified procedurally based on the text or even at random. The effect might be similar to Simlish in The Sims games, but even more random as it would be synthesized. There could be different "voice libraries" for when characters are speaking different languages, and these could be recorded by a few voice actors to provide a larger diversity in voices between characters beyond pitching up and down.

This system would add more feedback to indicate that a character is talking apart from just a speech bubble and the chat box, on top of contributing to an immersive environment by having some voices fill the soundscape. If the voices are distinct enough, you might even be able to tell which character is talking without looking! Voices for different characters could have pitch or other quantitative features adjusted based on size, gender, or other factors (with some player customization), and perhaps be based on a few different vocal profiles as mentioned above. Volume and range could also be varied based on different speech types (speaking, shouting, whispering). Together with an emote system, this could allow for plenty of variety in character interaction.

The hope would be that such a system would not be too expensive to design - it's definitely much cheaper than trying to record huge amounts of dialogue like some MMOs have done. Recording would need to be done for the possible phonemes (which could vary by language) and with a few different voice actors to give different vocal profiles. I'm not sure to what degree things like inflection could be simulated, so such variations may need to be recorded as well if they are to be implemented.

Many others had thoughts about this and similar systems in the discord chat, so I encourage you all to share your ideas.

EDIT: I forgot to mention that this concept could also be extended to the sensory map when/if it is implemented. Having the sensory map indicate sounds when a character is speaking would provide yet another source of feedback to the player.

EDIT 2: I also wish to give credit to BrightWard for urging me to make this thread. (So if you found the idea stupid, you know who to blame! :P)