Images and video · 2026-11-03

Microsoft patents AI that can change words and facial expressions in live streams

Microsoft has filed a patent application for technology that uses AI to replace a speaker's words and adjust their face during a live broadcast.

✨ What Microsoft has come up with

Microsoft has filed a patent application for an AI-based tool that can edit video broadcasts while they are happening. The system could replace or remove a speaker's words and at the same time adjust their facial expressions and lip movements on screen.

According to the inventors, this is an "enhancement of video or presentations in augmented reality (AR) format during live broadcasts using artificial intelligence in near real time." Futurism reported this, citing the patent tracking service Patentlyze.

📌 How it would work

According to the documentation, dedicated AI agents would identify unwanted fragments and correct the image of the speaker's face and mouth. The model would edit audio and video right before and after the changed section to make the transition smooth and free of visible splices.

The basic use cases listed include simultaneous translation into other languages, removing filler words, and correcting stuttering and speech impediments.

🔹 Why this raises questions

The described functions go much further in altering content. The system could cut colloquial expressions, dialect words, rude or offensive words, and mispronounced words from a live feed. It would look for replacement wording on the internet.

According to the text, the authors have considered the risk of disinformation and distortion of facts. The patent says the original recording of the event would be kept to ensure "objectivity." The technology is expected to be used in online lectures, corporate presentations, sports broadcasts, and TV shows.

It is worth remembering that this is a patent application. It does not mean the product already exists or will soon be on sale.

Sources: 🔗 DOU/dev.ua