Speech BCI

Speech brain-computer interfaces represent the highest-impact frontier in BCI — restoring communication for people with severe paralysis who cannot speak. As of mid-2026 the field has crossed from "high accuracy in a controlled lab session" to two harder milestones at once: durable, unsupervised real-world use (Card et al., Nature Medicine 2026) and dedicated commercial speech programs entering first-in-human trials (Paradromics Connexus, Neuralink VOICE).

The 2026 inflection — speech BCI leaves the lab. A man with ALS and severe dysarthria used a multimodal intracortical BCI (a brain-to-text speech decoder plus a cursor decoder) for >3,800 hours at home over nearly two years, with no researchers present — communicating 183,060 sentences / 1,960,163 words at an average 56 words per minute, and labelling 92% of sentences as decoded at least mostly correctly (Card et al., Nature Medicine 2026). This is the clinical-utility complement to the accuracy benchmarks below: not a one-session WER number, but a sustained daily communication tool. In parallel, two companies put speech-first BCIs into patients — Paradromics completed the first-in-human implant of its Connexus device (421-microelectrode wireless array, >200 bits/sec in preclinical models) in June 2026, and Neuralink opened its VOICE speech-restoration trial (FDA Breakthrough Device Designation).

The two earlier breakthroughs that set up this moment are converging to make practical speech BCI realistic within the next few years.

BraIn-to-Text (BIT) introduces an end-to-end differentiable network that decodes neural activity directly into sentences, achieving a 10% word error rate — down from the previous state-of-the-art 24.69% (a ~60% relative improvement). The key innovation is contrastive learning for cross-modal alignment: rather than decoding neural signals directly to text, BIT aligns neural embeddings with audio LLM representations, leveraging the linguistic knowledge already embedded in large audio-language models. This "neural-to-audio-to-text" bridge dramatically reduces the neural training data needed.

Stanford inner speech decoding demonstrates that private inner monologue — thinking words silently — can be decoded from motor cortex microelectrode arrays. The key neuroscience finding is that inner speech patterns are structurally similar to attempted speech patterns in motor cortex, just with reduced amplitude. This means BCIs designed for attempted speech can potentially be adapted for inner speech with sensitivity improvements. For patients with locked-in syndrome, this could enable direct thought-to-text communication without any physical effort.

Key Claims

• 10% word error rate achieved for brain-to-text — BIT framework, down from previous SOTA of 24.69%. Single end-to-end differentiable network. Evidence: strong (BIT Framework)
• Cross-modal alignment with audio LLMs is the key innovation — Contrastive learning bridges neural signals to language via audio representations. Reduces neural training data requirements. Evidence: strong (BIT Framework)
• Inner speech decoded from motor cortex — 4 patients with severe paralysis. Inner speech patterns are attenuated versions of attempted speech patterns. Evidence: strong (Stanford Inner Speech)
• Same neural substrate for inner and attempted speech — Motor cortex encodes both; BCIs for attempted speech may be adaptable for inner speech. Evidence: strong (Stanford Inner Speech)
• Speech BCI demonstrated as a durable, unsupervised at-home device — An ALS/dysarthria participant used a multimodal speech+cursor intracortical BCI for >3,800 hours over ~2 years with no researchers present, at 56 wpm with 92% of sentences decoded at least mostly correctly. The first strong evidence of real-world, long-term, independent use. Evidence: strong (Nature Medicine — Card)
• Dedicated commercial speech-BCI programs are now in humans — Paradromics implanted its Connexus speech BCI (421 microelectrodes, >200 bps preclinical) in its first patient (June 2026); Neuralink opened the VOICE speech-restoration trial with FDA Breakthrough Device Designation. Evidence: moderate (tech-report/multi-source) (Paradromics Connexus, Neuralink VOICE)

Benchmarks & Data

• 10% WER vs. 24.69% previous SOTA (~60% relative reduction) (BIT)
• 4 patients with severe paralysis (ALS, spinal cord injury) for inner speech (Stanford)
• Inner speech amplitude reduced vs. attempted speech but structurally similar (Stanford)
• Real-world at-home use: >3,800 hours over ~2 years; 183,060 sentences; 1,960,163 words; 56 wpm; 92% sentences ≥mostly-correct (Nature Medicine — Card)
• Paradromics Connexus: 421 microelectrodes; >200 bits/sec information transfer in preclinical models; chest transceiver + transcutaneous wireless link (Paradromics Connexus)

Open Questions

• Can 10% WER generalize across patients and recording modalities?
• What is the pathway from inner speech decoding to real-time thought-to-text?
• Can the BIT framework work with non-invasive (EEG) or minimally invasive (Stentrode) recordings?
• How does vocabulary size affect accuracy (open vocabulary vs. constrained)?
• What are the privacy implications of inner speech decoding?
• Does the Card at-home result generalize beyond a single, well-characterized participant to a broader patient population and to less-invasive recording hardware?
• Will Paradromics' >200 bps preclinical claim translate into proportionally faster, more natural speech output in the Connect-One trial — and how does raw bandwidth map to usable communication rate (Card achieved 56 wpm on a lower-bandwidth array)?

Related Concepts

• Invasive vs. Non-Invasive BCI — Recording modality determines signal quality for speech decoding
• Neural Signal Decoding — Underlying computational challenge
• Neuroprosthetics — Clinical application for communication restoration

Related Entities

• Paradromics — Speech-first high-bandwidth Connexus BCI; first-in-human implant June 2026
• Neuralink — VOICE speech-restoration trial (FDA Breakthrough)
• Stanford Neural Prosthetics Lab — Inner-speech and brain-to-text research lineage

Backlinks

Pages that reference this concept:

• BIT Framework
• Stanford Inner Speech
• Nature Medicine — Card at-home speech BCI
• Paradromics Connexus first implant
• Neuralink VOICE / Blindsight / scale

Changelog

• 2026-04-05 — Created from BIT framework + Stanford inner-speech sources
• 2026-06-24 — Added the 2026 "speech BCI leaves the lab" inflection: Card et al. (Nature Medicine) at-home long-term use (>3,800 hrs, 56 wpm, 92% correct), Paradromics Connexus first-in-human implant, and Neuralink VOICE trial; new claims, benchmarks, open questions, and entity links