Subvocal AI Glasses Turn Silent Mouth Movements into Text: How the Technology Works

Subvocal AI glasses convert silent mouth movements into text, offering hands-free communication without keyboards, microphones or voice commands.

By Indrani Priyadarshini

on November 5, 2026

Imagine sending a message, taking notes or asking an AI assistant a question without typing a single word or speaking aloud. A pair of smart glasses could make this possible by turning subtle mouth movements into written text.

This is what The Subvocal Company, a startup founded by Aarush Yadav and Arnav Kumar, is attempting to develop. The company is building AI-powered glasses that can recognise silent speech by detecting small movements in the facial and speech muscles.

The startup has raised $1.3 million in funding from Y Combinator and Afore Capital to develop the technology. Its aim is to offer an alternative way to interact with digital devices, without depending on keyboards, touchscreens or voice commands.

But how can glasses understand words that are never spoken aloud? Here is a closer look at how Subvocal’s AI glasses work, their expected price and what the technology could mean for everyday communication.

What are Subvocal’s AI glasses and how do they work?

Subvocal’s AI glasses are designed to recognise subvocal speech, a form of silent communication in which a person deliberately forms words without producing audible sounds.

Read More | BharatAI: Women-Led Indian AI Startup Enters Harvard’s Global Business School Network

Unlike conventional voice-to-text tools, which use microphones to capture spoken language, these glasses are designed to detect the subtle physical movements associated with forming words. The company says its technology uses non-contact sensing to pick up signals from internal speech muscles. These signals are then processed and converted into text.

Public reports suggest that the glasses may use radio-frequency (RF) signals to identify tiny changes caused by movements around the mouth and face. However, the company has not publicly disclosed all the technical details behind its sensing system.

In practical terms, users would simply form words silently with their mouths. The glasses would detect these movements and translate them into digital text. This means users would not need to speak aloud or type on a screen to communicate with their devices.

Are Subvocal’s glasses a mind-reading device?

No. Despite the technology’s unusual approach, Subvocal’s AI glasses are not designed to read people’s thoughts. The system does not directly interpret brain signals or identify words from a person’s thinking process. Instead, it relies on intentional physical movements made when someone silently forms words.

For example, if a user wants to send a message saying ‘I will be there in 10 minutes’, they would need to form those words with their mouth without making a sound. The glasses would then attempt to recognise the movements and convert them into text.

This distinction is important because the technology still requires deliberate user input. It is essentially a new way of entering text rather than a system that can understand unspoken thoughts.

Can silent speech replace typing and voice dictation?

Subvocal is attempting to address two common problems with digital communication: the time it takes to type and the lack of privacy associated with speaking aloud.

Read More | Indian Startup Unveils Vision Nano 250g UAV for Special Forces, Claims 25-Minute Flight Time

Typing messages on a smartphone or computer can be inconvenient, particularly when someone is moving or has their hands occupied. Voice dictation offers a faster alternative, but it is not always practical in offices, public transport, meetings or other shared spaces.

Silent speech recognition could offer another option by combining the speed of spoken communication with the privacy of text input.

According to the company, its prototype can convert quiet speech into text at speeds exceeding 200 words per minute. This is a claimed performance figure and has not yet been independently verified.

If the technology can deliver consistent results outside controlled testing environments, it could be useful for professionals, students and people who regularly interact with AI assistants.

It may also offer a hands-free way to write messages, record notes and enter prompts without interrupting conversations or disturbing people nearby.

How much do Subvocal AI glasses cost?

The Subvocal AI glasses are listed at $699 in the US, with an early reservation price of $499.

The reservation programme includes a refundable deposit of $199, while the remaining amount is payable later. Customers who require prescription lenses will have to pay an additional amount. The company expects to begin shipping the glasses in the first quarter of 2027. However, the product is still under development, and the final specifications could change before commercial release.

Deliveries will also depend on the necessary regulatory approvals.

What are the limitations of Subvocal’s AI glasses?

While the concept offers an interesting alternative to conventional typing, several questions remain unanswered. The first is accuracy. The company’s claimed typing speed of more than 200 words per minute needs to be tested independently, particularly under everyday conditions.

The technology will also need to demonstrate that it can recognise different accents, languages, speaking patterns and unfamiliar words. Maintaining accuracy during longer interactions could be another challenge.

Comfort and usability will matter just as much. Users will need glasses that can detect subtle mouth movements consistently without requiring exaggerated facial expressions or repeated attempts to enter text.

Read More | India’s First Networked Orbital Data Centre: TakeMe2Space Targets 2028 SpaceX Launch

Privacy is another important consideration. Since the glasses are designed to detect facial and speech muscle movements, users will want clear information about what data is collected, how it is processed and whether any of it is stored.

Until these details and real-world performance are established, it is too early to determine whether silent speech recognition can become a practical replacement for conventional input methods.

News Image