Nova-3 Realtime

# pip install websockets
# audio.pcm: mono, 16 kHz, signed PCM16 little-endian; <= 30 seconds.
import asyncio, json, os
from pathlib import Path
from websockets.asyncio.client import connect

async def main():
    headers = {"Authorization": "Bearer " + os.environ["SPEECH_API_KEY"]}
    async with connect("wss://api.speechinfra.com/v1/inference/nova-3-realtime/realtime", additional_headers=headers) as socket:
        await socket.send("{\"language\":\"en\",\"smart_format\":false,\"diarize_model\":\"off\",\"interim_results\":false,\"endpointing\":10,\"vad_events\":false,\"type\":\"session.configure\"}")
        while True:
            event = json.loads(await socket.recv())
            if event["type"] == "error":
                raise RuntimeError(event)
            if event["type"] == "session.updated":
                break
        async def send():
            audio = Path("audio.pcm").read_bytes()
            assert 0 < len(audio) <= 960000 and len(audio) % 2 == 0
            for i in range(0, len(audio), 3200):
                await socket.send(audio[i:i+3200])
                await asyncio.sleep(0.1)
            await socket.send('{"type":"session.close"}')
        sender = asyncio.create_task(send())
        try:
            async with asyncio.timeout(90):
                async for message in socket:
                    print(message)
        finally:
            sender.cancel()

asyncio.run(main())

// Use from the console origin with a Speech Cloud project key.
// apiKey is a Speech Cloud project key, never a provider key.
const socket = new WebSocket("wss://api.speechinfra.com/v1/inference/nova-3-realtime/realtime", ['speech', apiKey]);
socket.onopen = () => socket.send("{\"language\":\"en\",\"smart_format\":false,\"diarize_model\":\"off\",\"interim_results\":false,\"endpointing\":10,\"vad_events\":false,\"type\":\"session.configure\"}");
socket.onmessage = ({data}) => {
  const event = JSON.parse(data);
  console.log(event);
  if (event.type === 'session.updated') {
    // Send mono PCM16 / 16 kHz binary frames, <= 30 s in total.
    // After the final frame: socket.send(JSON.stringify({type:'session.close'}));
  }
};
// For a complete file-streaming example: scripts/test_models.py

{
  "type": "object",
  "additionalProperties": false,
  "required": [],
  "properties": {
    "language": {
      "type": "string",
      "default": "en",
      "description": "Single-language recognition, or multi for code-switching between: en, es, fr, de, hi, ru, pt, ja, it, nl.",
      "enum": [
        "af",
        "ar",
        "hy",
        "as",
        "be",
        "bn",
        "bs",
        "bg",
        "ca",
        "zh",
        "hr",
        "cs",
        "da",
        "nl",
        "en",
        "et",
        "fi",
        "fr",
        "ka",
        "de",
        "el",
        "gu",
        "he",
        "hi",
        "hu",
        "id",
        "it",
        "ja",
        "kn",
        "kk",
        "ko",
        "lv",
        "lt",
        "mk",
        "ms",
        "mr",
        "mn",
        "ne",
        "no",
        "ps",
        "fa",
        "pl",
        "pt",
        "pa",
        "ro",
        "ru",
        "sr",
        "sk",
        "sl",
        "es",
        "sv",
        "tl",
        "ta",
        "te",
        "th",
        "tr",
        "uk",
        "ur",
        "vi",
        "af-ZA",
        "ar-AE",
        "ar-SA",
        "ar-QA",
        "ar-KW",
        "ar-SY",
        "ar-LB",
        "ar-PS",
        "ar-JO",
        "ar-EG",
        "ar-SD",
        "ar-TD",
        "ar-MA",
        "ar-DZ",
        "ar-TN",
        "ar-IQ",
        "ar-IR",
        "as-IN",
        "zh-HK",
        "zh-CN",
        "zh-TW",
        "cs-CZ",
        "da-DK",
        "en-US",
        "en-AU",
        "en-GB",
        "en-IN",
        "en-NZ",
        "nl-BE",
        "fr-CA",
        "ka-GE",
        "de-CH",
        "gu-IN",
        "kk-KZ",
        "ko-KR",
        "ps-AF",
        "pt-BR",
        "pt-PT",
        "pa-IN",
        "es-419",
        "sv-SE",
        "th-TH",
        "tr-TR",
        "multi"
      ]
    },
    "keyterms": {
      "type": "array",
      "items": {
        "type": "string",
        "minLength": 1,
        "maxLength": 200
      },
      "maxItems": 100
    },
    "smart_format": {
      "type": "boolean",
      "default": false
    },
    "diarize_model": {
      "type": "string",
      "enum": [
        "off",
        "latest",
        "v1"
      ],
      "default": "off"
    },
    "interim_results": {
      "type": "boolean",
      "default": false
    },
    "endpointing": {
      "type": "integer",
      "minimum": 0,
      "maximum": 10000,
      "default": 10
    },
    "utterance_end_ms": {
      "type": "integer",
      "minimum": 0,
      "maximum": 10000,
      "visibleWhen": {
        "interim_results": true
      }
    },
    "vad_events": {
      "type": "boolean",
      "default": false
    },
    "type": {
      "type": "string",
      "enum": [
        "session.configure"
      ],
      "default": "session.configure"
    }
  }
}

{
  "type": "object",
  "description": "JSON server events. Send binary mono PCM16 at 16 kHz after session.updated; session.close drains and returns session.closed.",
  "properties": {
    "type": {
      "enum": [
        "session.created",
        "session.updated",
        "session.routed",
        "transcript.partial",
        "transcript.final",
        "error",
        "session.closed"
      ]
    }
  }
}