Modulate has raised $25 million to expand its audio-native AI technology, including tools for recognizing fraud, synthetic speech and other risks in spoken conversations. Future Ventures led the financing, with participation from Hyperplane and Lakestar, bringing the company’s total funding to $60 million.
The company plans to invest in research, product development and engineering, as well as developer relations and partnerships. It is also expanding access to its technology through APIs, model offerings and additional deployment options for organizations building audio-based services.
Modulate’s Velma system analyzes audio directly rather than relying only on a written transcript. The company says it evaluates features such as tone, emotion, intent and emphasis, alongside indicators that speech may have been generated or manipulated. Those signals are used to assess what is happening within a conversation.
The underlying Ensemble Listening Model combines more than 100 specialized audio models. Modulate describes the approach as a way to capture different aspects of speech and conversational behavior through models trained for particular listening tasks, instead of treating transcription as the complete account of an interaction.
The company says its technology processes more than 10 million hours of audio each month and has analyzed more than 600 million hours in total. The financing is intended to support broader deployment as customers and developers use audio analysis in fraud prevention, safety and other conversation-based applications.
Audio impersonation is also prompting changes in consumer content platforms. YouTube plans to add voice detection to its creator likeness tools, extending a system that currently focuses on faces. The planned function is aimed at helping creators identify unauthorized uses of their voice in uploaded content.
Other deepfake defenses are being evaluated in identity verification. The U.S. Department of Homeland Security has expanded its testing program to include AI-generated documents and facial deepfakes, with separate exercises for document verification and biometric attack detection.
Modulate says current applications include detecting synthetic voices in high-risk calls, monitoring voice agents for customer misunderstandings, and identifying harassment or grooming in online conversations. The company is expanding its SDKs and partner integrations so developers can incorporate those listening functions into communications and security products without building specialized audio models themselves.
Sources: Modulate
–
By the ID Tech Editorial Team
