Reka Speech is a multilingual speech model for speech-to-text and speech-to-speech, running on phones, cars, robots, and more. Supports more than 10 languages. With ~900 million parameters and a compact footprint of less than 1GB, this model delivers ultra low-latency performance while remaining lightweight enough for versatile deployment. It can be accessed through simple Reka API calls or run directly on-device, making it easy to integrate into a wide range of applications.

CLICK HERE to view the detailed user guide for more information. For more information about the product, please visit the Product Page.