Before I lose my train of thought, for text-to-speech, the mannerisms for speaking might be best serviced with a music generation system. My reasoning behind this would be it is not necessarily music but it has a certain meter, pitch variation and so on. If it were to come down to an AI coming up with something to say, mood can best be expressed by a variation in tone in much the way as something is sing-song. However, a certain clamp would be implied so that it isn't over the top, but at the same time breathe a little bit of life into the character.
I had something else but honestly couldn't tell you where I was going with the thought.
Powered by Qumana
No comments:
Post a Comment