Short answer: A post on 8 November 2026 saying Microsoft had patented an AI that could “rewrite what people say during livestreams in real time” passed 1.5 million impressions. The underlying document is US patent application 2026/0341309 A1, published on 29 October 2026. It describes a system that picks words in a live video to change, replaces them in the audio and video, and redraws the speaker’s lips, in about 1 to 10 seconds. It is an application, not a granted patent and not a product, and coverage reports the examiner rejected the claims as filed on 29 July 2026, a routine step. Coverage also says the filing mentions keeping the original feed as “ground truth” but that the main claim says nothing about telling viewers what was changed. For speakers and hosts, the lasting point is to keep the original recording of anything you say live. A TimeBind dates the files you hold. It does not show that a viewer saw an edited version.
Last checked 13 November 2026. General information for speakers, event hosts, broadcasters and creators, not legal advice. A published patent application is not a product, and we do not say Microsoft will build or ship what it describes.
Table of Contents
What the filing and coverage say
The Justia listing for application 20260341309, titled “Near real-time artificial intelligence (AI) enhancement of video or augmented reality (AR) presentations of live events,” says a computing system identifies spoken words to modify from a live audio recording, based on a set of parameters or a profile of a user, and has a generative AI system replace them in near real time in both the video and the audio, with smooth transitions. It defines near real time as within a few seconds, giving 1 to 10 seconds as an example, tied to the delay used in live broadcasts to allow manual filtering of profanity. It also describes changing the lips of the speaker to match the replacement words, and a live translation use.
TechPowerUp reports the filing was submitted in April 2025 and published on 29 October 2026, and lists examples of what the system could replace: filler words, stutters, offensive words and colloquial terms. It says the original feed may be stored as “ground truth” but the main claim says nothing about telling viewers what was changed. TechPowerUp and T-Break both stress that Microsoft files many patents that never become products, and T-Break reports the examiner rejected the claims as filed on 29 July 2026. Dexerto and Futurism report the same filing.
We did not read the full claims ourselves beyond the published abstract and description, and we cannot say what Microsoft plans to do with it.
Why this matters to anyone who speaks or streams live
The filing is a patent application, but the reaction shows a real worry: if a live video can be altered before a viewer sees it, a clip alone stops being a reliable record of what was said. Lectures, talks, award shows and sports are named as examples. For speakers, that makes the original recording more valuable, not less.
The same worry applies to benign uses. Live translation with matched lip movement could help audiences, and editing out stutters is a form of cleanup. The records you want do not depend on whether an edit is well-meant.
Why people are searching for this
Measured on 13 November 2026. The @Pirat_Nation post (8 November) had 1,566,864 impressions, 16,084 likes Replies had 90,177 impressions and 10,208 likes (@UnPhilteredGuy), 79,539 impressions and 9,468 likes (@ThatRetiredDude), 56,420 impressions and 6,340 likes (@SamuelDeats) and 41,244 impressions and 4,785 likes (@cem_uk_). X’s trending summary ran on 9 November. Counts keep moving.
Checklist for speakers, event hosts and broadcasters
- Record your own copy at the source. Keep a recording from your own microphone or camera, not only the platform’s feed.
- Keep the run sheet and transcript. Save the script, run of show and any live captions, with the date.
- Keep the platform and settings used. Note the service, any translation or enhancement features that were on, and the version or date.
- Say it when you edit. If you clean up or translate a recording, keep the original, the edited file and a short note of what changed.
- Keep the original of anything you publish. Store the unedited master with its metadata beside each published version.
- Write down who had access. Record who could edit or re-stream the feed.
- Disclose changes to viewers where you can. A plain label such as “translated,” “captions auto-generated” or “edited” helps people trust what they see.
- Date your recordings and notes. Fingerprint the master and the transcript after the event.
What a TimeBind does and does not show
A TimeBind is a dated, checkable record that an exact file existed no later than a recorded time and still matches. The SHA-256 fingerprint is made in your browser and the file is never uploaded. The record is anchored to Bitcoin (a Merkle root in an OP_RETURN). Your first Pooled TimeBind is free; after that, TimeBinds start from US$5, and an ASAP TimeBind is US$15. See how it works.
For a talk or stream, it can date your master recording, transcript and run sheet as you hold them. It does not show what viewers saw or heard, does not show a stream was unedited, and does not detect AI. Use it as one dated record next to the original recording and your notes.
Limits
This page relies on the Justia listing for US 2026/0341309 A1, TechPowerUp, T-Break, Dexerto and Futurism as seen in search results, and the X posts named above. We did not read the full claims or the examiner’s action, and the rejection date comes from T-Break’s report. We cannot say whether the application will be granted, amended or abandoned, or whether any product exists. This is not legal advice.
Tools that can help
Use the bulk SHA-256 manifest to fingerprint a folder of masters and transcripts, and the Content Credentials checker to see whether a file carries provenance data. For related reading see deepfake evidence: proving your files are genuine and spot AI fakes.
Frequently Asked Questions
-
Has Microsoft patented AI that edits live speech?
Microsoft has a published patent application, US 2026/0341309 A1, published on 29 October 2026. Coverage reports the examiner rejected the claims as filed, so it is not a granted patent.
-
Is Microsoft building this as a product?
We found no confirmation. Coverage notes that companies file many patents that never become products.
-
Does the filing say viewers would be told about edits?
TechPowerUp and T-Break report that the main claim says nothing about telling viewers what was changed. The filing mentions keeping the original feed as ground truth.
-
What should a speaker keep of a live talk?
Your own recording from the source, the transcript and run sheet, the platform settings used, and the original of anything you publish.
-
Can a TimeBind show a stream was not altered?
No. It shows that exact files existed no later than a recorded time and still match. It does not show what viewers saw or heard.
Sources checked
- Justia Patents, US patent application 20260341309, “Near real-time artificial intelligence (AI) enhancement of video or augmented reality (AR) presentations of live events,” https://patents.justia.com/patent/20260341309 (seen in search results 13 November 2026)
- TechPowerUp, “Microsoft Patent Application Describes AI That Edits Live Speech,” https://www.techpowerup.com/354585/microsoft-patent-application-describes-ai-that-edits-live-speech (seen in search results 13 November 2026)
- T-Break, “Microsoft AI patent rewrites live speech and redraws lips,” https://tbreak.com/microsoft-ai-patent-live-speech-lips/ (seen in search results 13 November 2026)
- Dexerto, “New Microsoft patent could use AI to alter speech and lip movements in real time,” and Futurism, “Microsoft Patents an AI That Can Literally Put Words in People’s Mouths in Real Time” (seen in search results 13 November 2026)
- X posts by @Pirat_Nation, @UnPhilteredGuy, @ThatRetiredDude, @SamuelDeats and @cem_uk_ (8 November 2026) and X’s trending summary of 9 November 2026, read 13 November 2026
Date your master recording and transcript after each event
Your first Pooled TimeBind is free, with no credit card required. After that, TimeBinds start from US$5, and an ASAP TimeBind is US$15. Click here to start, or see plans and pricing.
