# Cartesia uses state space models for sub-100ms voice AI on AWS

> Original title: Building ultra-low-latency voice AI applications on AWS

- Source: AWS Machine Learning Blog
- Published: 2026-11-03T16:40:54.000Z
- Bad Influence: https://badinfluence.news/items/g31rzjtysixd8aoodu2pn46af
- Original: https://aws.amazon.com/blogs/machine-learning/building-ultra-low-latency-voice-ai-applications-on-aws/

## Summary

Cartesia, an AWS Partner, uses State Space Models to deliver text-to-speech and speech-to-text with sub-100ms first-byte latency. Its Sonic text-to-speech model produces first audio in about 90ms across more than 44 languages, while its Ink speech-to-text model handles native turn detection. The post says Cartesia can be deployed on AWS for data sovereignty, latency control, or compliance needs.