Modulate

Voice AI · Deepfake Detection · Trust & Safety

Funding

Modulate raises $25M to scale audio-native voice AI platform Velma

Analyzing raw audio instead of transcripts lets its models catch deepfakes and intent that word-for-word transcription misses entirely.

• Modulate raised $25M led by Future Ventures, with Hyperplane and Lakestar participating, bringing its total funding to $60M.
• Founded in 2017 by MIT physics classmates Mike Pappas and Carter Huffman, the Boston startup builds audio-native models that read tone, emotion and intent directly from voice rather than text transcripts.
• Its Velma platform uses an 'Ensemble Listening Model' that selects among 100+ specialized small models per task, which Modulate says is up to 1,000x more efficient than one large model.
• Modulate processes over 10 million hours of audio monthly, has analyzed more than 600 million hours total, and ranks #1 on Hugging Face's Open ASR and Speech Deepfake Arena leaderboards.
• Its deepfake detection model reports 98.9% accuracy with a 1.1% equal error rate on public benchmarks, a bar Modulate positions as production-ready for fraud screening in healthcare and finance.
• The raise follows a $30M Series A led by Lakestar in 2022 and a prior valuation of roughly $170M, marking Modulate's pivot from gaming voice-chat moderation to enterprise fraud and AI-agent oversight.
• As enterprises roll out AI voice agents at scale, tools that verify authenticity and read intent directly from audio are emerging as a distinct security layer beyond transcript-based moderation.