The official Python API for ElevenLabs text-to-speech software. Eleven brings the most compelling, rich and lifelike voices to creators and developers in just a few lines of code.
Check out the HTTP API documentation.
pip install elevenlabs
The SDK was rewritten in v1 and is now programmatically generated from our OpenAPI spec. As part of this release there are some breaking changes.
The SDK now exports a client class that you must instantiate to call various endpoints in our API.
from elevenlabs.client import ElevenLabs
client = ElevenLabs(
api_key="..." # Defaults to ELEVEN_API_KEY
)
As part of this change, there is no longer a set_api_key
and get_api_key
method exported.
The SDK now uses httpx under the hood. This allows us to export an async client in addition to a synchronous client. Note that you can pass in your own httpx client as well.
from elevenlabs.client import AsyncElevenLabs
client = AsyncElevenLabs(
api_key="...", # Defaults to ELEVEN_API_KEY
httpx_client=httpx.AsyncClient(...)
)
There are no longer static methods exposed directly on objects. For example,
instead of Models.from_api()
you can now do client.models.get_all()
.
The renames are specified below:
User.from_api()
-> client.users.get()
Models.from_api()
-> client.models.get_all()
Voices.from_api()
-> client.voices.get_all()
History.from_api()
-> client.history.get_all()
The SDK no longer exports top level functions generate
, clone
, and voices
. Instead,
everything is now directly attached to the client instance.
The generate method is a helper function that makes it easier to consume the
text-to-speech APIs. If you'd rather access the raw APIs, simply use client.text_to_speech
.
The clone method is a helper function that wraps the voices add and
get APIs. If you'd rather access the raw APIs, simply use client.voices.add()
.
To get all your voices, use client.voices.get_all()
.
The SDK continues to export the play
, stream
and save
methods. Under the hood, these methods
use ffmpeg and mpv to play audio streams.
from elevenlabs import play, stream, save
# plays audio using ffmpeg
play(audio)
# streams audio using mpv
stream(audio)
# saves audio to file
save(audio, "my-file.mp3")
We support two main models: the newest eleven_multilingual_v2
, a single foundational model supporting 29 languages including English, Chinese, Spanish, Hindi, Portuguese, French, German, Japanese, Arabic, Korean, Indonesian, Italian, Dutch, Turkish, Polish, Swedish, Filipino, Malay, Russian, Romanian, Ukrainian, Greek, Czech, Danish, Finnish, Bulgarian, Croatian, Slovak, and Tamil; and eleven_monolingual_v1
, a low-latency model specifically trained for English speech.
from elevenlabs import play
from elevenlabs.client import ElevenLabs
client = ElevenLabs(
api_key="YOUR_API_KEY", # Defaults to ELEVEN_API_KEY
)
audio = client.generate(
text="Hello! 你好! Hola! नमस्ते! Bonjour! こんにちは! مرحبا! 안녕하세요! Ciao! Cześć! Привіт! வணக்கம்!",
voice="Rachel",
model="eleven_multilingual_v2"
)
play(audio)
Play
Don't forget to unmute the player!
audio.3.webm
List all your available voices with voices()
.
from elevenlabs.client import ElevenLabs
client = ElevenLabs(
api_key="YOUR_API_KEY", # Defaults to ELEVEN_API_KEY
)
response = client.voices.get_all()
audio = client.generate(text="Hello there!", voice=response.voices[0])
print(response.voices)
Show output
[
Voice(
voice_id='21m00Tcm4TlvDq8ikWAM',
name='Rachel',
category='premade',
settings=None,
),
Voice(
voice_id='AZnzlk1XvdvUeBnXmlld',
name='Domi',
category='premade',
settings=None,
),
]
Build a voice object with custom settings to personalize the voice style, or call
client.voices.get_settings("your-voice-id")
to get the default settings for the voice.
from elevenlabs import Voice, VoiceSettings, play
from elevenlabs.client import ElevenLabs
client = ElevenLabs(
api_key="YOUR_API_KEY", # Defaults to ELEVEN_API_KEY
)
audio = client.generate(
text="Hello! My name is Bella.",
voice=Voice(
voice_id='EXAVITQu4vr4xnSDxMaL',
settings=VoiceSettings(stability=0.71, similarity_boost=0.5, style=0.0, use_speaker_boost=True)
)
)
play(audio)
Clone your voice in an instant. Note that voice cloning requires an API key, see below.
from elevenlabs.client import ElevenLabs
from elevenlabs import play
client = ElevenLabs(
api_key="YOUR_API_KEY", # Defaults to ELEVEN_API_KEY
)
voice = client.clone(
name="Alex",
description="An old American male voice with a slight hoarseness in his throat. Perfect for news", # Optional
files=["./sample_0.mp3", "./sample_1.mp3", "./sample_2.mp3"],
)
audio = client.generate(text="Hi! I'm a cloned voice!", voice=voice)
play(audio)
Stream audio in real-time, as it's being generated.
from elevenlabs.client import ElevenLabs
from elevenlabs import stream
client = ElevenLabs(
api_key="YOUR_API_KEY", # Defaults to ELEVEN_API_KEY
)
audio_stream = client.generate(
text="This is a... streaming voice!!",
stream=True
)
stream(audio_stream)
Note that generate
is a helper function. If you'd like to access
the raw method, simply use client.text_to_speech.convert_as_stream
.
Stream text chunks into audio as it's being generated, with <1s latency. Note: if chunks don't end with space or punctuation (" ", ".", "?", "!"), the stream will wait for more text.
from elevenlabs.client import ElevenLabs
from elevenlabs import stream
client = ElevenLabs(
api_key="YOUR_API_KEY", # Defaults to ELEVEN_API_KEY
)
def text_stream():
yield "Hi there, I'm Eleven "
yield "I'm a text to speech API "
audio_stream = client.generate(
text=text_stream(),
voice="Nicole",
model="eleven_monolingual_v1",
stream=True
)
stream(audio_stream)
Note that generate
is a helper function. If you'd like to access
the raw method, simply use client.text_to_speech.convert_realtime
.
Use AsyncElevenLabs
if you want to make API calls asynchronously.
import asyncio
from elevenlabs.client import AsyncElevenLabs
eleven = AsyncElevenLabs(
api_key="MY_API_KEY" # Defaults to ELEVEN_API_KEY
)
async def print_models() -> None:
models = await eleven.models.get_all()
print(models)
asyncio.run(print_models())
All of the ElevenLabs models are nested within the elevenlabs module.
We support 29 languages and 100+ accents. Explore all languages.
While we value open-source contributions to this SDK, this library is generated programmatically. Additions made directly to this library would have to be moved over to our generation code, otherwise they would be overwritten upon the next generated release. Feel free to open a PR as a proof of concept, but know that we will not be able to merge it as-is. We suggest opening an issue first to discuss with us!
On the other hand, contributions to the README are always very welcome!