WebSockets

1
import asyncio
2
import websockets
3
import json
4
import base64
5
import pyaudio
6
# import os
7
8
9
API_KEY = "YOUR_API_KEY" # Or use os.getenv("MURF_API_KEY") if you have set the API key as an environment variable
10
WS_URL = "wss://global.api.murf.ai/v1/speech/stream-input"
11
PARAGRAPH = "With a single WebSocket connection, you can stream text input and receive synthesized audio continuously, without the overhead of repeated HTTP requests. This makes it ideal for use cases where your application sends or receives text in chunks and needs real-time audio to deliver a smooth, conversational experience"
12
13
# Audio format settings (must match your API output)
14
SAMPLE_RATE = 24000
15
CHANNELS = 1
16
FORMAT = pyaudio.paInt16
17
18
async def tts_stream():
19
  async with websockets.connect(
20
      f"{WS_URL}?api-key={API_KEY}&model=FALCON&sample_rate=24000&channel_type=MONO&format=WAV"
21
  ) as ws:
22
      # Send voice config first (optional)
23
      voice_config_msg = {
24
          "voice_config": {
25
              "voiceId": "Matthew",
26
              "multiNativeLocale":"en-US",
27
              "style": "Conversation",
28
              "rate": 0,
29
              "pitch": 0,
30
              "variation": 1
31
          }
32
      }
33
      print(f'Sending payload : {voice_config_msg}')
34
      await ws.send(json.dumps(voice_config_msg))
35
36
      # Send text in one go (or chunk if you want streaming)
37
      text_msg = {
38
          "text": PARAGRAPH,
39
          "end" : True # This will close the context. So you can re-run and concurrency is available.
40
      }
41
      print(f'Sending payload : {text_msg}')
42
      await ws.send(json.dumps(text_msg))
43
44
      # Setup audio stream
45
      pa = pyaudio.PyAudio()
46
      stream = pa.open(format=FORMAT, channels=CHANNELS, rate=SAMPLE_RATE, output=True)
47
48
      first_chunk = True
49
      try:
50
          while True:
51
              response = await ws.recv()
52
              data = json.loads(response)
53
              print(f'Received data:  {data}')
54
              if "audio" in data:
55
                  audio_bytes = base64.b64decode(data["audio"])
56
                  # Skip the first 44 bytes (WAV header) only for the first chunk
57
                  if first_chunk and len(audio_bytes) > 44:
58
                      audio_bytes = audio_bytes[44:]
59
                      first_chunk = False
60
                  stream.write(audio_bytes)
61
              if data.get("final"):
62
                  break
63
      finally:
64
          stream.stop_stream()
65
          stream.close()
66
          pa.terminate()
67
68
if __name__ == "__main__":
69
    asyncio.run(tts_stream())