A woman silenced by ALS sang a musical scale through a brain implant, and listeners now understand about 95% of words from UC Davis’ implant voice
The system turns brain signals straight into a voice that sounds like the person’s own, almost instantly, with no typing step. It is still experimental and has been tested in a handful of people.
A 58-year-old trained singer who lost her voice to ALS sang a musical scale using a brain implant, and an upgraded version of the system turns another patient’s brain signals into speech that listeners can understand almost word for word, University of California, Davis researchers report, according to The New York Times and a Society for Neuroscience release. The work is being presented at the society’s Neuroscience 2026 meeting.
Why it matters
ALS, or Lou Gehrig’s disease, slowly paralyzes the muscles, including those used to speak. Most speech implants so far turned brain signals into text first, then read it aloud after the person finished a sentence, which made conversation halting. The UC Davis system skips the text: it puts out sound in steps of 10 milliseconds, a hundredth of a second, so people hear themselves almost instantly as they try to speak and can stress words and change pitch.
In a preprint, a study not yet reviewed by outside scientists, the team says new software cut listeners’ word mistakes to 5.24%, meaning they understood about 95% of words, up from about 56% in the team’s 2025 Nature study using the same recordings. Casey Harrell, a climate activist with ALS who got four implants in 2023, has used the system to produce nearly 2 million words, the Times reported.
How it works
Surgeons placed four tiny grids holding 256 electrodes, hair-thin wires that pick up the electrical signals of individual brain cells, in the part of the brain that controls speech, UC Davis Health said. When the person tries to talk, artificial intelligence software built on the same basic design as ChatGPT, called a Transformer, predicts the sounds they meant to make and plays them in a voice made to sound like their own. The technology is experimental, part of a clinical trial called BrainGate2, and has been tested in very few people.
Source: Society for Neuroscience release via Medical Xpress, Nov. 14, 2026; The New York Times, Nov. 14, 2026; Wairagkar et al., bioRxiv preprint; Wairagkar et al., Nature, 2025; UC Davis Health, 2025.
- Society for Neuroscience (press release republished by Medical Xpress): Brain-computer interfaces: A new voice for people with paralysis (Nov. 14, 2026)
- The New York Times: Brain Implants Give A.L.S. Patients the Voice to Speak and Sing (Nov. 14, 2026)
- bioRxiv preprint: Wairagkar et al., Brain2voice 2.0: High-performance voice synthesis brain-computer interface (posted July 6, 2026)
- Nature: Wairagkar et al., An instantaneous voice-synthesis neuroprosthesis (2025)
- UC Davis Health: First-of-its-kind technology helps man with ALS ‘speak’ in real time (June 2025; 256 electrodes, four arrays, BrainGate2)