Live transcribe

Live transcribe API

One WebSocket: send your mic audio; get words as you speak and each finished sentence with its language.

1. 1

   Get an API key

   Create a key in the console. It uses the same balance as the playground.

   Create a key
2. 2

   Send your first request

   The same request the playground sends. Set OPENNAN_KEY to your key.

   JavaScriptPythonEvents

   Copy

   const ws = new WebSocket("wss://api.opennan.com/v1/realtime/stt");
   ws.binaryType = "arraybuffer";
   
   ws.onopen = () => {
     ws.send(JSON.stringify({ type: "auth", ticket: session.ticket }));
     ws.send(JSON.stringify({
       "type": "session.config",
       "language": "auto"
     }));
   };
   
   mic.ondata = (pcm16) => ws.send(pcm16); // 16 kHz mono 16-bit PCM, <= 65536 bytes a frame
   
   ws.onmessage = (e) => {
     const msg = JSON.parse(e.data);
     if (msg.type === "transcript.partial") showDraft(msg.text);
     if (msg.type === "transcript.final") addLine(msg.sentence_id, msg.text, msg.language);
     if (msg.type === "session.end") console.log("ended:", msg.reason);
   };
   
   stop.onclick = () => ws.send(JSON.stringify({ type: "session.end" }));

3. 3

   Read the events

   While you talk, the server sends JSON events; the voice comes as base64 audio chunks with their format.

   session.started

   Your session is open, with its price and limits

   transcript.partial

   Words while you are still talking

   transcript.final

   A finished sentence and its language

   usage

   Seconds and cost so far

   error

   A problem, with a code and plain words

   session.end

   Why it ended: you, balance, silence or time limit