Modulate raises $25M for its voice models and analysis suite
TechCrunch Ivan Mehta ● Covered by 2 sources
Modulate raised $25M for voice AI tools that read tone, detect deepfakes, and enforce rules. It matters because fake voices are easy now, and call centers need more than plain transcription.
Based on reporting by TechCrunch, Ivan Mehta — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Modulate, the Boston startup built around voice intelligence, has raised $25 million more for a platform that goes well beyond transcription. Its system uses a collection of smaller models to pick up emotional cues, spot deepfake or AI-made audio, detect synthetic music, and help enforce policy for voice agents in regulated industries.
The round was led by Future Ventures, with Hyperplane and Lakestar also taking part. PitchBook had previously pegged the company at $41 million raised and a $170 million valuation before this financing. Modulate was founded in 2017 by Mike Pappas and Carter Huffman, who met as MIT physics undergrads, and it started out in gaming voice modulation before shifting into moderation and, later, AI audio analysis.
That shift has only become more relevant as voice cloning gets easier and more enterprises add AI to customer service. Huffman says a lot of companies are focused on transcription, but the real gap is understanding the meaning and nuance of a conversation. Modulate is betting that knowing whether someone is frustrated, scamming, violating rules, or just sounding pleasant while being unhappy is a better way to judge a call than a simple positive-or-negative read.
The company says it now runs more than 100 models, split between signal extraction and analysis or detection. The first group looks at emotion, tone, language, and whether a voice is synthetic. The second tries to infer intent, such as what a caller wants or whether a rule is being broken.
Because those models are smaller, Modulate says it can avoid specialized hardware and heavy compute bills, which matters as token costs rise elsewhere. It also makes the system easier to expand with newer models and route requests through an orchestrator when needed. Modulate has 40 to 45 employees now, plans to add 10 more, and is working on more on-premises and on-device deployment for privacy. Its tools are already being used for scam detection, call-quality monitoring, compliance, and even voice-based cyberattack monitoring.
My take — AI-written commentary, not fact-checked reporting
Modulate is making the unglamorous bet that AI voice systems will need guardrails more than they need another cheerful demo. That sounds less sexy than “human-like” voice bots, which is exactly why it’s probably the smarter business. The market keeps rewarding companies that can fake a voice; the bill comes due when somebody has to tell whether that voice is lying.
Read more about this at: TechCrunch