Skip to main content
The ElevenLabs Python SDK provides powerful text-to-speech capabilities that convert written text into natural-sounding audio using advanced AI models.

​
Basic Usage

Convert text to speech with the convert() method: 
from elevenlabs.client import ElevenLabs
from elevenlabs.play import play

client = ElevenLabs(
    api_key="YOUR_API_KEY"
)

audio = client.text_to_speech.convert(
    text="The first move is what sets everything in motion.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_v3",
    output_format="mp3_44100_128"
)

play(audio)

​
Output Formats

The SDK supports various audio output formats formatted as codec_sample_rate_bitrate: 
audio = client.text_to_speech.convert(
    text="High quality MP3 audio.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    output_format="mp3_44100_128"  # MP3, 44.1kHz, 128kbps
)
  • MP3 with 192kbps requires Creator tier or above
  • PCM and WAV with 44.1kHz require Pro tier or above
  • μ-law format is commonly used for Twilio audio inputs

​
Model Selection

Choose from different models optimized for various use cases: 
# Eleven v3 - Dramatic delivery, 70+ languages
audio = client.text_to_speech.convert(
    text="Experience dramatic performances.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_v3"
)

# Eleven Multilingual v2 - Stability and accuracy
audio = client.text_to_speech.convert(
    text="Stability across 29 languages.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_multilingual_v2"
)

# Eleven Flash v2.5 - Ultra-low latency
audio = client.text_to_speech.convert(
    text="Fast generation with low latency.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_flash_v2_5"
)

# Eleven Turbo v2.5 - Balanced speed and quality
audio = client.text_to_speech.convert(
    text="Great balance for developers.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_turbo_v2_5"
)

​
Voice Settings

Customize voice characteristics with the VoiceSettings parameter: 
from elevenlabs.types import VoiceSettings

audio = client.text_to_speech.convert(
    text="Customized voice output.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_v3",
    voice_settings=VoiceSettings(
        stability=0.5,
        similarity_boost=0.75
    )
)

​
Advanced Options

​
Latency Optimization

Optimize streaming latency at some cost of quality: 
audio = client.text_to_speech.convert(
    text="Optimized for low latency.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    optimize_streaming_latency=3  # 0-4, where 4 is max optimization
)
​
optimize_streaming_latency
int
  • 0 - Default mode (no latency optimizations)
  • 1 - Normal optimizations (~50% improvement)
  • 2 - Strong optimizations (~75% improvement)
  • 3 - Max latency optimizations
  • 4 - Max with text normalizer off (best latency, may mispronounce numbers/dates)

​
Language Control

Enforce a specific language using ISO 639-1 codes: 
audio = client.text_to_speech.convert(
    text="Bonjour le monde",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_multilingual_v2",
    language_code="fr"  # French
)

​
Deterministic Generation

Use a seed for reproducible results: 
audio = client.text_to_speech.convert(
    text="Same seed produces same audio.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    seed=12345  # Integer between 0 and 4294967295
)

​
Context for Continuity

Improve speech continuity when generating multiple clips: 
# First generation
audio1 = client.text_to_speech.convert(
    text="This is the first sentence.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    next_text="This is the second sentence."
)

# Second generation with context
audio2 = client.text_to_speech.convert(
    text="This is the second sentence.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    previous_text="This is the first sentence."
)

​
Saving Audio Files

Save generated audio to a file: 
from elevenlabs.play import save

audio = client.text_to_speech.convert(
    text="Save this audio to disk.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    output_format="mp3_44100_128"
)

save(audio, "output.mp3")

​
Timestamps

Get character-level timing information for audio-text synchronization: 
response = client.text_to_speech.convert_with_timestamps(
    text="This is a test for the API of ElevenLabs.",
    voice_id="21m00Tcm4TlvDq8ikWAM",
    output_format="mp3_44100_128"
)

# Access audio data and alignment info
audio_data = response.audio
alignment = response.alignment

​
Async Usage

Use the async client for non-blocking operations: 
import asyncio
from elevenlabs.client import AsyncElevenLabs

client = AsyncElevenLabs(
    api_key="YOUR_API_KEY"
)

async def generate_speech():
    audio = await client.text_to_speech.convert(
        text="Async text-to-speech generation.",
        voice_id="JBFqnCBsd6RMkjVDRZzb",
        model_id="eleven_multilingual_v2"
    )
    # Process audio bytes
    async for chunk in audio:
        print(f"Received {len(chunk)} bytes")

asyncio.run(generate_speech())

Next Steps

Learn how to stream audio in real-time for lower latency

Build docs developers (and LLMs) love

Get started for freeTalk to us