Every mouth movement produces different 'QR codes' in the brain
In Melbourne, a biomedical startup is quietly rewriting what it means to lose one's voice. Fluent, born from the University of Melbourne, has developed a brain-computer interface that listens to the brain's intention to speak — even when the body can no longer follow through — and translates those silent electrical signals into words. Sitting just beneath the scalp rather than inside the skull, the device offers people with motor neurone disease and multiple sclerosis something rarer than a medical advance: a path back to spontaneous human connection.
- Thousands of Australians with progressive neurological conditions are losing speech — and with it, the effortless intimacy of ordinary conversation.
- Current tools like communication boards and eye-tracking systems demand exhausting physical effort for every single word, reducing language to a slow, laborious transaction.
- Existing brain implants can restore some communication but require high-risk neurosurgery, making them inaccessible to many of the people who need them most.
- Fluent's under-scalp electrode array captures motor cortex signals with 96% accuracy — matching the quality of invasive implants without breaching the skull.
- Backed by over $2 million in funding and validated in preliminary trials at St Vincent's Hospital Melbourne, the company is preparing to launch clinical trials before year's end.
In Melbourne, a startup called Fluent is tackling one of the quieter crises of progressive neurological disease: the slow erasure of speech. For people living with motor neurone disease or multiple sclerosis, communication often narrows to pointing at letters on a board or laboriously tracking words with their eyes — methods that are exhausting, slow, and strip away the spontaneity that makes conversation human.
Fluent, a spinout from the University of Melbourne, is building a brain-computer interface that works differently from anything currently on the market. Rather than implanting electrodes inside the skull — a procedure with serious surgical risks — the device sits under the scalp but outside the bone. Co-founder and biomedical engineer Dr Tim Mahoney says the safety profile is better than a routine cochlear implant. The electrodes read electrical signals from the motor cortex, the brain region that coordinates speech movement. Even when a person can no longer produce sound, their brain still generates the same distinct patterns it would during normal speech. Fluent's system captures those patterns and decodes them into text or audio using machine learning — no button-pressing, no pointing. Just the intention to speak.
Preliminary testing at St Vincent's Hospital Melbourne proved the concept was more than theoretical. With 144 electrodes placed on participants' scalps, researchers recorded brain activity during spoken, mimed, and imagined speech, building what Mahoney describes as the largest English dataset of its kind. Combined with data from a Japanese research partner, their model identified the correct phrase from 128 options with 96 percent accuracy — and crucially, from outside the skull.
The company has raised over $2 million from investors including the University of Melbourne Genesis Pre-Seed Fund, Galileo Ventures, and international backers in New York and London. Multiple Sclerosis Western Australia also invested, underscoring how directly the technology speaks to patient need. Clinical trials of Fluent's insertable electrodes are set to begin later this year. Success would not cure the underlying diseases, but it could restore something that matters enormously in daily life: the ability to speak, and to be heard.
In Melbourne, a startup called Fluent is working on a problem that affects thousands of Australians: the progressive loss of speech that comes with diseases like motor neurone disease and multiple sclerosis. Right now, people facing this reality have few options. They rely on communication boards, charts where they point to letters or words, or eye-tracking systems that require constant physical effort just to form a sentence. It's slow, exhausting, and it strips away the spontaneity of human conversation.
The technology that might change this sits at the intersection of neuroscience and engineering. Fluent, a spinout from the University of Melbourne, is developing a brain-computer interface that works differently from anything currently available. Instead of requiring electrodes to be surgically implanted inside the skull—a procedure that carries real risks and limits who can access the technology—Fluent's device sits under the scalp but outside the bone. This seemingly small difference is profound. It means less invasive surgery, faster recovery, and a safety profile that Dr Tim Mahoney, the company's co-founder and a biomedical engineer, says is better than a routine cochlear implant.
The device works by reading electrical signals from the motor cortex, the part of the brain that controls the muscles involved in speech. Mahoney describes the concept with an elegant metaphor: think of these electrical signals as QR codes. When someone speaks, each movement of the mouth and jaw produces a distinct pattern of electrical activity. When someone with impaired speech tries to speak—even if no sound comes out—their brain still generates those same patterns. Fluent's electrodes capture these sequences, and machine learning algorithms decode them into text or audio. The person doesn't need to push a button or point at anything. They just need to think about speaking.
The proof came through preliminary testing at St Vincent's Hospital Melbourne. Researchers placed 144 electrodes on participants' scalps and recorded brain activity while they spoke, mimed speech, and imagined saying different phrases. The team built what Mahoney describes as the largest English dataset of its kind for this application, then partnered with a Japanese research group that had an even larger dataset. The result: their model could identify the correct phrase from a pool of 128 options with 96 percent accuracy. This wasn't theoretical anymore. It worked.
What makes this achievement significant is that it proved something researchers thought was impossible without invasive surgery. The signal quality captured from outside the skull matched what you could get from electrodes placed inside the brain. That means Fluent could develop and refine the technology without putting people through complex neurosurgery. The company has already raised over $2 million from investors including the University of Melbourne Genesis Pre-Seed Fund, Galileo Ventures, and international backers like Jumpspace Ventures in New York and Founder's Factory in London. Multiple Sclerosis Western Australia also invested, a signal of how directly this technology addresses the needs of people living with these conditions.
Clinical trials of Fluent's insertable electrodes are scheduled to begin later this year. If they succeed, the implications are substantial. Thousands of Australians with progressive neurological conditions could regain the ability to communicate naturally, to maintain independence, to participate in conversations without the exhaustion of pointing and selecting. It's not a cure for the underlying disease, but it's something that might matter more in daily life: the restoration of voice.
Citações Notáveis
With a safety profile that's even better than a routine cochlear implant, the technology will be more accessible to the broader population.— Dr Tim Mahoney, Fluent co-founder
Fluent was born out of the University's innovation ecosystem and is working to improve the independence of people living with constraining illnesses.— Professor Mark Cassidy, University of Melbourne Deputy Vice-Chancellor
A Conversa do Hearth Outra perspectiva sobre a história
Why does it matter that the electrodes sit outside the skull rather than inside it?
Because surgery inside the brain carries real risks—infection, bleeding, damage to surrounding tissue. Putting the electrodes under the scalp but outside the bone gives you most of the signal quality with a fraction of the danger. That changes who can access it.
How does the device actually know what someone is trying to say?
The motor cortex produces electrical patterns for every speech movement—each vowel, each consonant, each word has its own signature. When someone tries to speak but can't, their brain still generates those patterns. The device reads them and a machine learning model translates them into words.
What was the breakthrough in the testing?
They proved the signal quality from outside the skull matched what you'd get from inside. That sounds technical, but it means they don't need to do invasive surgery to develop and refine the technology. They can iterate faster and safer.
Who benefits most from this?
Anyone losing speech to progressive neurological disease—motor neurone disease, multiple sclerosis, ALS. Right now they're trapped using communication boards or eye-tracking. This could give them back something closer to natural conversation.
What happens next?
Clinical trials start later this year. If those work, you're looking at a technology that could transform independence for thousands of people. But we're not there yet.