Modulate
Category: Voice / Speech AI
Modulate builds audio-native voice AI models that analyze raw conversation audio. The models detect fraud, deepfakes, toxicity and AI voice-agent failures that transcripts miss. Modulate was founded in 2017. The company is led by Carter Huffman. Based in Somerville (Boston area), Massachusetts, United States. Total funding raised: $60M. Latest round: Venture Round. Key investors include Future Ventures, Lakestar, Hyperplane.
- Founded
- 2017
- Headquarters
- Somerville (Boston area), Massachusetts, United States
- Total funding
- $60M
Value proposition
Modulate's audio-native AI reads both what is said and how it is said, including emotion, tone, intent and synthetic-voice signals. Analysis runs in real time or in batch. It is used to prevent fraud, moderate voice chat, guardrail AI voice agents and improve customer interactions. Modulate says its approach is far cheaper and more accurate than sending audio to large LLMs.
Products and solutions
Velma: the flagship platform and model suite, built on the Ensemble Listening Model architecture. Velma Detect covers deepfake detection, AI music and singing detection, language, emotion, accent, audio event and voicemail detection. Velma Transcribe offers multilingual and English fast speech-to-text, with medical and agentic transcription planned. Velma Redact removes PII and PHI. Velma Triage and Triage Mini surface risks. ToxMod provides real-time voice moderation for gaming and social platforms. The Modulate Platform is the enterprise product, and the Modulate API is self-serve (batch transcription costs $0.03 per hour).
Unique value
The Ensemble Listening Model picks from more than 100 small specialized audio models for each task and blends their results. Modulate claims this is up to 1,000x more efficient, and 2-4x more accurate, than using a single large model. It also delivers real-time analysis, such as real-time diarization, that transcript-based tools cannot match.
Target customer
Gaming and social platforms; contact centers and CX teams; banks, insurers and fintechs; healthcare institutions; IT helpdesks; companies deploying AI voice agents; developers building voice experiences.
Industries served
Gaming & Social, CX & Contact Centers, Insurance & Banking / Fintech, Healthcare, IT & Helpdesk, Trust & Safety, AI voice agents
Technology advantage
Modulate's models have been trained on very large volumes of real conversation audio: more than 600M lifetime hours and over 10M hours per month. Its transcription model ranked #1 on the Hugging Face Open ASR Leaderboard in July 2026. Velma Deepfake Detect ranks #1 on the Speech Deepfake Arena, with a claimed 98.9% accuracy on public benchmarks. The company also runs low-latency real-time inference and has an established trust-and-safety franchise with AAA game studios.
How they differentiate
Modulate analyzes audio natively instead of working from transcripts, and it uses an ensemble of small specialized models rather than one large model. This gives it lower cost and latency. It can also tell legitimate uses of synthetic voice apart from malicious ones. Detection, transcription, redaction and triage are all available through a single API key.
Main competitors
Pindrop (voice fraud / deepfake detection), Reality Defender (deepfake detection), Hume AI (emotion-aware voice AI), speech-to-text providers such as Deepgram and AssemblyAI
Key partnerships
Activision (ToxMod voice moderation for Call of Duty), Ubisoft (ToxMod for Tom Clancy's Rainbow Six Siege), co-founder Mike Pappas participates in regulatory and policy working groups in Washington, D.C.
Notable customers
Activision (Call of Duty), Ubisoft (Rainbow Six Siege), VoiceRun
Major milestones
2017: founded by MIT alumni Carter Huffman and Mike Pappas. 2019: $2M seed round (Hyperplane). Aug 2022: $30M Series A led by Lakestar. ToxMod was deployed in major titles, including a 50% cut in toxic voice chat in Rainbow Six Siege. Mar 2026: launched Velma Deepfake Detect, which ranks #1 on the Speech Deepfake Arena. Jul 2026: its transcription model topped the Hugging Face Open ASR Leaderboard. 2026: Carter Huffman became CEO and Mike Pappas became Chairman. Sep 2026: raised $25M led by Future Ventures, bringing total funding to $60M.
Growth metrics
More than 10M hours of audio processed per month and over 600M lifetime hours (the website shows about 563M hours of conversations improved). The company reports 40M+ unique business risks detected, over 40M consumers protected, and more than $100M in enterprise value protected.
Market positioning
Modulate is a frontier audio-intelligence layer for voice AI. It grew out of gaming voice moderation (ToxMod) and has expanded into enterprise fraud prevention, deepfake detection and AI voice-agent monitoring. It is now pushing developer and API adoption.
Geographic focus
Primarily United States and global gaming/social markets; European investor base (Lakestar).
About Carter Huffman
Co-founder and former CTO of Modulate (became CEO in 2026, succeeding co-founder Mike Pappas, who is now Chairman); ex-NASA Jet Propulsion Laboratory, where he built machine learning models optimized to run on spacecraft; MIT alumnus (physics research on dark matter and cosmic expansion).
Latest news about Modulate
More Voice / Speech AI companies
Official website: https://www.modulate.ai