Microsoft has filed a patent application describing an artificial intelligence system that could modify what people say during live video broadcasts, replace selected words and even adjust their lip movements to match the edited speech. The proposed technology could work in near real time, allowing viewers to watch a modified version of a live event rather than simply receiving subtitles or translated audio.
The application, published on October 29, 2026, describes a system that could remove filler words, correct mispronunciations, filter offensive language and translate speech into a viewer’s preferred language. However, the technology also raises questions about the authenticity of live broadcasts because different viewers could potentially receive different versions of the same speaker’s words. Importantly, this is a patent application, not a confirmed Microsoft product or an announcement of a commercial launch. <Cite refs={[“turn658331search2″,”turn658331search5″,”turn658331search4”]}/>
How Microsoft’s AI Broadcast Editing Technology Would Work
The patent application, titled “Near Real-Time Artificial Intelligence (AI) Enhancement of Video or Augmented Reality (AR) Presentations of Live Events”, describes a system that combines speech recognition, generative AI and video editing.
The technology would continuously analyse the audio and video of a person speaking. An AI agent would identify words or phrases that need to be modified based on predefined rules and information contained in a viewer’s profile.
A generative AI system would then replace the selected words in the audio and modify the corresponding video frames.
The system would also adjust the speaker’s mouth movements so that the edited speech appears to match the person’s lips.
The proposed process involves several steps:
- Capture live audio and video: The system receives a live broadcast through cameras and microphones.
- Identify words for modification: An AI agent analyses the speech and identifies words that match predefined editing rules.
- Generate replacement speech: Generative AI replaces the selected words with alternative wording.
- Modify video frames: The system adjusts the speaker’s mouth movements and surrounding frames to match the new audio.
- Deliver the edited broadcast: Viewers receive the modified audio and video with smooth transitions.
The application describes near-real-time processing, with reporting on the patent indicating a potential delay of approximately one to 10 seconds. This is a proposed capability rather than a demonstrated performance benchmark. <Cite refs={[“turn658331search2″,”turn658331search5″,”turn658331search8”]}/>
AI Could Remove Filler Words, Stutters and Offensive Language
One of the proposed applications is improving the quality of live presentations by removing unnecessary speech patterns.
The patent describes modifying several types of spoken content, including:
- Filler words such as “um” and “ah”.
- Stutters and repeated words.
- Mispronounced words.
- Mumbled or unclear speech.
- Colloquial expressions and regional phrases.
- Offensive or inappropriate language.
- Selected words that do not match predefined content preferences.
For example, someone watching a live business presentation could receive a version in which repeated words and unnecessary pauses have been removed.
Similarly, a viewer could potentially configure the system to filter certain expressions or offensive language.
The technology could also replace selected phrases with alternative wording generated using information from internet searches.
This goes beyond conventional content moderation. Instead of muting a word or displaying a warning, the system could generate replacement speech and modify the video so that the speaker appears to say the new words.
However, automatically deciding which expressions should be removed or corrected could create problems involving regional accents, cultural differences and individual speech patterns. <Cite refs={[“turn658331search2″,”turn658331search4”]}/>
Microsoft Patent Also Describes Real-Time Language Translation
Live translation is another major application described in the filing.
The proposed technology could identify speech in a language that is not included in a viewer’s language profile and replace it with translated words.
Unlike conventional subtitles or audio dubbing, the system would also modify the speaker’s visible lip movements to match the translated speech.
This could make international broadcasts feel more natural to viewers who prefer watching content in their own language.
Potential applications include:
- International business conferences.
- University lectures.
- Award ceremonies.
- Live television programmes.
- Sports broadcasts.
- Public presentations and events.
For example, someone watching an English-language lecture could receive an AI-generated version in Hindi or another preferred language, with the speaker’s mouth movements adjusted to match the translated audio.
The patent also describes potential use on smartphones, tablets, augmented reality glasses and video recording devices.
If implemented successfully, the technology could reduce language barriers in live digital content and make international events more accessible. <Cite refs={[“turn658331search5″,”turn658331search3″,”turn658331search6”]}/>
Viewers Could Receive Different Versions of the Same Broadcast
One of the more significant aspects of the proposed system is its ability to customise content based on individual viewer profiles.
The patent describes profiles that may contain information such as age, preferred languages, location and other user-specific parameters.
The AI system could use this information to determine which words should be modified for a particular viewer.
