Cartesia

Cartesia builds voice AI on state-space models rather than the transformer architecture most competitors use, and the payoff is speed — its Sonic model can produce natural-sounding speech in under 90 milliseconds end to end, which matters a lot for a phone bot or voice agent where any noticeable delay breaks the illusion of a real conversation. It's become the go-to TTS provider for teams building real-time voice products where ElevenLabs' latency is a dealbreaker.