In This Article
Microsoft has published a patent application describing AI that could modify speech in live video and adjust a speaker’s mouth movements to match. Proposed uses include translation, profanity filtering and correcting verbal slips. The filing outlines a possible technology, not a confirmed product launch, and raises questions about how audiences can verify what was originally said.
Key takeaways
The patent proposes near-real-time changes to audio and video, potentially tailored to viewer preferences. Its applications range from accessibility and translation to content filtering, but realistic edits could complicate trust in broadcasts and other recorded evidence.
- AI could replace or remove selected words and adjust corresponding facial movements.
- The application describes storing original footage as a reference record.
- The filing does not confirm that Microsoft plans to release the technology.
How the proposed system would work
The application, titled “Near Real-Time Artificial Intelligence (AI) Enhancement of Video or Augmented Reality (AR) Presentations of Live Events,” describes AI analyzing a live audio-video feed, identifying words for modification and generating replacements. It would then alter the audio and relevant video frames so the speaker’s lips appear to match the revised speech.
Potential uses include translating a speaker into a viewer’s language, removing filler words, correcting mispronunciations, replacing slang and filtering profanity. The patent also discusses using preferences or profiles to guide changes. It gives a one-to-10-second window as an example of near-real-time processing; that figure is not a demonstrated performance benchmark.
A patent application, not a product announcement
The application was filed on April 29, 2025, and published on October 29, 2026, under publication number US 2026/0341309 A1. A patent filing describes an invention for consideration and protection; it does not establish that the technology is built, commercially available or scheduled for release.
The proposed system also describes keeping original audio and video as “ground truth.” That archive could help establish what was said, but its value would depend on secure preservation and whether people reviewing an edited version can readily identify and access the unaltered record.
Why authenticity and governance matter
Changing a speaker’s words while making their face appear to confirm those words could make modified footage difficult to distinguish from the original. Personalized edits also raise the possibility that different viewers could receive different versions of the same event. Clear disclosure, consent, reliable provenance and auditable originals would therefore be central considerations for any real-world deployment.
For enterprise leaders, the broader lesson is that AI-generated outputs need traceable inputs and accountable governance. Spargent helps organizations build trusted data foundations with Microsoft Fabric and Power BI; similar discipline around lineage, access and auditability is important whenever AI transformations may influence decisions or public understanding. The patent does not say those Microsoft products are part of the proposed system.
- Microsoft files patent for AI that can alter what people say during live broadcasts – Computerworld, Computerworld.
- Microsoft files patent for AI that can change what people say in live video, Yahoo Tech.
- Microsoft Patents an AI That Can Literally Put Words in People’s Mouths in Real Time – Futurism, Sites at Penn State.
- Microsoft’s AI Patent Could Put Words in Your Mouth Mid-Sentence, Then Make Your Lips Match, International Business Times UK.
- Microsoft Patents AI That Can Change Your Speech And Face In Real Time, Times Now.