Speech neuroprosthesis

Also: speech BCI, speech brain-computer interface, brain-to-text interface, brain-to-voice interface

By neuraspeak editorial · Updated 2026-10-07 · 1 min read

A speech neuroprosthesis is a brain-computer interface that decodes neural activity from speech-related areas of the brain while a person attempts to speak, and converts it into text, synthesized voice, or an animated face, to restore communication for people who have lost the ability to speak because of paralysis.

How does a speech neuroprosthesis work?

Electrodes are implanted over or into the parts of the cortex that control the lips, tongue, jaw, and larynx. When the user attempts to say a sentence, even if no sound comes out, those areas still produce patterns of activity. A neural decoder, typically a recurrent neural network, converts these patterns into likely speech sounds (phonemes), and a language model assembles the sounds into the most plausible words and sentences. The output can be displayed as text, spoken through a text-to-speech voice, or used to drive a digital avatar.

How fast and accurate are speech neuroprostheses?

Performance has improved quickly. In 2021 a UCSF team decoded sentences from a 50-word vocabulary at a median 15.2 words per minute with a 25.6% word error rate (Moses et al., NEJM). In 2023, Stanford's BrainGate team reached 62 words per minute with a 23.8% word error rate on a 125,000-word vocabulary (Willett et al., Nature), and UCSF reached a median 78 words per minute with a 25% word error rate on a 1,024-word vocabulary, plus a voice and facial avatar (Metzger et al., Nature). In 2024 a UC Davis team reported 97.5% word accuracy over 8.4 months, with conversation at about 32 words per minute (Card et al., NEJM). Natural conversation runs around 160 words per minute, as Willett et al. note.

What are the current limitations?

Published results come from a small number of clinical-trial participants, each studied intensively. Systems require brain surgery, need regular recalibration because signals change over time, and are not yet approved commercial medical devices. A 2025 Nature study showed instantaneous voice synthesis, including changes in intonation and simple singing, but in a single participant (Wairagkar et al.).

Why it matters for speech BCI

This term is the core subject of speech BCI research. Every other concept in this glossary, from the electrode type to the decoder and the problem of signal drift, matters because it affects how fast, accurate, and reliable a speech neuroprosthesis can be for someone who cannot speak.

Questions

Who could benefit from a speech neuroprosthesis?+

People with intact speech-related brain areas who cannot speak because of paralysis, such as those with ALS (amyotrophic lateral sclerosis) or brainstem stroke. The participants in the landmark studies had these conditions.

Does the user have to try to speak out loud?+

The leading systems decode attempted speech, meaning the user tries to move their speech muscles as if talking, even if no intelligible sound results. This differs from purely imagined or inner speech.

Sources

Facts on this page are drawn from the sources listed and were accurate to the best of our knowledge on the update date shown. Company and product names are trademarks of their owners and are used for identification only. Nothing here is medical, investment or legal advice. Companies and individuals can request a correction via the corrections page.