---
title: >-
  University of California, San Francisco and UC Berkeley: AI speech neuroprosthesis for
  text, audio, and avatar control
slug: ucsf-speech-neuroprosthesis
stable_id: c42c26c667a37013
company: University of California, San Francisco and UC Berkeley
function_code: clinical_ops
pattern_codes:
  - creator_to_judge
evidence_strength: verified
publication_tier: showcase
freshness: current
reviewed_at: '2026-08-23'
updated_at: '2026-08-22'
source_quality_summary: 1 peer reviewed; publication outcomes are verified.
caveat_summary: >-
  Single-participant investigational system, not an approved general product. A 25%
  error rate requires correction and limits high-stakes communication.
collections:
  - human-still-decides
  - regulated-autonomy
bundle_version: 1.0.0
bundle_fingerprint: sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93
canonical_url: https://brianletort.ai/transformations/ucsf-speech-neuroprosthesis
---

# University of California, San Francisco and UC Berkeley: AI speech neuroprosthesis for text, audio, and avatar control

Attempted-speech brain signals can produce real-time text, audio, and facial animation as a faster channel.

Function: Clinical operations. Patterns: Creator to judge. Evidence: verified.

Freshness: current. Reviewed: 2026-08-23. Updated: 2026-08-22.


Source quality: 1 peer reviewed; publication outcomes are verified.

## Before

1. **Participant** — Uses head tracking to select letters on a commercial assistive device. (control: User intent and device selection.)
2. **Communication partner** — Reads the slowly composed text. (control: Human clarification.)

## After

1. **Participant** — Attempts to speak while 253 cortical electrodes capture speech-motor signals. (control: Participant-initiated use and clinical trial protocol.)
2. **AI decoder** — Decodes text and synthesizes speech and avatar movements in real time. (control: Language model, confidence, and user correction.)

## Decision rights

The participant owns intended communication and decides use; clinical teams control the investigational device and safety.

## Exception path

Incorrect text is corrected or repeated; the participant can revert to the existing assistive device.

## Outcomes

- **Communication rate** (verified): 14.2 words per minute with the participant's commercial head-tracking device. → Median real-time decoding rate of 78.3 words per minute.. Single-participant investigational system, not an approved general product.
- **Text decoding error** (verified): Naturalistic speech speed and accuracy were previously elusive for neuroprostheses. → Median word error rate of 25%.. A 25% error rate requires correction and limits high-stakes communication.

## Executive lesson

A breakthrough accessibility outcome can coexist with material error, invasiveness, and single-user limits; all belong in the case.

## Anti-pattern

Calling an investigational single-participant result a deployable cure.

## Questions for leaders

- How is consent verified when decoding is wrong?
- What noninvasive alternatives remain?
- How durable is calibration?

## Sources

- [A high-performance neuroprosthesis for speech decoding and avatar control](https://par.nsf.gov/servlets/purl/10582752) — Nature