Microsoft Technology Licensing, LLC

Jan Pavlovsky

Adam Czeisler

Described herein is a technique for securely synthesizing a voice of a speaker during a language translated voice call. When a voice call is first initiated between the speaker and one or more other call participants, the audio data representing the speech of the speaker is received at a server computer where it is processed by obtaining a sample of a fixed duration (e.g., 8 seconds). This fixed-duration sample is then processed to generate a voice profile of the speaker for use in generating synthesized speech in a voice of the speaker, in a second language. This process of sampling the audio data and generating the voice profile is repeated at a fixed interval (e.g., every 30 seconds), such that the voice profile of the speaker is continuously updated during the voice call.

Growing community of inventors

Prague, Czechia

Jan Pavlovsky