Microsoft patents AI for modifying speech and faces of people in live broadcasts
The technology will be able to replace undesirable words, correct speech, and synchronously adjust the speaker's lip movements.
Microsoft has filed a patent application for a system that will use AI to edit the speech and appearance of people directly during broadcasts. Special AI agents will be able to remove or replace individual words while simultaneously correcting lip movements and facial expressions to make the changes appear natural.
Among the possible functions are translating speeches into the viewer's preferred language, removing filler words, correcting stuttering and pronunciation errors. The system will also be able to replace dialectal expressions and offensive language.
The technology is proposed for use during lectures, sports competitions, TV shows, and award ceremonies. However, the preservation of original recordings as reference data raises questions about potential misuse. Microsoft has not yet commented on the application.
