Meta Neural Band
The Meta Neural Band is a wrist-worn input device that reads surface electromyography (sEMG) signals from the forearm and converts them into commands for Meta Ray-Ban Display glasses. Meta announced it on 2025-09-17 at Meta Connect and put it on sale in the United States on 2025-09-30, bundled with the glasses for $799; it is not sold as a standalone product [1]. Meta says the band is "the product of years of surface electromyography (EMG) research with nearly 200,000 research participants" and that it "has the fidelity to measure movement even before it's visually perceptible" [1]. The peer-reviewed work underlying it appeared in Nature on 2025-07-23 under the title "A generic non-invasive neuromotor interface for human-computer interaction", authored by Patrick Kaifosh, Thomas R. Reardon and CTRL-labs at Reality Labs [2][3]. The band is the first shipping product from the team Facebook acquired when it bought CTRL-labs in 2019 [11], and Meta presents sEMG as a general input layer for its AI glasses and other devices rather than as an accessory for one product [6].
What it senses, and what it does not
sEMG places electrodes on the skin and measures the electrical activity of muscle fibers when motor neurons fire. The band sits on the wrist, where the tendons and muscles that move the fingers are close to the surface, so a movement can be detected from the muscle activation that produces it rather than from the visible motion itself. Because the sensing happens on the body, nothing needs to be in a camera's line of sight: reviewers confirmed the gestures work with the hand in a pocket, at the side, or in complete darkness [9].
This is not a brain-computer interface in the sense of implanted electrode arrays of the kind Neuralink develops, and it is not electroencephalography. Meta's own technical page states that "EMG sensors can only sense user outputs using electrical signals from muscle activations", that no information is sent back into the body, and that the technology requires no implants or surgery [6][7]. Dario Farina of Imperial College London, who was not involved in the work, put it plainly to IEEE Spectrum: "It is not a mind-reading system. It cannot make you act in a different way as imposed by your will; it does not connect you to other people neurally; it does not predict your intentions" [8].
Press coverage has been inconsistent on this point since 2019, when the CTRL-labs deal was widely described as a purchase of mind-reading technology. IEEE Spectrum's own 2025 piece carried the page title "Meta Wristband Interface: AI Translates Brain Signals" over an article headlined "Meta's New Bracelet Reads Hand Gestures" whose text stated that "the bracelet is not a direct interface with the brain" [8]. The distinction bounds what the device can do: it reads motor commands after they have left the spinal cord, so it can register an intended finger movement that never becomes visible, but it has no access to thought or to any intent a muscle does not act on. The machine learning that maps signals to commands runs on-device, according to Meta [6].
Hardware and availability
| Attribute | Value | Source |
|---|---|---|
| Announced | 2025-09-17, Meta Connect | Meta [1] |
| On sale | 2025-09-30, United States only | Meta [1] |
| Price | Included in the $799 Meta Ray-Ban Display bundle; not sold separately | Meta [1] |
| Sensing | Surface electromyography at the wrist | Meta [1][6] |
| Battery | Up to 18 hours (vendor-reported) | Meta [1] |
| Water resistance | IPX7 | Meta [1], UploadVR [9] |
| Band material | Vectran | Meta [1] |
| Sizes | Three | Meta [1] |
| Colors | Black and Sand | Meta [1] |
| Charging | Proprietary magnetic pin charger, included | UploadVR [9] |
Meta compares the Vectran band to the material used on the crash pads of the Mars Rover [1]. The company has not published an electrode count, channel count or sampling rate for the shipping band; those figures exist only for the research device described in the Nature paper.
Because the band ships only with Meta Ray-Ban Display, its availability follows that product's. Meta had planned to bring the glasses to Canada, France, Italy and the United Kingdom in early 2026, but announced on 2026-01-06 that it was pausing the expansion, citing "unprecedented demand and limited inventory" [17]. As of 2026-08-01 the glasses, and therefore the Neural Band, remain a United States product.
Gestures
Meta describes the band as letting the wearer "silently scroll, click, and, in the near future, even write out messages using subtle finger movements" [1]. UploadVR's David Heaney documented the shipping gesture set in a November 2025 review [9]:
| Gesture | Function |
|---|---|
| Thumb to index finger pinch | Select or click |
| Thumb to middle finger pinch | Toggle the display, go back; hold for shortcuts |
| Thumb double-tap on the side of the index finger | Summon Meta AI |
| Thumb swipe along the side of the index finger | Directional navigation, like a virtual d-pad |
| Pinch and twist the wrist | Volume and camera zoom |
Meta's help documentation describes the band as using EMG "to interpret the natural signals created by your muscle activity when you make certain gestures" [19], and its technology page says haptic feedback confirms that a gesture has registered [6].
