Non-Invasive Brain-to-Speech

Thought becomes
speech.

The first non-surgical, no-movement system that decodes imagined speech and gives voice to people who have none — without a single electrode beneath the skin.

79%
Top-5 Accuracy
70M
People Unreached
0
Prior Benchmarks
See the Results Get in Touch
Scroll
See It Work

Not a claim. A recording.

Real-Time Demo
The system running on actual EEG data — live decoding, actual output.
Lab Session (Conceptual)
Shot in a real EEG lab — a conceptual visualization of how imagined speech decoding works.
The Problem
Who we build for
People whose minds are alive and whose voices are gone

ALS. Locked-in syndrome. Severe cerebral palsy. 70 million people live with complete or near-complete loss of voluntary movement. Their cognition is intact. Their ability to communicate is not. Existing systems require surgery, residual movement, or both. For the fully locked-in, nothing reaches them.

What we do
Decode imagined speech from brain signals. No surgery. No movement.

We read electrical signals from the scalp — the same signals the brain produces when a person imagines speaking — and decode them into sentences. No implants. No physical action. The person thinks the sentence. The system hears it. That product does not exist today without surgery. It exists now.

Proven Results

The system works.
Here is how well.

Every result is real, reproducible, and holds across multiple subjects. No benchmark existed for this task before we built one.

79%
Sentence decoding accuracy
In the top 5 guesses · 197-sentence vocabulary
1.92×
Signal quality improvement
Through custom research-identified electrode selection
24.5%
Exact match accuracy
First guess, correct — improves as the system learns each user
4
Subjects
Consistent results across everyone tested
As the system learns each individual's specific brain signals over time, the top-5 result becomes a single precise answer — the right one, every time. That is the product arc for each user.
How It Works

Brain signal to sentence.
Three steps.

The system is built on two discoveries: which brain regions to listen to, and how to translate what they say.

01
Finding the right signal
The brain has a language circuit — a specific network of regions that activate during imagined speech. We developed a custom mathematical approach that identifies the exact scalp electrode positions that capture this circuit, without being told where to look. The result: a cleaner signal from fewer electrodes. The hardware is unchanged. The methodology is new.
02
Decoding the sentence
The cleaned brain signal is translated into a sentence by matching it against a learned representation of language. The system retrieves the closest match from a vocabulary of 197 sentences. The thought reliably decoded, gives a fully locked-in patient the ability to communicate pain, need, love, and practical necessity. 79% of the time, the correct sentence is in the top 5.
03
Getting to open vocabulary
Beyond retrieval, a generation system produces free-form sentences directly from brain signal — any thought, not a predefined list. This is an early-stage prototype — it becomes clinically usable with more data. The closed-vocabulary product funds the data that trains the open-vocabulary one.
04
Building the wearable
The same circuit-mapping work that decodes speech also defines the physical device: a minimal, portable headset built around exactly the electrodes that matter — natural to wear, no setup. As the same technology moves from the clinic into daily life, that headset gets simpler, not more complex. What comes after that is a different kind of device entirely — one that doesn't just listen, but responds.
Where This Goes

The next interface
is not operated.
It is felt.

We start where the need is most urgent. Every phase that follows is built on what Phase One proves — about the brain, about the signal, about what becomes possible.

Phase One · Now
Thought to Speech

A working communication device for fully locked-in patients. The right vocabulary, reliably decoded, giving back the ability to communicate pain, need, love — without surgery.

Phase Two
Open Vocabulary + Everyday Wearable

Any sentence, not a fixed list — decoded directly from thought. The same decoder, built into a minimal wearable made for daily life, not just the clinic. The clinical work funds both.

Horizon
An Interface That Listens and Speaks

Today's systems only read. The next one reads and writes — sensing the brain's signal and responding to it directly, the way a body and a limb stay in constant, effortless sync. Neither side has to learn the other. They adapt together, automatically.

Let's talk.

Research collaborations, clinical partnerships, press inquiries, or anything else — reach out directly.

huzyafa@excelleve.com
Email Us