Live speech-to-text, built for speed.

# pip install websockets   (Python 3.10+)
# Streams a 16 kHz mono 16-bit WAV in real time and prints words as they arrive.
import asyncio, json, wave
from websockets.asyncio.client import connect

URL = ("wss://api.saidwire.com/v1/listen"
       "?encoding=linear16&sample_rate=16000&interim_results=true"
       "&punctuate=true&endpointing=300&early_words=true")
KEY = "sk_live_..."

async def main(path):
    w = wave.open(path)
    frame = w.getframerate() // 50                     # 20 ms of audio
    async with connect(URL, additional_headers={"Authorization": f"Token {KEY}"}) as ws:
        async def send():
            while chunk := w.readframes(frame):
                await ws.send(chunk)
                await asyncio.sleep(0.02)              # real time, like a live mic
            await ws.send(json.dumps({"type": "CloseStream"}))
        sender = asyncio.create_task(send())
        async for raw in ws:
            m = json.loads(raw)
            if m["type"] == "EarlyWord":
                print("word  ", m["punctuated_word"])
            elif m["type"] == "Results" and m["is_final"]:
                print("final ", m["channel"]["alternatives"][0]["transcript"],
                      "(end of turn)" if m["speech_final"] else "")
        await sender

asyncio.run(main("hello.wav"))

// Browsers can't set headers on a WebSocket: pass the key as a subprotocol.
// Use a short-lived key in browsers, never your server key.
// Send 16-bit PCM (e.g. from an AudioWorklet); compressed browser formats are not supported yet.
const url = "wss://api.saidwire.com/v1/listen?encoding=linear16&sample_rate=16000&interim_results=true&early_words=true";
const ws = new WebSocket(url, ["token", KEY]);
ws.binaryType = "arraybuffer";
ws.onmessage = (e) => {
  const m = JSON.parse(e.data);
  if (m.type === "EarlyWord") showWord(m.punctuated_word);
  if (m.type === "Results" && m.is_final) showFinal(m.channel.alternatives[0].transcript, m.speech_final);
};
// pcm16 = Int16Array of 20-100 ms of mono audio
function onAudio(pcm16) { if (ws.readyState === 1) ws.send(pcm16.buffer); }
// when done:
ws.send(JSON.stringify({ type: "CloseStream" }));

# Stream a raw 8 kHz μ-law phone recording with websocat (https://github.com/vi/websocat).
# --binary sends stdin as binary frames; replies print one JSON message per line.
websocat --binary -H="Authorization: Token sk_live_..." \
  "wss://api.saidwire.com/v1/listen?encoding=mulaw&sample_rate=8000&early_words=true" \
  < call.ulaw