Brain implants are giving people who lost their speech a voice again

Brain implants are helping people who can no longer speak to communicate again. How a speech decoder turns brain signals into words, what one man with ALS has managed to say, and what the technology still cannot do.

Ctrl+Alt Science 4 min
A head in profile outlined in purple light, with glowing neurons and sound waves leaving the mouth

Brain implants are helping some people who have lost the ability to speak to communicate again, Science News reports. The systems read signals from the part of the brain that controls speech and turn them into text, and sometimes into synthesised speech.

The article follows one man, Casey Harrell, who has used such a system every day for years. Here is how it works, what he has done with it and what still stands in the way.

One man, nearly 2 million words

Harrell was diagnosed with amyotrophic lateral sclerosis, or ALS, in 2020, and lost his speech soon afterwards. In 2023, in a clinical trial at the University of California, Davis, he had four small grids of electrodes placed in the speech-related part of his brain’s left motor cortex.

He now uses the system for 10 to 14 hours a day. By a count reported in June 2026 in Nature Medicine, he had communicated nearly 2 million words. Science News says he uses it for email, Slack, playing Wordle with his family and work.

The early days were striking. On day one, the system had a vocabulary of 50 words, and the test sentence was “What good is that?” On day two, the vocabulary was more than 125,000 words.

How a speech decoder works

Science News lays out the process, using a diagram adapted from a 2024 paper in the New England Journal of Medicine:

1. Electrodes detect electrical signals from neurons. 2. The signals are recorded as raw neural data. 3. Machine learning links chunks of that data to phonemes, the units of sound. 4. A predictive language model assembles the phonemes into words and sentences. 5. Text appears on a screen, and it can be turned into synthesised speech.

Speech involves upwards of 100 muscles, the article notes, which is part of why decoding it is so hard.

Faster sounds and smaller devices

Other teams are working on speed and size. A 2025 study in Nature by Wairagkar and colleagues used an algorithm that turned brain signals directly into sounds with a delay of about 10 milliseconds. Just over half of the sounds were intelligible to listeners.

A company called Paradromics, based in Austin, Texas, makes a fully implantable device, reported in June 2026. Its cortical module is about the size of a dime and holds 421 electrodes. A transceiver in the chest relays the signals to a receiver that produces sound. Matthew Willsey of the University of Michigan implanted it in a 68-year-old woman who could no longer speak intelligibly. The main goal is safety testing, with real-time speech as a longer-term aim. In September she used it to say “I have a lot to say” and to call her grandchildren.

A separate study in Nature Neuroscience, reported in September, found that two people with paralysis could use brain signals to control both the words and the gestures of digital avatars.

What still stands in the way

Harrell’s system has limits. It is not portable: it uses a wired port on top of his head, so it works only at home. It is bulky and slower than conversation; in one exchange, an answer took about two minutes. It is also limited in decoding sounds that are not words.

More widely, Science News says outcomes are not guaranteed and availability is uncertain. Implants would likely be expensive, need invasive surgery and face high demand compared with the number of skilled clinicians. Changes in the brain over time, especially as a disease progresses, could make the signals unreliable. The researcher Stavisky put it plainly: “But they’re not solved yet.”

How far away is this for patients?

One of the researchers, Brandman, estimates he is about five years from offering an implant to a patient with ALS, Science News reports. The article says ALS affects an estimated 35,000 people in the US. We have found nothing about when, or whether, a system like this could be offered in the UK.

Harrell described what the system means to him: “when I am one of many, instead of being one of one.”

What we do not know

We have relied on a single article, and we have not read the papers it cites. We will update this piece if the researchers publish new results or if a system moves toward clinical use.

ShareXWhatsApp

Leave a Reply

Your email address will not be published. Required fields are marked *