I think it needs to be continuous speech-to-text as well.