This means two people watching the same live event could potentially receive different versions of the speaker’s message.
For example, one viewer might receive a translated version, while another could see a version with selected expressions removed.
Such personalisation could make broadcasts more accessible and relevant. However, it also introduces questions about how much control viewers, broadcasters and technology providers should have over the original content.
If different audiences hear different versions of the same speech, establishing exactly what was communicated could become more difficult.
The concern is particularly relevant for political speeches, public announcements, interviews and other events where the precise wording can matter. <Cite refs={[“turn658331search4″,”turn658331search8”]}/>
Could Microsoft’s Technology Create New Deepfake and Misinformation Risks?
The ability to change spoken words and synchronise a speaker’s lips with replacement audio raises concerns about misinformation.
Traditionally, live video has been treated as a relatively direct record of what someone said at a particular moment. Microsoft’s proposed technology could introduce another layer between the original broadcast and what viewers actually see.
If words can be replaced without obvious visual or audio interruptions, audiences may struggle to distinguish original speech from AI-modified content.
Several questions emerge:
- Would viewers be informed when a broadcast has been edited by AI?
- Would the original audio and video always be available?
- Who would decide which words are inappropriate or require replacement?
- Could broadcasters use the technology to change statements rather than simply clean up speech?
- How would regulators verify the original content during disputes?
The patent describes preserving the original recording as a reference or “ground truth” in certain implementations.
That could help investigators and broadcasters establish what was originally said. However, preserving an original recording does not automatically ensure that viewers know when they are watching an altered version.
The application does not establish a universal requirement for visible disclosure of every modification. <Cite refs={[“turn658331search5″,”turn658331search4″,”turn658331search8”]}/>
Microsoft Patent Is Still an Application, Not a Confirmed Product
The filing was submitted by Microsoft Technology Licensing LLC on April 29, 2025, and published as US 2026/0341309 A1 on October 29, 2026.
It contains proposed technical claims covering AI-based modification of live video and audio, including a separate claim describing real-time language translation with matching lip movements.
However, a published patent application does not mean that a technology has been approved for commercial use or that a company intends to launch it.
Patent applications are examined by the relevant patent office, and claims can be amended, narrowed or rejected during the process.
Patentlyze reported that the examiner had raised objections to the claims as originally filed in July 2026. Microsoft may respond or amend the application as the examination continues.
There is currently no confirmed product launch date or commercial release announcement for this technology.
The distinction is important because technology companies regularly file patents covering ideas and potential applications that may never become consumer products. <Cite refs={[“turn658331search2″,”turn658331search5″,”turn658331search7”]}/>
The Bigger Picture
Microsoft’s proposed technology highlights how generative AI could move beyond creating new digital content to modifying existing content while it is being consumed. Real-time translation, speech cleanup and accessibility features could make live broadcasts more useful for international audiences and people who struggle with certain audio or language formats.
At the same time, the ability to change what a person appears to say creates a significant trust challenge. If AI can modify speech and facial movements without obvious signs of editing, the distinction between an original recording and a personalised version could become increasingly difficult to establish. The technology therefore raises questions about transparency, consent, authenticity and accountability that will matter if similar systems eventually reach the market.
Looking Ahead
The immediate next step is the patent examination process, which will determine whether Microsoft’s proposed claims are granted and in what form. Even if the application progresses, the company would still need to develop a reliable system capable of modifying speech and video with minimal delay while maintaining natural audio and facial movements. Microsoft has not announced plans to launch a commercial product based on the filing.
Longer term, AI-powered broadcast editing could become part of a wider shift towards personalised media experiences. The technology could improve accessibility and international communication, but its use in news, politics and public information would require particularly strong safeguards. Clear disclosure of AI modifications and reliable access to original recordings could become essential to maintaining public trust in live digital content.
SEO Metadata
SEO Title: Microsoft Patent Could Let AI Rewrite Live Broadcasts and Change Speech
Meta Description: Microsoft has filed a patent for AI that could rewrite speech in live broadcasts, remove unwanted words, translate languages and adjust speakers’ lip movements.
URL Slug: microsoft-patent-ai-rewrite-live-online-broadcasts
Focus Keyphrase: Microsoft AI broadcast editing patent
Secondary Keywords: Microsoft AI patent 2026, AI live broadcast editing, generative AI video editing, real-time AI translation, AI lip synchronization, live speech modification, Microsoft patent application
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
