Voice, Speech & Realtime AI

Modulate raises $25M to push voice AI past transcription into intent and integrity

Source: TechCrunch · Sep 28, 2026

Modulate, a Boston voice-intelligence startup, has raised $25 million led by Future Ventures, adding to roughly $41 million raised previously at a $170 million valuation. Rather than one large model, it runs a fleet of more than a hundred small models that read vocal emotion and tone, detect synthetic-voice (deepfake) audio, and analyze conversation intent, policy adherence, and scam patterns — aimed at compliance and trust for voice agents in regulated industries.

This matters for our space because it is voice AI moving decisively past transcription. The frontier is no longer 'what words were said' but 'what was actually happening in the conversation' — affect, authenticity, intent. That is exactly the layer a multimodal interview record carries: not just the transcript, but how a real person reasoned, hesitated, and responded under real conditions.

It also underlines a growing demand signal around interview integrity. Detecting deepfaked or coached responses, and distinguishing genuine reasoning from performance, requires models trained on large volumes of authentic human conversation to know what 'real' sounds like. The more the market invests in analyzing live conversation for trust and intent, the more valuable a consented, real-human conversational corpus becomes as the ground truth those systems learn from.

Key Points

• Boston voice-intelligence startup raised $25M (Future Ventures), ~$41M raised prior at a $170M mark
• Runs 100+ small models for vocal emotion/tone, synthetic-voice (deepfake) detection, and conversation intent/policy analysis
• Targets compliance and trust for regulated-industry voice agents
• Emotion + deepfake + intent analysis over live conversation sits directly adjacent to interview integrity