Australian neurotech startup Fluent is developing a brain-computer interface designed to help people who have lost speech due to conditions like motor neuron disease and multiple sclerosis. Their device captures brain signals emitted during an attempt to speak and converts them into text or synthetic voice. Early tests show the system can correctly identify intended phrases from a set of 128 options with up to 96% accuracy.
Fluent positions its technology as a middle ground between invasive implants and external sensors. Unlike many brain-computer interfaces (BCIs) that implant electrodes deep into brain tissue, Fluent’s device sits just under the skin but above the skull-a so-called sub-scalp interface. This approach avoids brain surgery while providing a cleaner neural signal than fully external methods.
The company focuses on decoding activity from the motor cortex, the brain region controlling muscle movements involved in speech, including those of the tongue, lips, and jaw. When a person tries to speak, neurons produce unique electrical patterns that Fluent likens to QR codes: each articulatory movement creates a distinct neural signature, which their algorithms translate into phrases.
To train their AI, Fluent collected a large dataset of neural speech signals using 144-electrode caps placed over the motor cortex. Volunteers participated by speaking aloud, mouthing words silently, and subvocalizing phrases internally. Collaborating with Japanese researchers, Fluent’s team developed a system that, in laboratory conditions, could guess the correct phrase from 128 choices with 96% accuracy. While impressive, these results are limited to a controlled phrase set-not free-flowing natural speech.
How Fluent’s sub-scalp brain-computer interface works
The brain-computer interface space has been buzzing recently. Elon Musk’s Neuralink demonstrated cursor control and text entry through implanted microelectrodes, while Synchron is pursuing a less invasive vascular implant. Both focus primarily on patients with severe paralysis, but restoring speech has emerged as a distinct challenge within BCI development.
In 2023 and 2024, US research groups published findings showing implantable BCIs converting brain signals into text at speeds reaching tens of words per minute. These implants offer high signal fidelity because electrodes sit directly on or inside brain tissue. However, they require brain surgery, carry lengthy recovery times, and face rigorous regulatory scrutiny.
Fluent takes a different route. It openly acknowledges that sub-scalp implants produce noisier, lower-quality signals than deep brain implants. The startup aims to bridge this gap by applying AI-including advanced language models-that infer missing parts of phrases from context. This approach resembles modern speech recognition systems, which predict words not only from sound input but also based on surrounding words. Here, the ”sound” is the faint and noisy neural activity beneath the scalp.
This strategy also makes commercial sense. According to industry analysts, the global brain-computer interface market is worth billions and growing rapidly, but mass adoption is hampered by high costs and complex implantation processes. Should sub-scalp solutions prove clinically effective, they could fit more naturally into rehabilitation technology portfolios rather than experimental neurosurgery. For patients, reducing medical risks and barriers to access is far more important than flashy demos.
Fluent still faces a long road. Transitioning from lab prototypes to certified medical devices requires extensive clinical trials and regulatory approvals. However, Australia’s supportive ecosystem-offering research grants, tax incentives, and a relatively accommodating regulatory environment-could give Fluent an advantage in early-stage testing compared to startups in stricter jurisdictions. If the company can move beyond a fixed phrase set to more natural speech decoding, it will represent a major milestone: demonstrating a practical communication tool for people who have lost the ability to speak.
As Fluent continues development, the major question remains how well sub-scalp brain-computer interfaces can scale in accuracy and vocabulary size without invasive surgery. Success here could reshape assistive communication technology by balancing effectiveness with safety-a critical factor for widespread adoption in healthcare.