Funding news · Sep 28, 2026
Modulate raises $25M Series B to scale audio-native AI models
Modulate, a Boston-based frontier audio AI company, has raised $25 million in Series B funding. The round was led by Future Ventures, with participation from Lakestar and Hyperplane.
What Modulate Does
Modulate builds audio-native models designed to help machines understand voice conversations beyond the literal words being spoken. Its flagship platform, Velma, uses an Ensemble Listening Model architecture that orchestrates more than 100 specialised audio models to detect emotion, tone, intent and synthetic speech in real time.
That capability set underpins several commercial applications:
- Transcription
- Deepfake detection
- Voice agent supervision
- Trust and safety use cases, including fraud prevention and voice chat moderation
The company's approach differs from text-first AI systems. By treating audio as the primary modality rather than a downstream transcription problem, Modulate aims to capture signal — emotional state, intent, authenticity — that is lost when voice is reduced to text.
Background and Traction
Modulate was founded in 2017 by MIT alumni Mike Pappas and Carter Huffman. The company first became known for ToxMod, its voice moderation product used in Call of Duty. Across its history, Modulate says it has analysed more than 600 million hours of audio.
That volume of audio data is a meaningful asset in a category where model performance depends heavily on exposure to real-world, in-the-wild conversational audio — particularly for tasks like synthetic speech detection, where adversarial conditions are the norm.
Why the Category Is Drawing Capital
Audio AI has moved from a niche research area to a commercially contested layer of the AI stack. Two forces are driving investment:
Voice agents are proliferating. As enterprises deploy AI voice agents for customer service, sales and support, the need for supervision tooling — systems that can monitor, evaluate and intervene in live voice interactions — grows alongside deployment.
Synthetic audio is a rising threat surface. Voice cloning and AI-generated speech have become practical attack vectors for fraud and impersonation. Detection capabilities that operate in real time are increasingly a compliance and security requirement rather than a nice-to-have.
Modulate's Ensemble Listening Model architecture is positioned against both trends simultaneously, which is likely central to the investor thesis behind the round.
The Round
| Detail | Value |
|---|---|
| Round type | Series B |
| Amount | $25,000,000 |
| Currency | USD |
| Date | 28 September 2026 |
| Lead investor | Future Ventures |
| Participating investors | Lakestar, Future Ventures, Hyperplane |
What to Watch
The Series B gives Modulate capital to scale its audio-native model development and expand Velma's commercial footprint across trust and safety, voice agent supervision and fraud prevention.
For strategic acquirers, Modulate sits in a consolidating intersection: voice AI infrastructure, content authenticity and trust-and-safety tooling. The company's combination of a proprietary multi-model architecture, a large analysed-audio corpus and an existing enterprise deployment history in a major gaming franchise makes it an asset with both technical depth and demonstrated commercial reach.
Key questions going forward include how Modulate converts its ToxMod heritage into broader enterprise trust-and-safety contracts, and whether its deepfake detection capabilities become a standard component of voice agent deployments as that market matures.