Modulate

Modulate is a voice AI company whose models analyze audio for what is said and how it is said. Its Ensemble Listening Model architecture combines transcript content with dozens of audio signals in each utterance, aiming to catch fraud, synthetic voices, customer churn and AI agent failures that transcripts alone miss. The offering has two sides. The platform, including the Velma conversation understanding suite and the ToxMod voice moderation product for games, is aimed at operations, trust and safety, and contact center teams. The Modulate API is aimed at developers and groups its models into families for detecting, transcribing, redacting and triaging audio. Models include deepfake voice detection, AI generated music detection, emotion and accent identification, language detection, audio event detection, multilingual speech to text covering more than 70 languages with diarization, and PII and PHI redaction from audio and transcripts. Batch processing runs over a REST API and most models also stream in real time over WebSocket. API pricing is usage based, per hour of audio processed. Examples from the published list: English fast speech to text at $0.025 per hour batch and $0.05 streaming, multilingual speech to text at $0.03 batch and $0.06 streaming, deepfake detection at $0.25, and Velma conversation analysis at $0.75 for up to 10 behaviors or $1.25 for up to 25. Credits cost $1.00 per 100 on pay as you go with no commitment. Annual commitment tiers are priced at $30K, $180K and $450K, with rates of $0.98, $0.95 and $0.90 per 100 credits.