Neuralink Demonstrates Thought-to-Speech in ALS Patient as VOICE Trial Begins Decoding Silent Neural Signals
Neuralink's VOICE trial decoded silent neural signals from an ALS patient's brain into synthesized speech as the company's BCI program scales to 21 participants worldwide.
Overview
Neuralink has demonstrated a brain-computer interface that converts silent neural activity into synthesized speech, marking a significant milestone for the company’s VOICE clinical trial. The demonstration, shared on March 31, 2026, featured a participant with amyotrophic lateral sclerosis who communicated audible words without moving his mouth or producing any sound, according to video evidence posted by the company and confirmed by CEO Elon Musk on social media.
The participant, identified as Kenneth Shock, received Neuralink’s N1 brain implant in January 2026. Within weeks of the procedure, he was using the device to speak, edit video, and control a computer. The achievement represents the first public demonstration of Neuralink’s speech-decoding capabilities in a human subject.
How the System Works
The VOICE trial targets brain regions involved in speech production rather than the motor cortex areas used in Neuralink’s earlier PRIME study for cursor control. When a person intends to speak, the brain sends instructions to the muscles of the mouth, tongue, and vocal cords. The N1 chip intercepts these neural signals before they reach the muscles, decodes them into phonemes, assembles the phonemes into words, and routes them to a speech synthesizer.
Neuralink engineers trained the system through a three-stage process. Shock first spoke sentences aloud while the implant recorded corresponding neural patterns. He then silently mouthed the same sentences as the decoder learned to map diminished motor signals to speech. Finally, he progressed to imagining speech with no physical movement at all, relying entirely on the neural intent captured by the implant’s electrodes.
The company used recordings of Shock’s voice from before his ALS-related speech deterioration to build a personalized voice model. The synthesizer reproduces his original vocal timbre rather than generating a generic computer voice, allowing his family to hear him speak in what they recognized as his own voice for the first time in years.
Current Limitations
The system introduces a multi-second delay between the moment Shock formulates a thought and when the synthesized audio plays. Neuralink has stated its long-term goal for the VOICE trial is to achieve conversational speeds of 140 words per minute by reading signals from speech-production brain regions, though the current demonstration falls short of that target.
Accuracy remains a challenge. The technology is still in an early feasibility phase and is not FDA-approved for commercial use. Neuralink has characterized the VOICE trial as an evaluation of initial clinical safety and preliminary efficacy of the N1 implant and its R1 surgical robot in restoring communication.
Scaling the Program
The speech demonstration arrives as Neuralink’s broader brain-computer interface program reaches its largest scale to date. The company reported in January 2026 that 21 participants, whom it calls “Neuralnauts,” are now enrolled in its trials worldwide, up from 12 in September 2025 and just three at the end of 2024, according to a company blog post. Participants have accumulated more than 670 combined days of implant use and over 4,900 hours of active device usage.
Neuralink’s expansion reflects a wider trend in the BCI sector. As STAT News reported in its 2026 outlook, clinical trial participation across the brain-computer interface industry is growing from single digits to dozens of patients, with leading companies expanding trials internationally.
Beyond speech, the company’s PRIME study participants have demonstrated typing speeds of up to 40 words per minute using a mental ten-finger keyboard system, and information transfer rates of over 10 bits per second, comparable to able-bodied mouse users. One participant, Bradford G. Smith, an ALS patient who is completely nonverbal, has used his implant to create and narrate YouTube videos using AI-synthesized speech built from archived recordings of his voice, as Interesting Engineering reported.
Production and Surgical Advances
Neuralink plans to begin high-volume production of its brain-computer interface devices in 2026 and transition to a nearly fully automated surgical procedure, according to Interesting Engineering. The company is also developing a technique for threading electrodes through the dura mater without removing it, which Musk described as “a big deal” for reducing surgical invasiveness.
Hardware upgrades are also planned. The current N1 implant uses approximately 1,000 electrodes, and Neuralink intends to increase that count to 3,000, which would substantially expand the volume of neural data available for decoding complex behaviors like speech.
Regulatory and Scientific Context
The VOICE trial operates under an FDA Breakthrough Device Designation that Neuralink secured for speech restoration, covering individuals with severe speech impairment caused by ALS, stroke, spinal cord injury, cerebral palsy, and other neurological conditions. The designation provides prioritized FDA review and increased agency interaction but does not constitute approval.
Companies developing brain-computer interfaces face a fundamental regulatory challenge: these devices do not cure disease, making it difficult to define and measure therapeutic benefit in the way regulators require for device approval, as STAT News has reported. Whether Neuralink’s VOICE trial can establish speech restoration metrics that satisfy FDA requirements for a pivotal trial remains an open question.
Neuralink has reported zero serious device-related adverse events across all 21 participants to date. The company has not disclosed a timeline for advancing from early feasibility studies to larger pivotal trials that would be required for commercial approval.