Neural handwriting and the software track
Handwriting was announced as a future capability in September 2025 and shipped in stages, tracked in Meta's release notes under a Neural Band version series separate from the glasses' own [4]. The wearer traces letters with an index finger on any surface, on a leg, or in the air at their side, and the decoder turns the muscle activity into text.
| Version | Date | Added | Scope |
|---|---|---|---|
| v2 | 2026-01-06 | Neural handwriting for WhatsApp and Messenger, with conversation-based suggested replies | Early Access, US, English only |
| v3 | 2026-03-04 | Instagram Direct and Android phone messages | Early Access, US, English only |
| v4 | 2026-04-13 | Replies to notifications, including iPhone messages, with suggested quick replies | Early Access |
| v5 | 2026-05-14 | Handwriting across Instagram, WhatsApp, Messenger and phone message notifications, plus contact search | General release |
Meta introduced the first of these at CES 2026 alongside a teleprompter feature, and Alex Himel, Meta's vice president of wearables, said at the time that "Meta Neural Band and its EMG technology could be the best way to control any device" [5]. A further update on 2026-07-27 extended handwriting to querying Meta AI silently, by double-tapping the thumb and writing rather than speaking a wake phrase; Meta labeled that feature Early Access and limited to the United States and Canada [18].
The Nature paper
The Nature paper is the public evidence base for the product, and it is more specific than the marketing. The research hardware, called the sEMG-RD, carries 48 gold-plated dry electrode pins arranged as 16 bipolar channels sampled at 2 kHz, made in four wrist sizes and streaming over Bluetooth [2]. Meta collected training data from thousands of consenting participants and used it to train generic decoders, which are neural network models evaluated on people whose data was never in the training set.
| Task | Training participants | Closed-loop test participants | Median result |
|---|---|---|---|
| Continuous wrist navigation | 162 | 17 | 0.66 target acquisitions per second |
| Discrete gestures | 4,900 | 24 | 0.88 gesture detections per second |
| Handwriting | 6,627 | 20 | 20.9 words per minute |
Across all studies the paper reports 11,236 unique participants [2]. That is the figure to use for the published work; Meta's "nearly 200,000 research participants" is a separate, broader, vendor-reported count for its whole sEMG program and is not the paper's dataset [1][2].
The word "generic" is the claim that matters. Most prior high-bandwidth neuromotor interfaces have been invasive and tuned to one person, with decoders trained on that individual's own signals. Anatomy, muscle geometry, skin impedance and how a band happens to sit on a given wrist all change the signal, so a model fitted to one person transfers poorly to another; the paper shows exactly that for single-participant models. The finding is that with enough people in the training set, performance on entirely held-out participants improves in a power-law fashion, to the point where a new user can put the band on and it works without calibration. The authors call it "the first high-bandwidth neuromotor interface with performant out-of-the-box generalization across people" [2]. That is a scaling result as much as a sensing one, and it is what makes a mass-market wristband plausible where a lab device was not.
Personalization still helps. Fine-tuning a generic model on roughly 20 minutes of a specific user's own handwriting data produced a 16% improvement in median performance for models pretrained on 6,400 participants, with the largest gains going to the users the generic model served worst [2]. The paper also reports that personalized models do not transfer: using one participant's data to personalize a model for another improved performance in only 7% of such pairs, and on average made it 8.86% worse [2].
The authors list their own limits. Conventional input still wins on raw speed: a MacBook trackpad completed the same target acquisitions in a median 0.68 seconds against 1.51 seconds for the sEMG decoder, and a game controller managed 1.45 gesture completions per second against 0.88 [2]. IEEE Spectrum noted that 20.9 words per minute compares with roughly 36 words per minute for typing on a phone keyboard [8]. The paper demonstrates fully continuous control over only one degree of freedom, wrist flexion and extension, and states that it is unclear whether models trained on able-bodied participants will generalize to clinical populations, though early work looks promising [2].
From CTRL-labs to Reality Labs
CTRL-labs was founded in 2015 [13]. Its chief executive and co-founder Thomas Reardon had earlier founded the team at Microsoft that built Internet Explorer [11], then completed a doctorate in neuroscience [20]. The company raised $67 million from investors including GV, Lux Capital, Amazon's Alexa Fund, Spark Capital, Founders Fund, Matrix Partners and Breyer Capital [11].
