Modulate raises $25M for its voice models and analysis suite

Start-up voice intelligence company Modulate has secured $25 million in fresh funding for its platform that provides transcription, emotion analysis, deepfake detection, AI music detection, and policy enforcement for voice agents across various industries. This investment follows a pattern seen in the emerging voice AI sector: supporting firms endeavoring to make AI voices appear more natural.

Furthermore, it competes with other entities striving to discern the intent behind human conversation through analysis and safeguard individuals and businesses from voice cloning, which has become increasingly feasible. Modulate's latest round of funding was spearheaded by Future Ventures, with participation from Hyperplane and Lakestar.

According to PitchBook data, the startup had amassed $41 million in funding at a $170 million valuation prior to this round. The company was established in 2017 by Mike Pappas and Carter Huffman, who crossed paths as undergraduates at MIT. Initially, Modulate focused on voice modulation for gaming but has since transitioned towards a voice-based moderation tool.

With the rise of voice AI models, the firm is now prioritizing the detection of various types of AI audio generation and the analysis of the intent behind a person's words. Huffman explained to TechCrunch that their unique understanding of the voice AI landscape is that many are adept at transcription, but there is a lack of tools that provide a comprehensive understanding of a conversation, which is crucial when engaging with another human being.

Modulate currently operates over 100 models classified into two categories: Signal extraction models to comprehend vocal emotion, tone, language, and synthetic voice determination; and Analysis/detection models that examine intent, such as the customer's objective, whether the caller is breaking rules, or if they are attempting to deceive the recipient.

Huffman noted that their smaller model infrastructure eliminates the need for specialized hardware and substantial computational resources, which is vital when token costs escalate. Moreover, it facilitates the company's ability to develop newer models, incorporate them into their system, and trigger them as needed. Modulate's clientele is diverse, but the company specializes in deepfake detection and alerting organizations like call centers to potential scams.

It also monitors AI agents' interaction with customers to evaluate call quality and ensure compliance with regulatory requirements. Due to these offerings, Modulate typically operates alongside a company's voice stack to analyze calls. As more enterprises embrace AI-powered customer service, it becomes increasingly vital for them to understand why a customer call was successful or unsuccessful.

In such scenarios, assessing customers' intent and response is critical, beyond mere analysis. Huffman emphasized that Modulate can provide detailed data to enterprises regarding this aspect. "When companies think of emotion analysis, they often consider if the customer was neutral or positive, meaning the call was a success, and if the customer was negative, the call was a failure.

However, in reality, people may remain polite even to AI agents or bots. They may not appear angry, but they might be highly dissatisfied," he explained. The startup asserts that its technology is also utilized to monitor cyberattacks via voice calls. Presently, Modulate employs a team of 40-45 individuals and intends to add 10 more personnel in the coming months to enhance model development.

The company is currently expanding its on-premises and on-device deployment capabilities to bolster privacy.