Microsoft seeks patent for AI that alters live speech and syncs lips

Microsoft filed a patent application for AI technology that could alter speech during video broadcasts with a 1-to-10-second delay and modify the presenter’s mouth area to match the new audio. According to Tecnoblog, the system could translate speeches, remove stutters and mispronounced words, filter inappropriate terms and replace information during a presentation, including based on internet searches conducted during the broadcast.

The proposal raises authenticity concerns because it could create versions of a speech that the presenter did not originally say. According to Patentlyze, the claims received a rejection in the preliminary examination, but Microsoft can still revise the application. The filing does not indicate that the technology is tied to a specific product, and a patent does not mean it will be commercially released.