~/ARTIFICIAL I/microsoft-patents-ai-system-for-real-time-live-speech-modification-and-lip

Microsoft Patents AI System for Real-Time Live Speech Modification and Lip-Sync

Microsoft has filed a patent application for an AI system that can nearly-real-time modify spoken content in live broadcasts and redraw the speaker's lip movements to match the changes. The technology introduces a 1-to-10-second broadcast delay to filter out filler words, correct mispronunciations, or provide real-time translation. This development has significant implications for media broadcasting, censorship, and accessibility, enabling automated content moderation and seamless live translation. However, it also raises critical concerns regarding the expansion of real-time deepfake technology and viewer trust in media authenticity. The patent, numbered US 2026/0341309 A1, covers applications across lectures, award ceremonies, and sports broadcasts using devices ranging from AR glasses to mobile phones. Patent examiners initially rejected the claims in July 2026, which is a routine part of the application process that Microsoft can appeal.

## BACKGROUND

Real-time video and audio processing has traditionally suffered from latency barriers when performing complex generative AI manipulations like mouth redrawing. While voice changers and offline lip-sync tools are widespread, applying these transformations seamlessly on live streams while managing natural audio transitions remains an active frontier in AI research.

## REFERENCES

## KEYWORDS

#Artificial Intelligence#Generative AI#Video Processing#Microsoft#Patent

$ subscribe --daily