Back to news
researchMicrosoft2026-11-09

Microsoft disclosed a patent application on November 9 for near real time AI editing of live speech with synchronized lip movement

The patent, numbered US 2026/0341309 A1, describes an agent that listens to audio in real time and flags words for modification under preset rules or viewer profiles, with generative AI replacing the words and redrawing the speaker lip motion, with processed video playing about 1 to 10 seconds behind

On November 9, Microsoft disclosed a recent patent application for using artificial intelligence to modify what people say during live broadcasts and other real time video programs in near real time, together with redrawing the speaker lip movements so viewers have difficulty noticing that the original words were replaced.

The patent is numbered US 2026/0341309 A1 and is titled near real time artificial intelligence enhancement of video or augmented reality presentations of live events. According to the filing, the system includes an agent that listens to audio in real time and identifies words to modify or remove according to preset rules or viewer profiles including age, spoken language and place of residence. Generative AI then replaces those words in the audio, smooths the seams around each edit, and generates new video frames so the speaker mouth movements stay consistent with the modified audio.

Content that could be replaced includes filler words, stutters, mispronounced words, offensive vocabulary and certain colloquial expressions. Processed video plays about 1 to 10 seconds behind to allow the audio and video edits to complete. The patent also covers real time translation and could run on augmented reality glasses, phones, tablets or cameras. It should be noted this is only a published application, examiners rejected some of the claims on July 29, 2026, and it was filed on April 29, 2025, with Microsoft not saying whether it plans to build the technology.

微软专利实时翻译直播AR