Modulate, a Boston-based voice intelligence company, has raised $25 million in a new funding round to develop its enterprise-focused platform. The platform uses a collection of small models to provide conversation transcription, sentiment and tone analysis, and detection of deepfakes and AI-generated music, in addition to monitoring voice agents’ compliance with policies in regulated sectors.
The round was led by Future Ventures, with participation from Hyperplane and Lakestar. According to PitchBook data, Modulate had raised $41 million before this round, at a valuation of $170 million.
From Voice Modification to Conversation Analysis
Mike Pappas and Carter Huffman founded the company in 2017 after meeting while studying physics at the Massachusetts Institute of Technology. Modulate began by providing voice-modification technologies for games, before moving to audio-content moderation tools and then, as generative voice models became more widespread, focusing on detecting types of synthetic audio and analyzing speaker intent.
The company currently operates more than 100 models, broadly divided between models for extracting signals and models for analysis and detection. The first group analyzes emotion, tone, and language and determines whether audio is synthetic, while the second group examines the caller’s intent, the likelihood that they are violating rules, or whether they are attempting to carry out fraud.
What Changes in Practice?
Modulate generally works alongside the voice system used by an organization rather than replacing it. Its uses include alerting call centers to calls that may involve deepfakes or fraud, evaluating the quality of AI agents’ interactions with customers, and verifying these agents’ compliance with regulatory requirements in regulated fields.
The company believes that sentiment analysis alone is not sufficient to evaluate a call’s success. A customer may appear polite and show no obvious anger while actually being dissatisfied with the service. The platform therefore also focuses on understanding the customer’s intent and response, with the aim of providing more detailed data about why an interaction succeeded or failed.
Carter Huffman said that the company’s reliance on smaller models reduces its need for specialized hardware and substantial computing capacity, while also making it easier to train new models, add them to the platform, and invoke them when needed through a central coordinator. Modulate said that its technologies are also used to monitor cyberattacks through voice calls.
Expansion and Open Questions
Modulate currently employs between 40 and 45 people and plans to add ten employees in the coming months to support model development. It is also working to increase deployment capabilities at the customer’s premises and on devices in response to privacy requirements.
The funding is significant at a time when voice-based customer service is expanding, but the source does not identify the customers or provide independent results measuring the accuracy of deepfake and fraud detection. It also does not provide details about pricing or the regulated markets in which the platform is being deployed—points that will remain important for organizations evaluating call-analysis and compliance solutions.