sadly not currently, you really need a end to end speech to speech model for that