A distributed dictation/transcription system is provided. The system provides a client station, dictation manager, and dictation server networked such that the dictation manager selects a dictation server to transcribe audio from the client station. The dictation manager selects one of a plurality of dictation servers based on conventional load balancing and on a determination of whether the user profile is already uploaded to a dictation server. While selecting a dictation server or uploading a profile, the client may begin dictating, which audio would be stored in a buffer of dictation manager until a dictation server was selected or available. The user may receive in real time or near real time a display of the textual data that may be corrected by the user to update the user profile.