Microsoft Files Patent Application for AI System That Alters Speaker’s Words Live

Must read

Microsoft has filed a patent application with the U.S. Patent and Trademark Office (USPTO) for an AI system that changes a speaker’s words in video recordings and live broadcasts in near real time. The system also matches the speaker’s lip movements to the altered words.

The application, published Oct. 29 and filed last year, describes a system in which an AI agent analyzes the audio from a recording and selects words to change based on set parameters and a viewer profile. Audio and video are then generated to carry out those changes, smoothing the transitions before and after the altered segment. The change happens in near real time, with a delay of 1 to 10 seconds, comparable to the delay used in live television broadcasts.

Publication of an application does not mean a patent has been granted or that a product is planned. According to the website Patentlyze, a USPTO examiner rejected the application in a first review, but that is common practice, and applicants typically respond by making amendments.

According to the Microsoft inventors, the system could have many uses, such as removing filler words, slips of the tongue and words deemed offensive. Which words would be considered inappropriate would depend on the viewer’s profile, including age, languages, background or religion.

Microsoft anticipates that the system could also be used for live translation. Examples of possible uses include diplomatic meetings, international conferences, podcasts, films and lectures.

The description of the invention also mentions “live correction of false or misleading information.” The system could check speakers’ claims against information online and update them “in line with current knowledge.”

The application does not mention the speaker’s consent to the changes, nor does it provide for informing viewers that the material has been modified. One of the claims provides only that the original recording be stored as a reference point.

Microsoft justifies the invention by saying that correcting speakers’ imperfections in video currently requires specialized tools and skilled technicians.

From Washington, Oskar Górzyński (PAP)

More articles

Latest article