This is really great. I know it's not a new model, and it does often hallucinate, but it's really great frontier open-source voice AI models.