Skip to main content
ElevenLabs offers multiple state-of-the-art TTS models, each optimized for different use cases, languages, and performance requirements.

​
Listing Available Models

Retrieve all available models for your account: 
from elevenlabs.client import ElevenLabs

client = ElevenLabs(
    api_key="YOUR_API_KEY"
)

models = client.models.list()

for model in models:
    print(f"Model: {model.model_id}")
    print(f"Name: {model.name}")
    print(f"Languages: {len(model.languages)} supported")
    print(f"Can do TTS: {model.can_do_text_to_speech}")
    print("---")

​
Main Models Overview

ElevenLabs provides four main TTS models, each with unique characteristics:

Eleven v3

Model ID: eleven_v3Dramatic delivery and performances with support for 70+ languages and natural multi-speaker dialogue.

Eleven Multilingual v2

Model ID: eleven_multilingual_v2Excels in stability, language diversity, and accent accuracy across 29 languages. Recommended for most use cases.

Eleven Flash v2.5

Model ID: eleven_flash_v2_5Ultra-low latency with support for 32 languages. Faster model at 50% lower price per character.

Eleven Turbo v2.5

Model ID: eleven_turbo_v2_5Good balance of quality and latency, ideal for developer use cases where speed is crucial. Supports 32 languages.

​
Eleven v3

The latest generation model with dramatic performances and extensive language support. 
audio = client.text_to_speech.convert(
    text="Experience the most dramatic and lifelike performances.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_v3",
    output_format="mp3_44100_128"
)

​
Key Features

Eleven v3 excels at emotional expression and dramatic performances, making it ideal for:
  • Audiobook narration
  • Character voices
  • Storytelling
  • Expressive dialogue
The most multilingual model with support for over 70 languages, providing global reach for your applications.
Natural multi-speaker dialogue capabilities for:
  • Conversational AI
  • Podcast generation
  • Interview simulations
  • Interactive narratives
Eleven v3 may have higher latency compared to Flash or Turbo models. Not recommended for real-time streaming applications where low latency is critical.

​
Eleven Multilingual v2

The recommended model for most production use cases, balancing quality, stability, and language support. 
audio = client.text_to_speech.convert(
    text="The first move is what sets everything in motion.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_multilingual_v2",
    output_format="mp3_44100_128"
)

​
Key Features

  • Stability: Consistent voice quality across generations
  • Accuracy: Excellent accent and pronunciation accuracy
  • 29 Languages: Broad language support for global applications
  • Reliability: Proven performance in production environments

​
Use Cases

Content Creation

  • Video voiceovers
  • E-learning modules
  • Marketing materials
  • Product demos

Applications

  • Mobile apps
  • Web applications
  • IVR systems
  • Accessibility tools

​
Eleven Flash v2.5

Ultra-low latency model optimized for speed and cost efficiency. 
audio = client.text_to_speech.convert(
    text="Fast generation with minimal latency.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_flash_v2_5",
    output_format="mp3_22050_32"  # Lower format for even faster streaming
)

​
Key Features

  • Ultra-Low Latency: Fastest generation times
  • Cost Effective: 50% lower price per character
  • 32 Languages: Broad language support
  • Optimized for Streaming: Ideal for real-time applications

​
Best For

1

Real-Time Streaming

Low-latency audio streaming for conversational AI and live applications
2

High-Volume Processing

Batch processing large amounts of text with cost constraints
3

Prototyping

Rapid development and testing with quick iteration cycles
# Optimized for streaming
audio_stream = client.text_to_speech.stream(
    text="Real-time streaming with Flash v2.5",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_flash_v2_5",
    optimize_streaming_latency=4
)

​
Eleven Turbo v2.5

Balanced model providing excellent quality-to-speed ratio for developer applications. 
audio = client.text_to_speech.convert(
    text="Great balance of quality and latency.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_turbo_v2_5",
    output_format="mp3_44100_128"
)