Facebook announced the acquisition on 2019-09-23 in a post by Andrew Bosworth, then vice president of AR and VR, who wrote that "we know there are more natural, intuitive ways to interact with devices and technology. And we want to build them" [11]; Road to VR reported it the following day [13]. Bosworth also gave what remains the clearest short description of the technology: "You have neurons in your spinal cord that send electrical signals to your hand muscles telling them to move in specific ways such as to click a mouse or press a button" [12]. Facebook did not disclose a price. Bloomberg reported a range of $500 million to $1 billion, and TechCrunch said a source close to the matter confirmed that range [11][12]. Meta has never confirmed a figure.
The group became CTRL-labs at Reality Labs and kept publishing. A March 2021 Facebook Reality Labs post laid out the roadmap, quoting Reardon: "What we're trying to do with neural interfaces is to let you control the machine directly, using the output of the peripheral nervous system" [14]. It described starting with one or two bits of control from a pinch and working toward richer manipulation and eventually high-speed typing, and showed two wrist haptics prototypes, Bellowband and Tasbi, that have not shipped in a product [14]. Meta published an sEMG white paper in January 2025 [7], the Nature paper followed in July, and the consumer band arrived two months after that, demonstrated on stage by Mark Zuckerberg at Connect [1].
Meta has also begun showing the band outside its own glasses. At CES 2026 it demonstrated a Garmin Unified Cabin proof of concept in which a passenger controlled in-vehicle infotainment by scrolling and pinching [5]. Engadget's Karissa Bell, who tried it, called the in-car demonstrations fairly limited and the demo content disconnected from realistic driving, describing the work as exploratory rather than close to shipping [15].
Accessibility
Meta's launch post argued that sEMG has a specific advantage for people whose movements are small: "muscle signals at the wrist can provide control signals for people who can't produce large movements", including people who experience tremors or who have fewer than five fingers [1]. Because the band reads muscle activation rather than motion, a signal can exist where a camera-based system would see nothing.
Two research collaborations back this up. Meta's technology page names a partnership with Doug Weber's lab at Carnegie Mellon University aimed at hand paralysis [6]; a May 2026 Meta post described that work as a three-year collaboration and showed a participant with a spinal cord injury steering and boosting in a multiplayer racing game using two of the bands [16]. At CES 2026 Meta announced a University of Utah collaboration led by Jacob A. George, the Solzbacher-Chen Endowed Professor, evaluating custom gestures for people with muscular dystrophy, ALS and stroke, targeting smart-home control and the university's TetraSki adaptive skiing device as alternatives to joystick or sip-and-puff controls [5]. Engadget quoted Meta's claim that the band "is sensitive enough to detect subtle muscle activity in the wrist, even for people who can't move their hands" [15].
None of this is independently validated at scale yet, and the Nature paper's own caveat about clinical populations applies. Solomon Romney, formerly head of Microsoft's Inclusive Design Lab, told IEEE Spectrum that "this type of wearable assumes typical limb shape and fine motor control", which cuts against the accessibility framing for some of the users it is aimed at [8].
Reception
Early hands-on coverage was unusually positive about the band and much cooler about the glasses it comes with. Road to VR's Ben Lang wrote after the September 2025 demo that the Neural Band "seems like a real win for Meta" and that "if I could use the Neural Band for pinch detection on Vision Pro, I absolutely would" [10]. UploadVR found gesture recognition close to 100% accurate, with sideways swipes occasionally failing while walking, and no accidental display wakes [9].
The criticisms are mostly about the band as an object rather than as a sensor. UploadVR noted that it is a second wrist device to charge every night on its own proprietary connector, that it functions "solely as an sEMG input device, and nothing more" despite containing an accelerometer, and that display wake after a gesture often takes a second or two [9]. Eben Kirksey of the University of Oxford told IEEE Spectrum he doubted most computer users would rush out to buy the wristband if it became commercially available [8]; Joe Paradiso of the MIT Media Lab was more sympathetic, arguing that "interfaces like these are needed for the everywhere-computing eyewear that's emerging" [8]. Farina, while rejecting the mind-reading framing, said that "the amount of information decoded with this device is very large, far larger than any previous attempt" [8].
