who knew finding real-time TTS would be so hard

Replica has (very recently, as in yesterday) removed access to their API for creating runtime Text to Speech. Which is very annoying since they were the perfect candidate for voices that don't just sound like robots.

There's a couple other options out there, but all of them are equally annoying as hell to implement. Amazon and Google have systems, but boy do they look like a pain to set up for less-than-optimal results.

Oculus has a text to speech system, which would be great to use with their model lip syncing SDK as well. I'm trying to get that working, but naturally I'm getting errors that are getting in my way. I guess I'll keep working with it and see what I can get. I can't really tell if it'll give me the results I'm looking for (it seems like its mostly meant to be paired with their speech to text, for people using Oculus for VR. It's like an alternative for the STT, not really a standalone TTS system).

It seems like ReadSpeaker.AI has a pretty good system? They have a Unity SDK available for free demo, so I reached out to them about obtaining it. It can do real time text to speech, and the voices have some emotion and a bit of character to them. Not just robot voice, which is a big concern of mine.