Continua após a publicidade

From “live” to “edited”: Microsoft’s patent proposes AI that swaps spoken words and redraws speakers’ lips, allowing viewers in different locations to hear different versions of the same speech.

Image: Pixabay / Neva79

What You Need to Know

  • A Different Message for Every Viewer: Microsoft’s new patent describes an AI system capable of swapping out words during live broadcasts based on the audience’s profile—such as age, location, and language—in just 1 to 10 seconds.
  • Seamless Lip Syncing: Beyond modifying the audio track, the AI dynamically alters the speaker’s mouth movements to match the updated wording, making the edits virtually undetectable.
  • A Patent, Not a Finished Product: Published on October 29, 2026, the patent does not guarantee a commercial release. While Microsoft highlights practical uses such as audio cleanup and real-time translation, the technology also raises concerns about deepfakes, misinformation, and hyper-personalized filter bubbles.

Continua após a publicidade

Microsoft has filed a patent application for an AI system that can change what people say in live broadcasts within seconds. The technology could also modify a speaker’s lip movements to match the altered audio.

The application was revealed by patent tracker Patentlyze.

What the Filing Says

The inventors describe “near real-time AI enhancement of video or augmented reality presentations of live events,” according to patent application No. 20260341309. The examples cite delays of 1 to 10 seconds.

  • AI agents identify target words in the audio and either remove or replace them.
  • The system edits the audio and the video frames before and after the swap to create seamless transitions.
  • In some cases, the speaker’s lips are redrawn to match the new speech.

Every Viewer Gets Their Own Version

Under the first claim, substitutions would follow predefined rules or the viewer’s profile, including age, location, and native language. That means two audiences watching the same event could hear different speeches. The replacement words could come from an internet search.

Transparency

The filing notes that the original audio and video could be stored as “ground truth” for future reference. According to TechSpot, the first claim doesn’t specify whether viewers would be told about the changes.

What Would Be Edited

  • Filler sounds, such as “um” and “uh,” plus pauses and stutters
  • Profanity and offensive, insensitive, or inappropriate language
  • Personal and colloquial expressions, such as “y’all”

Real-Time Translation With Lip Sync

One independent claim covers live translation. YouTube already offers AI dubbing, but Microsoft goes further. It proposes translating speech and generating new lip movements synced to the audio in the viewer’s language. The original recording could also be retained for reference, according to TechSpot.

The application lists lectures, award shows, TV programs, sports, and concerts as potential uses. TechSpot also notes that the first claim could support instant fact-checking, correcting inaccurate or misleading statements during a broadcast.

What the Press Is Saying

The concerns below come from media reports, not from the patent application itself.

  • Lack of awareness: According to TechSpot, the ability to hide edits by adjusting lip movements suggests audiences could watch modified content without realizing it.
  • Fraud and disinformation: TechSpot points out that real-time AI face-swapping is already popular among scammers. Altering live speech could take disinformation and fraud “to new heights.”
  • Filter bubbles: Futurism notes that today, viewers who dislike what they hear can simply turn off the stream. This system would instead tweak other people’s speech to fit personal taste, edging us closer to personalized bubbles. “What’s so bad about hearing a Southerner say ‘y’all’?” the outlet asks.

Microsoft’s Response

According to TechSpot, Microsoft says the technology is meant for cleaning up and translating broadcasts. Futurism reports that the company did not immediately respond to a request for comment. According to Computerworld, it’s unclear when, or even if, the technology will launch.

An Industry-Wide Trend

TechSpot notes that this filing isn’t an isolated case:

  • In July, Microsoft filed a patent application for AI that rewrites a game’s code based on player behavior.
  • Sony has also filed patent applications for AI that predicts a player’s next frames and button presses to reduce latency.
  • Another Sony application would use AI to manage a game’s progression and difficulty, or to stand in for absent players.

What to Expect

A patent application is not a product. This one was published on October 29, 2026, and originally filed on April 29, 2025. There’s no guarantee it will ever launch.

Still, the filing shows the industry’s growing interest in real-time AI content editing. If “live” stops being truly live, how can transparency and trust be protected for audiences?


Sources & References

  • Patentlyze: Patent search database (link)
  • Justia: US App US 2026/0341309 A1 – Microsoft (link)
  • TechSpot: Live speech editing AI patent (link)
  • Futurism: Live speech alteration report (link)
  • Computerworld: Broadcast AI speech modification news (link)