Microsoft patent outlines AI system for rewriting livestream speech
The proposed technology would tailor audio to viewer profiles and adjust lip movements, while a separate claim covers live translation.
Microsoft has filed a patent describing an AI system that would change words spoken during livestreams according to viewer information. The proposal, uncovered by Patentlyze, also describes altering lip movements to match the replacement speech.
One claim outlines a process for detecting particular words in broadcast audio and substituting others using profile details, including the viewer's location, age and native language. Microsoft says the technology would eliminate fillers and pauses. It could also filter profanity or use internet search results to inform word replacements.
The claim does not specify whether viewers would be told what had changed or the reasons for an edit. A separate translation claim describes matching lip movements to translated livestream audio. That version would preserve the original recording as a reference.
A patent does not necessarily become a product. Microsoft also filed a claim in July describing AI that would modify game code, adding content in response to how someone plays.
