Powered by Reka Spark, our low latency on-device speech-to-text translation model is state-of-the-art with less than 1 billion parameters and less than 1GB memory size. Suitable for deployments on cellphones, robots, and other edge devices.

We support speech translation to and from 10 languages including English, Japanese, Korean, Chinese, Arabic, French, Spanish, German, Italian, Portuguese.

Our model outputs both the transcription of audio input and text translation. You can try it yourself here (English-Korean). Contact us to customize audio output style.