Carol Espy-Wilson’s research into acoustic phonetics and speech perception has been published in journals for decades. Now, as founder and CTO of OmniSpeech, she’s shipping that research as two products: one that strips background noise from any audio stream, and another that scans for AI-generated deepfake voices in real time. The company, based in College Park, Maryland, operates with the precision of a lab and the nascent ambition of a venture-scale bet, having raised a total of $670,000, including a $500,000 pre-seed check from TEDCO in November 2025 [The Daily Record, November 2025].
The Academic Wedge
OmniSpeech’s differentiation is its academic pedigree, not a sales playbook. Espy-Wilson, a professor of electrical and computer engineering at the University of Maryland and the first African-American woman to earn a PhD in electrical engineering from MIT, leads the technical vision [MIT Technology Review, 2015]. CEO David Przygoda, who joined from a CMO role at Auto-Tune, handles commercial strategy [PR Newswire]. This split is common in deep tech spinouts, but the early traction is notable. The company’s AI Detect tool, a "zero-shot" model that can identify deepfakes from previously unseen AI voice generators, was a co-winner of the Federal Trade Commission’s Voice Cloning Challenge in 2024 [Federal Trade Commission, 2024]. That regulatory validation provides a credibility wedge into markets like public safety, defense, and enterprise communications where trust is non-negotiable.
Early Traction in Hardware and Software
The path to market runs through two distinct channels. OmniClear, the noise suppression engine, is already built into headsets for an unnamed major hardware company, targeting consumer electronics and hearing aids [Technical.ly]. In parallel, AI Detect is available as an app on Zoom, scanning calls in real time, with a planned Chrome extension for platforms like YouTube [Technical.ly]. The company is also listed as an Arm partner, with its detection software available on Arm core IP. This dual approach,embedding in silicon and layering onto SaaS platforms,shows a pragmatic understanding of distribution. Revenue is estimated under $5 million, and headcount is under 25, fitting the profile of a team focused on product integration before a broad sales push [ZoomInfo].
| Product | Core Function | Key Integration / Partner | Target Market |
|---|---|---|---|
| OmniClear | Background noise suppression | Built into headsets for a major hardware company | Consumer electronics, hearing aids, automotive [CBInsights] |
| AI Detect | Real-time deepfake voice detection | Zoom app, Arm IP, planned Chrome extension [Technical.ly] | Enterprise communications, public safety, defense |
The Scaling Equation
For a tools company, integration is only half the battle. The other half is performance at scale. OmniSpeech’s technical breakdown rests on a key claim: its AI Detect uses a zero-shot learning model. In practice, this means the system doesn’t require prior training on a specific AI voice generator to flag its output as fake. This is a significant architectural advantage for detecting novel threats, but it introduces a critical variable,latency and accuracy under load. Processing audio streams in real time on a platform like Zoom, with potentially millions of concurrent users, is a different class of problem than running a batch analysis in a research lab.
The sober assessment is that the company’s current infrastructure is untested at that magnitude. The pre-seed funding level, while sufficient for R&D and early partnerships, does not provide the war chest typically required to build the global, low-latency inference infrastructure that a standalone security service would demand. The most plausible near-term path is deeper embedding within larger platforms (like Zoom or hardware OEMs) that already own the scaling problem. The risk is becoming a feature, not a destination. If a platform partner decides to build or buy a competing solution, OmniSpeech’s standalone market narrows considerably. Their defense is a decade of proprietary research and a head start in regulated, high-stakes verticals where accuracy trumps cost.
Sources
- [The Daily Record, November 2025] TEDCO invests $500K in College Park-based OmniSpeech | https://thedailyrecord.com/2025/11/25/tedco-invests-500k-in-college-park-based-omnispeech/
- [MIT Technology Review, 2015] Carol Espy-Wilson, SM ’81, EE ’84, PhD ’87 | https://www.technologyreview.com/2015/10/20/165680/carol-espy-wilson-sm-81-ee-84-phd-87/
- [PR Newswire] Former Auto-Tune CMO joins OmniSpeech as CEO | https://www.wfmz.com/news/pr_newswire/pr_newswire_business/former-auto-tune-cmo-joins-omnispeech-as-ceo/article_17c03b56-6e3a-5366-8104-4ec0c4c8f9a8.html
- [Federal Trade Commission, 2024] FTC Voice Cloning Challenge | https://www.ftc.gov/news-events/contests/ftc-voice-cloning-challenge
- [Technical.ly] OmniSpeech uses AI to spot deepfake voices in real time | https://technical.ly/entrepreneurship/omnispeech-ai-voice-deepfake-detection-software/
- [ZoomInfo] OmniSpeech company information | https://www.zoominfo.com/c/omnispeech-llc/355963482
- [CBInsights] OmniSpeech company profile | https://www.cbinsights.com/company/omnispeech-1