​
Key Features

  • Balanced Performance: Good quality with reduced latency
  • Developer-Focused: Optimized for common development scenarios
  • 32 Languages: Wide language coverage
  • Versatile: Suitable for most application types

​
Ideal Use Cases

  • Conversational AI: Chatbots and virtual assistants
  • Gaming: Character dialogue and narration
  • Notifications: Audio alerts and announcements
  • Interactive Systems: IVR and phone systems

​
Model Comparison

​
Model Characteristics
comparison
Featurev3Multilingual v2Flash v2.5Turbo v2.5
QualityHighestHighGoodHigh
LatencyHigherMediumLowestLow
Languages70+293232
PriceStandardStandard50% lowerStandard
Best ForDramatic contentProductionReal-time/CostDeveloper apps
Multi-speakerYesLimitedNoLimited

​
Choosing the Right Model

1

Assess Your Requirements

Determine your priorities: quality, latency, cost, or language support
2

Consider Your Use Case

  • Content creation → Multilingual v2 or v3
  • Real-time streaming → Flash v2.5 or Turbo v2.5
  • High-volume processing → Flash v2.5
  • Dramatic narration → v3
3

Test and Compare

models_to_test = [
    "eleven_v3",
    "eleven_multilingual_v2",
    "eleven_flash_v2_5",
    "eleven_turbo_v2_5"
]

for model_id in models_to_test:
    audio = client.text_to_speech.convert(
        text="Compare model outputs.",
        voice_id="JBFqnCBsd6RMkjVDRZzb",
        model_id=model_id
    )
    # Save and compare
    save(audio, f"output_{model_id}.mp3")

​
Language Support

Check which languages a model supports: 
models = client.models.list()

for model in models:
    if model.model_id == "eleven_v3":
        print(f"Languages supported by {model.name}:")
        for lang in model.languages:
            print(f"  - {lang.name} ({lang.language_id})")

​
Model-Specific Settings

Some models support additional parameters: 
# Using language code enforcement
audio = client.text_to_speech.convert(
    text="Bonjour le monde",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_multilingual_v2",
    language_code="fr"  # Enforce French
)

# Optimizing for streaming with Turbo
audio_stream = client.text_to_speech.stream(
    text="Optimized streaming.",
    voice_id="JBFqnCBsd6RMkjVDRZzb",
    model_id="eleven_turbo_v2_5",
    optimize_streaming_latency=3
)

​
Async Model Operations

List models asynchronously: 
import asyncio
from elevenlabs.client import AsyncElevenLabs

client = AsyncElevenLabs(
    api_key="YOUR_API_KEY"
)

async def print_models():
    models = await client.models.list()
    for model in models:
        print(f"{model.name}: {model.model_id}")

asyncio.run(print_models())

​
Best Practices

  • Use Multilingual v2 for stable, production-ready applications
  • Test with your actual content before committing to a model
  • Monitor usage and costs across different models
  • Consider fallback strategies if a model is unavailable
  • Choose Flash v2.5 or Turbo v2.5 for low-latency requirements
  • Combine with optimize_streaming_latency parameter
  • Use lower output formats (22050Hz, 32kbps) to reduce bandwidth
  • Implement proper error handling and retry logic
  • Use Flash v2.5 for high-volume processing (50% cost savings)
  • Cache generated audio when possible
  • Monitor character usage across models
  • Batch requests when real-time isn’t required
  • Use v3 for the highest quality and dramatic expression
  • Use higher output formats (44100Hz, 128kbps+)
  • Fine-tune voice settings for each model
  • Test different models with your specific content

​
Additional Resources

Models Documentation

Detailed information about all models and languages

Voice Lab

Try different models with various voices
For the most up-to-date information about model capabilities, pricing, and language support, visit the ElevenLabs Models documentation.

Build docs developers (and LLMs) love

Get started for freeTalk to us