stonkbuilderCOMMUNITY NeuralinkSpeech restoration (VOICE)
Independent Research Notice: Independent StonkBuilder Community research publication. Not affiliated with, operated by, or endorsed by Neuralink Corp. or PreStocks. Neuralink devices are investigational and have not received FDA marketing authorization. Nothing here is medical, legal, or financial advice. PreStocks tokens are not Neuralink shares, confer no shareholder rights, and are unavailable to U.S. persons and residents of prohibited jurisdictions.
Research Explainer VOICE Study (NCT07224256) · Communication Neuroprosthesis Verified 30 Oct 2026

Speech restoration with Neuralink: inside the VOICE trial and real-time brain-to-voice decoding

Losing the ability to speak to ALS or brainstem stroke cuts people off from natural human conversation. While cursor control lets users type letter-by-letter, Neuralink's VOICE study (NCT07224256) records directly from the speech-motor cortex to decode intended words in real time—synthesizing spoken audio in the participant's own pre-ALS voice and reaching a company-reported 75 words per minute (5.8% word error rate) within the first week of use in one participant.

NCT07224256VOICE trial registryRegistered communication restoration study
75 WPMFirst-week speech rateCompany-reported single-participant benchmark
5.8% WERWord error rateReported alongside the 75 WPM first-week milestone
~140 WPMNatural speech targetLong-term conversational benchmark
Key points
  • The VOICE Study (NCT07224256) targets speech-motor cortical regions that coordinate the lips, tongue, jaw, and larynx when a person attempts or silently intends to speak.
  • In March–October 2026 official videos (u8mmn9lOIyw, _j806JHhCRo, emi5O7LeiyU), participants Kenneth and Terry demonstrated real-time speech synthesis in their own cloned natural voices.
  • Neuralink reported a first-week benchmark of 75 words per minute (WPM) at a 5.8% word error rate (WER) in one VOICE participant—roughly half the pace of natural ~140 WPM speech.
  • Peer-reviewed 2025 studies in *Nature* and *Nature Neuroscience* independently validated instantaneous and streaming brain-to-voice synthesis in academic intracortical cohorts.

Why cursor typing is too slow for natural conversation

Natural human conversation flows at roughly 130 to 150 words per minute (WPM). Even an expert BCI cursor user or eye-tracker operator selecting keys on an on-screen keyboard typically communicates at 15 to 35 WPM. By the time a person with bulbar ALS finishes typing a sentence, the conversation around the dinner table has already moved on.

Instead of moving a cursor across a QWERTY keyboard, a speech neuroprosthesis places electrodes in the cortical regions that orchestrate vocal-tract movements (ventral premotor and primary motor cortex). When the user attempts to speak—or thinks of speaking a phrase—the decoder predicts phonemes, acoustic units, and words directly from neural firing patterns.

VOICE Study Pipeline · NCT07224256 How intended speech is decoded from speech-motor cortex into text and a participant's own voice
01 · SPEECH MOTOR CORTEX Articulatory Intent Participant attempts or silently thinks words. Neurons encoding lips, tongue, jaw & larynx fire in structured sequences. 02 · STREAMING DECODER Phonemes & Acoustic Units Neural networks map 1,024-ch spiking activity into streaming phoneme probabilities and spectral voice features. 03 · INTENT GATING & LM 75 WPM · 5.8% WER Context models assemble words while gating filters prevent unintended inner thoughts from triggering speech output. 04 · PERSONALIZED VOICE Natural Voice Synthesis Synthesizer renders live audio cloned from pre-ALS recordings plus on-screen transcript (Kenneth & Terry, 2026).
Four-stage speech decoding pipeline in the VOICE clinical trial from articulatory intent to personalized voice synthesis.

What the 2026 VOICE trial videos show: Kenneth, Terry, and the 72-hour home sprint

Between March and October 2026, Neuralink released three landmark videos documenting early human progress in the VOICE program:

Poster illustration for Speaking With The Mind | Neuralink
Official @neuralink Video 18 Sep 2026 · 6:56 Tier 1 · Human Trial (VOICE)
Speaking With The Mind | Neuralink

Click to load the privacy-enhanced youtube-nocookie.com player. No third-party cookies or requests are sent until you press play.

Watch on YouTube ↗ Read video analysis →
Source: Neuralink (@neuralink) · Channel UCLt4d8cACHzrVvAz9gtaARA · Video _j806JHhCRo · Checked 30 Oct 2026 Direct YouTube link ↗Read video analysis →

Official September 2026 update ('Restoring Speech with Neuralink') showing VOICE trial participants Kenneth and Terry training speech decoders and speaking aloud in their own personalized voices.

