Streaming text to speech

1
import os
2
import uuid
3
from dotenv import load_dotenv
4
from elevenlabs import VoiceSettings
5
from elevenlabs.client import ElevenLabs
6
7
load_dotenv()
8
9
ELEVENLABS_API_KEY = os.getenv("ELEVENLABS_API_KEY")
10
elevenlabs = ElevenLabs(
11
    api_key=ELEVENLABS_API_KEY,
12
)
13
14
15
def text_to_speech_file(text: str) -> str:
16
    # Calling the text_to_speech conversion API with detailed parameters
17
    response = elevenlabs.text_to_speech.convert(
18
        voice_id="pNInz6obpgDQGcFmaJgB", # Adam pre-made voice
19
        output_format="mp3_22050_32",
20
        text=text,
21
        model_id="eleven_flash_v2_5", # use the flash model for low latency
22
        # Optional voice settings that allow you to customize the output
23
        voice_settings=VoiceSettings(
24
            stability=0.0,
25
            similarity_boost=1.0,
26
            style=0.0,
27
            use_speaker_boost=True,
28
            speed=1.0,
29
        ),
30
    )
31
32
    # uncomment the line below to play the audio back
33
    # play(response)
34
35
    # Generating a unique file name for the output MP3 file
36
    save_file_path = f"{uuid.uuid4()}.mp3"
37
38
    # Writing the audio to a file
39
    with open(save_file_path, "wb") as f:
40
        for chunk in response:
41
            if chunk:
42
                f.write(chunk)
43
44
    print(f"{save_file_path}: A new audio file was saved successfully!")
45
46
    # Return the path of the saved audio file
47
    return save_file_path

1
text_to_speech_file("Hello World")

1
import os
2
from typing import IO
3
from io import BytesIO
4
from dotenv import load_dotenv
5
from elevenlabs import VoiceSettings
6
from elevenlabs.client import ElevenLabs
7
8
load_dotenv()
9
10
ELEVENLABS_API_KEY = os.getenv("ELEVENLABS_API_KEY")
11
elevenlabs = ElevenLabs(
12
    api_key=ELEVENLABS_API_KEY,
13
)
14
15
16
def text_to_speech_stream(text: str) -> IO[bytes]:
17
    # Perform the text-to-speech conversion
18
    response = elevenlabs.text_to_speech.stream(
19
        voice_id="pNInz6obpgDQGcFmaJgB", # Adam pre-made voice
20
        output_format="mp3_22050_32",
21
        text=text,
22
        model_id="eleven_multilingual_v2",
23
        # Optional voice settings that allow you to customize the output
24
        voice_settings=VoiceSettings(
25
            stability=0.0,
26
            similarity_boost=1.0,
27
            style=0.0,
28
            use_speaker_boost=True,
29
            speed=1.0,
30
        ),
31
    )
32
33
    # Create a BytesIO object to hold the audio data in memory
34
    audio_stream = BytesIO()
35
36
    # Write each chunk of audio data to the stream
37
    for chunk in response:
38
        if chunk:
39
            audio_stream.write(chunk)
40
41
    # Reset stream position to the beginning
42
    audio_stream.seek(0)
43
44
    # Return the stream for further use
45
    return audio_stream

1
text_to_speech_stream("This is James")