Better voice-to-text recognition needed. Maybe add the possibility to use other (cloud-based) apis for those people who do not care much for their privacy or similar.