Voxmind Raises £546K to Build On-Device Voice Authentication and Deepfake Detection

Voxmind, a UK voice biometrics start-up, has raised 546,491 pounds in a pre-seed round led by Ascension Ventures, with angel participation from Mark McDermott, co-founder of ScreenCloud, Russell Hart as lead angel, and members of the Cambridge angel network. The company is building voice authentication and deepfake detection that runs on the device, rather than calling a cloud service, and is targeting banks, contact centres, telecom operators, and unified communications hardware vendors.

The technical approach is what differentiates Voxmind in a market that has just been thinned out by the largest cloud providers. The platform uses what the company calls physics-based analysis of vocal tract biomechanics through a patent-pending phoneme-frequency extraction engine, rather than relying on language-trained machine learning features. Voxmind claims 99.8 percent deepfake detection accuracy in under three seconds on its internal benchmark, and says the model runs in under 500 megabytes of memory with no GPU and no cloud connectivity required.

The market backdrop is the recent exit of large cloud providers from voice biometrics as a managed service. Microsoft retired Azure Speaker Recognition in September 2025, and Amazon Web Services has said it will end support for Voice ID in May 2026, leaving customers that relied on those services to find alternatives. The cloud-provider retreat from voice biometrics has been driven in part by deepfake risk and by the operational difficulty of maintaining a hosted speaker-recognition service that can keep up with synthetic-speech models.

“We built for this architecture before the gap existed, physics-based, on-device, deployable anywhere,” said Jai Keerthi, Founder and Chief Executive Officer of Voxmind. The company offers three deployment models: an OEM software development kit for hardware manufacturers, a platform-native integration over WebSocket, gRPC, SIP, or SIPREC for contact centre and telephony platforms, and a cloud API for customers that want the hosted option. Voxmind says it has signed an OEM agreement with a major unified communications hardware provider.

The target customer set is shaped by the cloud-exit market gap: banks and financial services firms needing voice authentication that the cloud providers will no longer host, contact centres facing increasing volumes of deepfake-driven fraud calls, telecom operators, US community banks, IT services providers, and the OEM hardware vendors building enterprise IP phone and conferencing equipment.

Voxmind plans to use the funding to accelerate go-to-market and continue product development on the on-device model.

Sources: Tech.eu

–

By Ji-seo Kim