Cartesia is a real-time multimodal AI voice platform built around the Sonic-3 streaming text-to-speech model, the first commercial TTS to natively generate laughter, emotion, and conversational nuance with industry-leading latency under 100 ms (90 ms time-to-first-audio). The platform spans three core products: Sonic-3 for ultra-low-latency TTS in 40+ languages, Ink-Whisper for high-accuracy streaming speech-to-text, and Line, a full voice-agent development framework for building production-grade conversational agents. Sonic-3 supports inline emotion tags such as [laughter], [sigh], and [whisper], fine-grained dials for speed and volume, instant voice cloning from just three seconds of audio, and Pro voice cloning for studio-quality character voices. Cartesia is engineered on state-space model (SSM) architectures rather than traditional transformers, giving it dramatically lower inference cost and the ability to stream audio token-by-token in real time. The platform is used in healthcare scribing, customer service, gaming NPCs, hospitality call automation, and AI sales reps, with enterprise-grade compliance (SOC 2 Type II, HIPAA, PCI Level 1). Pricing is credit-based starting at a generous free tier with 20K credits, scaling to Pro at $4/month and Scale at $239/month for high-volume teams. With backers including Index Ventures, Lightspeed, and Conviction, Cartesia has rapidly become the developer-favorite voice AI infrastructure for latency-sensitive real-time applications.