  • 'Translating Thought To Speech' (u8mmn9lOIyw, 18 Mar 2026): Introduced the first VOICE trial demonstration, showing intended speech decoded into live text and synthesized audio.
  • 'Restoring Speech with Neuralink' (_j806JHhCRo, 28 Sep 2026): Showed participants Kenneth and Terry progressing from mimed vocalization to silent speech intent, with audio rendered in a voice clone trained on recordings of their voice before ALS progression.
  • '72 Hours to Build for a Neuralink Participant' (emi5O7LeiyU, 29 Oct 2026): Documented a three-day engineering sprint inside a participant's home to build practical conversation tools, quick-response controls, and intent gating for daily family dialogue.
Poster illustration for An Engineering Sprint For Speech Restoration | Neuralink
Official @neuralink Video 29 Oct 2026 · 4:39 Tier 1 · Human Trial (VOICE)
An Engineering Sprint For Speech Restoration | Neuralink

Click to load the privacy-enhanced youtube-nocookie.com player. No third-party cookies or requests are sent until you press play.

Watch on YouTube ↗ Read video analysis →
Source: Neuralink (@neuralink) · Channel UCLt4d8cACHzrVvAz9gtaARA · Video emi5O7LeiyU · Checked 30 Oct 2026 Direct YouTube link ↗Read video analysis →

Official 29 October 2026 documentary ('72 Hours to Build for a Neuralink Participant') following Neuralink engineers inside a participant's home as they refine real-time speech tools.

To evaluate Neuralink's company-reported VOICE updates rigorously, it helps to compare them with the peer-reviewed neuroprosthesis literature published in 2025:

DimensionNeuralink VOICE Trial (2026 disclosures)Peer-reviewed academic benchmarks (Nature / Nature Neuroscience 2025)
Hardware interfaceWireless, fully implanted 1,024-channel N1 array (nl_speech, ct_voice)Intracortical Utah arrays (UC Davis, *Nature* 2025) & high-density ECoG (*Nature Neuroscience* 2025) with percutaneous connectors.
Reported speed & accuracy75 WPM at 5.8% word error rate in week 1 for one participant (nl_speech)Instantaneous (<10 ms latency) paralinguistic + pitch voice synthesis (*Nature* s41586-025-09127-3) and streaming brain-to-voice (*Nat. Neurosci.* s41593-025-01905-6).
Home usabilityFully wireless implant used inside participants' homes (emi5O7LeiyU)Primarily cart-based laboratory setups tethered to external neural signal processors.
Publication statusCompany clinical updates & registered trial NCT07224256; peer-reviewed paper pendingFull peer-reviewed datasets, open methods, and error-rate distributions.

Continue Reading in the Neuralink Cluster

Research Explainer

PRIME, CONVOY & VOICE trials

Compare Neuralink's human clinical trials—PRIME, CONVOY and VOICE—including ClinicalTrials.gov IDs, eligibility, 21 enrolled participants and endpoints.

Read the guide →Research Explainer

Telepathy & cursor control

Explore how Neuralink PRIME study participants use Telepathy for wireless cursor control, CAD design, digital painting, gaming and handwriting decoding.

Read the guide →Research Explainer

Safety, privacy & ethics

Objective analysis of Neuralink's clinical and ethical questions: surgical risks, thread retraction, biocompatibility, neural privacy and explant support.

Read the guide →Research Explainer

Future possibilities

An evidence-graded guide to Neuralink's future possibilities: what works in human trials today, plausible extensions, Blindsight vision R&D, and speculation.

Read the guide →

Sources checked 30 Oct 2026

Sources and official records

  1. [1] Neuralink: Speech Restoration (VOICE study overview) · Verified 30 Oct 2026
  2. [2] ClinicalTrials.gov: The VOICE Study for Communication Restoration (NCT07224256) · Verified 30 Oct 2026
  3. [3] Neuralink YouTube (2026-03-24): Translating Thought To Speech | Neuralink · Verified 30 Oct 2026
  4. [4] Neuralink YouTube (2026-09-18): Speaking With The Mind | Neuralink · Verified 30 Oct 2026
  5. [5] Neuralink YouTube (2026-10-29): An Engineering Sprint For Speech Restoration | Neuralink · Verified 30 Oct 2026
  6. [6] Nature (2025): An instantaneous voice-synthesis neuroprosthesis · Verified 30 Oct 2026
  7. [7] Nature Neuroscience (2025): A streaming brain-to-voice neuroprosthesis · Verified 30 Oct 2026