Meta has not published sales figures for the Neural Band, which would in any case track Meta Ray-Ban Display sales, and no independent teardown of the band's electrode array had been published as of 2026-08-01. Meta sells and documents the band only as part of Meta Ray-Ban Display, not as a controller for its display-less Ray-Ban Meta glasses [1][19]. The privacy argument around Meta's eyewear has centered on the cameras rather than the wristband, and is covered separately under smart glasses privacy; on the band itself, Meta's position is that the deep learning decoding runs on-device and that EMG sensing only reads outward [6].
References
- ^"Meta Ray-Ban Display: AI Glasses With an EMG Wristband." Meta Newsroom, 2025-09-17. about.fb.com/...n-display-ai-glasses-emg-wristband
- ^Kaifosh, P., Reardon, T. R., and CTRL-labs at Reality Labs. "A generic non-invasive neuromotor interface for human-computer interaction." Nature 645(8081), pp. 702-711, 2025-07-23. pmc.ncbi.nlm.nih.gov/...PMC12443603
- ^Europe PMC record for DOI 10.1038/s41586-025-09255-w (PMID 40702190, PMCID PMC12443603). Europe PMC REST API, retrieved 2026-08-01. ebi.ac.uk/...search
- ^"AI glasses release notes." Meta Help Center, retrieved 2026-08-01. meta.com/...1809764829519902
- ^"CES 2026: Meta Ray-Ban Display Teleprompter, Neural Handwriting, Industry & Research Collaborations, & More." Meta Blog, 2026-01-06. meta.com/...fied-cabin-university-of-utah-tetraski
- ^"EMG Wristbands and Technology." Meta Emerging Tech, retrieved 2026-08-01. meta.com/...emg-wearable-technology
- ^"Human-Computer Input via a Wrist-Based sEMG Wearable." Meta Blog, 2025-01-09. meta.com/...ace-emg-wrist-white-paper-reality-labs
- ^Choi, Charles Q. "Meta's New Bracelet Reads Hand Gestures." IEEE Spectrum, 2025-08-13. spectrum.ieee.org/meta-wristband-interface
- ^Heaney, David. "Meta Ray-Ban Display Review: First Generation Heads-Up Mobile Computing." UploadVR, 2025-11-10. uploadvr.com/meta-ray-ban-display-review
- ^Lang, Ben. "Hands-on: Meta Ray-Ban Display Glasses & Neural Band Offer a Glimpse of Future AR Glasses." Road to VR, 2025-09-17. roadtovr.com/...fer-a-glimpse-of-future-ar-glasses
- ^Matney, Lucas. "Facebook buys startup building neural monitoring armband." TechCrunch, 2019-09-23. techcrunch.com/...ilding-neural-monitoring-armband
- ^"Facebook acquires neural monitoring startup CTRL-labs." Engadget, 2019-09-23. engadget.com/...ntrl-labs-acquisition-emg-bracelet
- ^"Facebook Acquires BCI Startup CTRL-Labs to Develop Neural Input Device." Road to VR, 2019-09-24. roadtovr.com/...bs-develop-wrist-worn-input-device
- ^"Inside Facebook Reality Labs: Wrist-Based Interaction for the Next Computing Platform." Facebook Reality Labs, 2021-03-18. tech.facebook.com/...r-the-next-computing-platform
- ^Bell, Karissa. "Meta's EMG wristband is moving beyond its AR glasses." Engadget, 2026-01-06. engadget.com/...ng-beyond-its-ar-glasses-120000503
- ^"Our AI Wearables Are 'Changing the Game' for Disabled People." Meta Newsroom, 2026-05-18. about.fb.com/...nging-the-game-for-disabled-people
- ^Hayden, Scott. "Meta Pauses International Release of Meta Ray-Ban Display Glasses." Road to VR, 2026-01-07. roadtovr.com/...-ban-display-international-release
- ^"Meta Ray-Ban Display Glasses Get Even Better with Threads and New Version of Meta AI." Meta Blog, 2026-07-27. meta.com/...isplay-glasses-v127-muse-spark-threads
- ^"Get started with Meta Ray-Ban Display and Meta Neural Band." Meta Help Center, retrieved 2026-08-01. meta.com/...1864740664283499
- ^"Solving Billion-User Problems With CTRL-labs." Spark Capital (an investor in the company), retrieved 2026-08-01. sparkcapital.com/...n-user-problems-with-ctrl-labs
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.
v1 · 2,965 words · full history
Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify
Reviewer note: Independent adversarial fact-check, 2026-08-02: every load-bearing number, date, price, spec and quotation re-checked against primary sources by a second agent; citation markers audited against their references.