Cartesia AI

Cartesia AI is the fastest real-time text-to-speech company, built on a novel State Space Model (SSM) architecture that achieves 90ms full model latency and 40ms turbo latency, the fastest production TTS available in 2026. Founded in 2023 by researchers from MIT and Cornell University, Cartesia raised $64 million in Series A funding from Kleiner Perkins in March 2025, bringing total funding to $86 million.

The SSM architecture provides a fundamental speed advantage over transformer-based TTS systems because it processes audio state incrementally rather than recomputing attention over the full context window, enabling real-time generation that keeps pace with human speech production. Sonic-3, the flagship model, adds native emotion control, AI laughter synthesis, expressive prosody, and 40+ languages.

TTS latency matters for conversational AI voice agents because the total voice agent response loop requires under 300ms for natural conversation flow. Cartesia's 90ms latency leaves sufficient budget for LLM inference and network round-trips within the perceptual threshold.

Enterprise infrastructure includes 99.9% uptime SLA, SOC 2 Type II and HIPAA compliance, and on-premise deployment for data sovereignty requirements. Pricing starts at $0.03 per thousand characters.

Primary use cases include AI voice agents, interactive voice response (IVR) modernisation, real-time dubbing, accessibility tools, and any application requiring ultra-low-latency voice generation.