Thanks for the feedback. We only support English now. While the underlying model generalizes to speak Chinese, but the whole pipeline is optimized for English and specific English speakers.