Meta Files Patent for AR Glasses That Could Speak on Your Behalf

By Patentlyze Team · Updated Jul 3, 2026

Meta is patenting a system that listens to a conversation, generates several possible reply options using an AI language model, and then speaks the one you pick out loud through your mixed reality headset. It's essentially an AI voice for people who need one.

How Meta's AR device would generate spoken replies

Imagine you're in a conversation but can't easily speak on your own. A device on your face listens to what the other person says, then your display shows a few suggested replies. You tap or glance at the one you want, and the device says it out loud for you.

That's the core idea here. Meta's mixed reality headset would use a large language model (the same category of AI behind chatbots) to come up with those reply choices, then a separate text-to-speech system to actually vocalize the one you select.

The filing doesn't specify exactly what triggers the suggestions or how you'd make your selection, but the direction is clear: the device acts as a real-time voice on your behalf, turning your choice into spoken words during a live conversation.

How the LLM picks and voices candidate responses

The patent describes a three-step loop running on or through a mixed reality device:

• Step 1 - Listen: The system captures incoming audio, presumably the other person speaking to you, as the trigger for generating responses.
• Step 2 - Generate options: A trained large language model (LLM) produces a set of candidate replies to what was just said. Think of it like a smarter version of autocomplete, but for full spoken sentences.
• Step 3 - Speak the chosen reply: Once the user selects one of those options (the method of selection isn't detailed in the claim), a text-to-speech model converts that text into audio and plays it back, effectively speaking for the user.

The claim is intentionally broad. It covers any mixed reality device, any LLM, and any text-to-speech system, which means Meta is staking out the general concept rather than a specific implementation.

The patent doesn't describe how the user makes a selection, whether that's eye-tracking, a hand gesture, a controller tap, or something else. That detail may live in related filings or simply isn't locked down yet.

What this means for AAC and hands-free communication

This filing sits squarely in the territory of augmentative and alternative communication (AAC), the field of assistive technology that helps people with speech or language disabilities communicate. Current AAC devices exist, but integrating this capability directly into a wearable mixed reality headset would make it far less stigmatizing and more practical for everyday use. You'd just be wearing glasses, not carrying a dedicated tablet or speech device.

For Meta, this also fits a broader pattern of positioning its Ray-Ban smart glasses and the Quest platform as more than gaming or social tools. A device that can speak for you in real time is a genuinely useful capability, and getting there first with a patent filing puts Meta in a strong position as the assistive-tech and wearable-AI markets overlap.

Editorial commentary on a publicly published patent application. Not legal advice.