• Text-to-Speech REST
  • Getting Started
  • Feature Overview
  • Starter Apps
• Text-to-Speech Streaming
  • Getting Started
  • Feature Overview
  • Starter Apps
  • Control Messages
• Models and Languages
  • Voices and Languages
• Media Output Settings
  • Encoding
  • Bit Rate
  • Container
  • Sample Rate
• Results Processing
  • TTS Tagging
  • TTS Callback
  • Audio Output Streaming
• Tips and Tricks
  • Real-Time TTS with WebSockets
  • Text Chunking for TTS
  • Formatting text for Aura-2
  • Handling Audio Issues in Text To Speech
  • Sending LLM Outputs to a WebSocket
  • Text Chunking for TTS REST Optimization
  • Text to Speech Latency
  • Text to Speech Prompting
  • TTS Troubleshooting WebSocket, NET, and DATA Errors

• Why Use WebSockets for TTS?
• WebSocket Implementation Examples

1
# For help migrating to the new Python SDK, check out our migration guide:
2
# https://github.com/deepgram/deepgram-python-sdk/blob/main/docs/Migrating-v3-to-v5.md
3
4
import sounddevice as sd
5
import numpy as np
6
import time
7
8
from deepgram import DeepgramClient
9
from deepgram.core.events import EventType
10
from deepgram.extensions.types.sockets import SpeakV1SocketClientResponse
11
12
TTS_TEXT = "Hello, this is a text to speech example using Deepgram."
13
14
def main():
15
    try:
16
        # Create a Deepgram client using the API key from environment variables
17
        deepgram = DeepgramClient()
18
19
        # Create a websocket connection to Deepgram
20
        with deepgram.speak.v1.connect(
21
            model="aura-2-thalia-en",
22
            encoding="linear16",
23
            sample_rate=48000
24
        ) as dg_connection:
25
26
            def on_message(message: SpeakV1SocketClientResponse) -> None:
27
                if isinstance(message, bytes):
28
                    print("Received audio chunk")
29
                    # Convert binary data to audio format playback devices understand
30
                    array = np.frombuffer(message, dtype=np.int16)
31
                    # Play the audio immediately upon receiving each chunk
32
                    sd.play(array, 48000)
33
                    sd.wait()
34
                else:
35
                    msg_type = getattr(message, "type", "Unknown")
36
                    print(f"Received {msg_type} event")
37
38
            dg_connection.on(EventType.OPEN, lambda _: print("Connection opened"))
39
            dg_connection.on(EventType.MESSAGE, on_message)
40
            dg_connection.on(EventType.CLOSE, lambda _: print("Connection closed"))
41
            dg_connection.on(EventType.ERROR, lambda error: print(f"Error: {error}"))
42
43
            dg_connection.start_listening()
44
45
            # Send text to be converted to speech
46
            from deepgram.extensions.types.sockets import SpeakV1TextMessage
47
            dg_connection.send_text(SpeakV1TextMessage(text=TTS_TEXT))
48
49
            # Send control messages
50
            from deepgram.extensions.types.sockets import SpeakV1ControlMessage
51
            dg_connection.send_control(SpeakV1ControlMessage(type="Flush"))
52
53
            # Allow time for playback
54
            time.sleep(5)
55
56
            dg_connection.send_control(SpeakV1ControlMessage(type="Close"))
57
            print("TTS stream completed")
58
59
    except Exception as e:
60
        print(f"An error occurred: {e}")
61
62
if __name__ == "__main__":
63
    main()