|
 |
OmniSpeech CEO, David Przygoda, giving AI Detect demos in the CM media suite at the Venetian hotel at CES. |
|
When Carol Espy-Wilson (ECE/ISR) founded OmniSpeech in 2014, she aimed to transform years of research in speech communication and noise suppression into practical tools that could make a difference in everyday life. This month, that mission reached a new milestone with the launch of OmniSpeech AI Detect™, a real-time voice deepfake detection tool now available in the Zoom App Marketplace.
The product made its debut at CES 2026, drawing interest from national media and industry leaders for its novel approach to securing online communication. AI Detect equips users with the ability to identify synthetic voices during live video calls by analyzing speech patterns using a proprietary zero-shot machine learning model. With a single click, the software evaluates voices in the meeting and provides a simple, color-coded result: green for likely human, red for deepfake, and yellow when uncertainty remains. 
The need for such tools has become increasingly urgent. As generative AI technologies advance, voice cloning has become more accessible to the public, making it easier for malicious actors to impersonate others using just a few seconds of recorded speech. These threats extend beyond celebrity impersonations or political disinformation. Everyday users are now vulnerable to fraud and misinformation in their most routine conversations.
The company’s approach is grounded in scientific rigor. Its zero-shot AI model does not rely on preloaded examples of known voice forgeries. Instead, it learns general characteristics of natural human speech from diverse voice datasets and applies that knowledge to detect anomalies on the fly. This method allows AI Detect to identify synthetic voices generated by systems it has never encountered, providing a flexible and adaptive layer of security.
OmniSpeech was recognized by the Federal Trade Commission in 2024 as one of three winners in the small organization category of its Voice Cloning Challenge. The company is preparing for a new round of seed funding and plans to expand the product to Zoom for Government, Zoom Phone, and as a browser extension in the coming months.
“Deepfakes are a new form of noise — one that distorts reality and undermines trust in our voices”, said Epsy-Wilson. “With AI Detect, we’re building technology that cuts through that noise to protect authenticity in human communication. And this is just the beginning — we have many more innovations ahead.”
As people across the globe continue to rely on video conferencing for work, education, and connection, the ability to distinguish genuine voices from artificial ones is becoming a critical element of safe and effective online interaction. Espy-Wilson’s research journey exemplifies ISR’s mission to solve complex societal challenges through interdisciplinary systems science. With AI Detect, that mission extends into the virtual spaces where so much of modern life now unfolds.
Related Articles:
Sensor Advancement Breaks Barriers in Brain-Behavior Research
MATRIX Faculty to Present at International Conference
University of Maryland Represented at International Forum
Small Business Connections Help Meet Researcher Needs
UMD Engineering & HII Partner to Accelerate Defense Technologies
UMD Semiconductor Retreat Builds Strategic Momentum
UMD Student Bridging Research Innovation and Education Advocacy
MATRIX-Affiliated Faculty Solving Challenges with Solutions from Nature
AI Robotics Research at UMD Puts Safety and Trust First
Clark School Seed Grant Awards Announced
February 3, 2026
|
