Realtime voice APIs
The foundational voice AI model for a borderless world
Stream audio over WebSocket and receive transcription, translation, or synthesized speech with sub-second latency. Built for real-time, multilingual speech; drop-in ready for Python.
Speech APIs
Realtime and batch speech APIs, one voice model.
Overview
Start here
→
Quickstart, authentication, audio formats, and per-capability guides — everything you need to make your first call.
GuidesQuickstart
s2tSpeech to Text
→
Stream audio and receive transcription deltas live, or POST a file and poll for a batch transcript.
EN · JA · KO · ZHWebSocketREST
s2stSpeech to Speech Translation
→
Simultaneous translation between languages with sub‑second latency. Voice in, voice out; no waiting for the sentence to end.
EN · JA · KO · ZH · ESRealtime
t2sText to Speech
→
Synthesize natural speech from text. Streaming output, designed for agents and devices.
EN · JA · KO · ZH · ESStreaming
SDKs
Install and start streaming.
One client, three APIs.
Python:
pip install kotoba-sdk