Subvocal Silent Speech Glasses: What to Test in 2026
A practical buyer test for quiet drafting, corrections and a development-stage reservation.
November 03, 2026
Subvocal silent speech glasses are designed to turn deliberate, inaudible speech into text. They're still in development, with preorders open in the US and shipments estimated for Q1 2027. For a shared-office worker, I'd ask whether an unprepared message survives capture and correction. The advertised speed of 200+ words per minute doesn't answer that.[2][11]
On November 3, I found the launch discussion on Hacker News's front page, with 783 points and 363 comments. Tech Times published a report that morning.[1][5] That response makes sense. Speaking a draft aloud can feel awkward when someone nearby is trying to concentrate, especially if the draft is about them.
I haven't tried the glasses. After reading the founders' explanation and preorder terms, I'd want to see someone compose a message silently, change their mind halfway through, and correct it. Then I'd consider paying the balance.
Silent speech still asks you to form the words
The founders say the frames use radio-frequency sensing. In their launch post, they say radio waves change as they pass the tongue, lips, and soft palate when someone silently mouths words. A model converts those changes into text. They also describe tests of closed-mouth articulation, where people deliberately form words.[6]
A headline saying glasses can read your thoughts blurs an important distinction. You choose the words and perform an input action. The material I read doesn't show that the glasses can read thoughts. Calling it silent speech gives readers a clearer idea of how the proposed interaction works.
The company's page shows an office and a train, plus a coding-agent example with coworkers nearby.[2] I can see the appeal. A better microphone won't make you comfortable saying a sensitive sentence aloud.
Tech Times gets the shared-office problem right. Its November 3 report also warns that independent testing hasn't established everyday accuracy.[5] I'd be cautious about claims that it could end typing at work. A demo that makes text appear doesn't answer questions about editing, fatigue, or which messages people choose not to send.
Ask for an unprepared paragraph, then change it
The founder says the person in the Hacker News video completed a 20-sentence calibration. He also says the team is still working through edge cases.[6] That gives some context, but it doesn't show how much practice another buyer would need.
Imagine handing a project over to a colleague. This is a hypothetical test, not a report from a Subvocal user. The draft uses a project nickname and a changed date. You start to promise delivery on Friday, then remember the supplier hasn't confirmed it.
Compose the message silently without reading a script. Stop when you reach the uncertainty. Change the promise to a request for confirmation, then go back and correct the project name mentioned earlier. Don't send the message.
Now check the result. Did the abandoned promise stick around? Is the nickname spelled the same way as in the original? Would you bother fixing the date on a crowded train? You can check these results without knowing how the sensor works.
For a fair comparison, type a similar message using your usual setup. If you can speak where you are, dictate one too. Track how long you spend finding and fixing errors. Don't judge precision from a single try. Repeat the task another day, after taking the glasses off and putting them back on.
The speed claim needs a clearer test. Does it count raw decoded words, or finished text after corrections? Buyers need the finished-text rate to estimate how long writing will take. Until a repeatable test provides that context, treat 200+ WPM as the company's claim.[2]
Try silence when you aren't entering text
Test the off switch by itself. In a shared office, you might wear the device while reading, eating, or listening to someone. Ask how capture starts. Check its active indicator, then find out how you stop without losing the draft.
In the launch discussion, the founder says the connected device will have a trigger key. Automatic detection of silent speech is still being researched and may come after launch.[6] Those are separate promises. Don't plan your day around automatic detection.
Before asking for a more automatic version, I'd test the trigger by hand. Use a disposable note, deliberately start input, silently compose a sentence, then stop. Leave the field open and do something else. Check whether anything appears. Then switch to another app and try again. This is a suggested acceptance test, not proof that the prototype passes or fails.
Staying quiet won't hide what's on your screen. If the message is sensitive, consider who might see your draft. Someone nearby can read it as you type, even if you don't say a word.
What the reservation buys today
The early bird price for Subvocal silent speech glasses is $499 per pair. In the US, reservations require a $199 deposit that goes toward the price, followed by a separate $300 final payment. Taxes, shipping, and optional prescription lenses cost extra.[3][11]
The terms say you can cancel and get your full deposit back before making the separate final payment. Q1 2027 is an estimate, not a promised delivery date. The preorder page also says delivery depends on FCC approval.[3][11] A development-stage reservation won't solve tomorrow's deadline. I would keep the current writing setup in place while waiting.
Before you make the final payment, ask for the latest software compatibility list and service terms. The terms allow development changes. They say the product and required service terms will be described before final payment, with notice and a cancellation choice for material changes.[11] Save a copy of the version you reviewed. If you need a specific laptop or app, ask whether that exact combination is supported rather than relying on a general compatibility promise.
Keep today's writing setup separate from the preorder
If you can speak freely at home but need quiet text input on your commute, these glasses may be useful. You can appreciate that without giving up the work tools you already use.
DictaFlow, the publisher of this article, is a standard voice-dictation app. On desktop, hold the shortcut, speak, then release to insert text at the active cursor. It can't decode silent articulation, and the article doesn't establish any integration with Subvocal.[7]
Use a keyboard for quiet work today. If speaking feels comfortable, an existing dictation tool may help you finish your draft. If you're evaluating DictaFlow, its current product facts explain the input and insertion limits.[7] Neither option shows how well the glasses will perform.
If Subvocal offers you a test session, bring the handover message you haven't sent. Change the delivery commitment halfway through. Correct the project nickname afterward. Keep the text you wrote and a record of the corrections so you can compare them with your next session.
Disclosure: A DictaFlow editorial written with AI from published sources. We have not tested Subvocal. The handover and capture tests are hypothetical.
Sources
[2] Subvocal - The Subvocal Company
[3] Subvocal - The Subvocal Company
[6] Show HN: Subvocal - Talk to your computer without making noise | Hacker News
[7] DictaFlow: Consumer Facts on Mac, Windows, Mobile, and Pricing