Artificial intelligence and brain-computer interface research is moving closer to enabling direct communication between human thoughts and machines. Tether’s neurotechnology division, Tether Evo, has introduced BrainWhisperer, a brain-to-text decoding project designed to translate neural signals into written speech using AI-augmented BCI implants. The system has achieved a variable 98.3% accuracy in translating brain signals, marking a significant step forward in neurotechnology for speech-impaired and paralyzed individuals.
During research trials, BrainWhisperer successfully decoded sentences generated from phonemes transmitted by intracranial BCI implants. One example included the line: “Do you know where it might have gone? I am an artist, lost in my own vision. I don’t think so anymore.” The project is part of Tether’s broader Brain OS initiative, an open-source brain operating system built on the company’s QVAC AI platform. Brain OS aims to connect with personal brain-computer interfaces and wearable devices while processing data directly on the device to protect user privacy. The goal is to enhance cognitive expression and intelligence using modern AI while ensuring users remain in control of their most personal thoughts.
BrainWhisperer uses a neural signal-decoding framework based on OpenAI’s Whisper Automatic Speech Recognition model. The system tokenizes neural signals and integrates a LoRA-Fine-Tuned AI model to improve transcription accuracy and reduce Word Error Rate (WER) over repeated trials. The project recently ranked 4th among 466 participants in the Brain-to-Text ‘25 Kaggle Competition, recording a 1.78% WER, just 0.25% behind the top result. The competition measures advanced brain-to-text decoding models using standardized performance benchmarks. Tether Evo’s entry featured a multi-stage transcription pipeline using ensemble learning with 5 models, trained with the Adam optimizer and a 100-epoch cosine learning rate, converting phoneme sequences into text through a Weighted-Finite-State-Transducer (WFST) system.
Ongoing research by Tether Evo is also exploring cross-subject neural signal decoding to reduce the lengthy calibration process typically required for each participant. The company’s Cross-subject contextual training system uses hierarchical CTC methods and affine transformation formulas to translate signals from multiple subjects while maintaining accuracy. Early results show 7.5% WER using standard CTC and 6.67% WER with hierarchical CTC, compared with 6.39% and 5.9% WER in current cutting-edge models. The team is also researching non-invasive BCI alternatives, such as wearable or ear-based devices using Surface Electromyography (sEMG) sensors to detect electrical signals from muscles. These efforts aim to improve usability, expand adoption and accelerate the future of brain-machine communication through the Brain OS ecosystem.
Also read: Viksit Workforce for a Viksit Bharat
Do Follow: The Mainstream LinkedIn | The Mainstream Facebook | The Mainstream Youtube | The Mainstream Twitter
About us:
The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.
