The Patent That Could Alter and Personalize Live Events
Updated
Thanks to AI, we are approaching a point where a live video can show different people different versions of what is being said, with the ability to adjust the speaker’s lips to match each different version. The idea, as detailed in a patent filed by Microsoft and published on October 29, 2026, is to automatically clean up live speech, remove filler words or language that might be offensive to some, fix unclear delivery, and translate what is being said into a viewer’s own language, all while looking and sounding completely natural. What could possibly go wrong? Still just a filing and not yet a product, the details are specific enough to matter right now. The document, numbered US 2026/0341309 and titled “Near Real-Time Artificial Intelligence (AI) Enhancement of Video or Augmented Reality (AR) Presentations of Live Events,” was filed by Microsoft Technology Licensing and lists inventors David Callaghan and Dinesh Kumar.
The system operates by capturing live video and audio, then, almost right as it happens, an AI agent scans the speech within one to ten seconds of the words leaving the speaker’s mouth. The agent then selects specific words or phrases to change based on how it is programmed, often depending on the profile of the viewer, which, they claim, would be based on things like age, location, and the languages they speak. After the AI system picks the unwanted words, it replaces them in the audio, blends the cuts so they don’t sound unnatural, and generates new video, making the speaker’s mouth match the new words. The viewer sees what appears to be an ordinary stream, with no indication that anything has changed. Indeed, the filing notes that the word and facial movement swaps would occur in “near real-time” because the program would work inside the current few-second delay that broadcasters have long used to catch and mute profanity. Look out. Soon, lectures, award shows, live sporting events, and television shows could be unknowingly altered, with the ability to operate on phones, tablets, cameras, and AR glasses, to name a few.
Most of the examples of use in the application are, on their face, modest. Things like trimming out someone saying “um” or mumbling or mispronouncing a word. Likewise, filtering out a word or phrase the system flags as offensive or translating a speech into another language and redrawing the lips to fit the translation. The slightly scary part of the filing does not call for a notice, watermark, or any kind of notification that the speech and face of the speaker have been altered. Some have suggested the option of keeping the original recording on file, or embedding it so it can be checked later, but those are add-on features and are not part of what makes it operational. When stripped down to its essential function, the outcome is simple. Two viewers with two different versions of the same live moment, each with facial expressions that match, and neither has been told of the swap. It seems a safe bet that those seeking control spot the nefarious possibilities.
The invisible swap is what catapults Microsoft’s proposed technology leaps and bounds beyond the ordinary video editing that has occurred for years. Why? When a producer bleeps a word, everyone watching and listening hears and sees the same gap. But in this AI-driven reality, the change is undetectable and individually tailored all the way down to each viewer. In this scenario, even if the original video was saved somewhere in a server, one neighbor’s screen could differ from another, and there would be nothing to point out or flag the difference.
For years, live video has been relied upon as a fairly consistent record of public moments. During those years, we have trusted the interview, the speech, the hearing, the game, the lecture, and so on, largely because we believe that the camera is showing us the real deal. Microsoft’s proposed product will put an end to that, pushing society into altered realities. Remarks that get softened for one crowd and left sharp for another (or swapped out completely) will leave viewers unable to tell the real deal from the altered version crafted especially for them.
Along with the concerns mentioned above, the swift pace of AI is itself unsettling. Software that can generate convincing talking faces from audio recordings already exists. IN 2024, Microsoft itself highlighted a research project called VASA-1, which produced talking-face video from an audio clip and a single still image. Sure, that work was separate from Microsoft’s current patent filing, that sketches a product capable of targeting a running live feed, altering it, and delivering it a few short seconds. In other words, what should concern everyone is not that the delay is shrinking. It is that the alteration of what we believe to be a live event—whatever it may be—now fits inside a delay short enough to make it look and feel like the original. And no one will be the wiser.