Overview
Instant Voice Cloning (IVC) allows you to create a custom voice clone from audio samples. This feature requires an API key and enables you to generate voices that match the characteristics of your provided samples.
Basic Voice Cloning
Create a voice clone with multiple audio samples:
from elevenlabs.client import ElevenLabs
client = ElevenLabs(
api_key="YOUR_API_KEY"
)
voice = client.voices.ivc.create(
name="Alex",
description="An old American male voice with a slight hoarseness in his throat. Perfect for news",
files=["./sample_0.mp3", "./sample_1.mp3", "./sample_2.mp3"]
)
print(f"Voice created with ID: {voice.voice_id}")
Parameters
The name that identifies this voice. This will be displayed in the dropdown of the website.
List of audio file paths to use for voice cloning. Multiple samples improve quality.
A description of the voice to help identify it later.
If set, will remove background noise from voice samples using the audio isolation model. If the samples do not include background noise, it can make the quality worse.
Labels for the voice. Keys can be language, accent, gender, or age.
With Background Noise Removal
Use the audio isolation model to clean up samples:
voice = client.voices.ivc.create(
name="Sarah",
description="Professional female voice",
files=["./recording1.mp3", "./recording2.mp3"],
remove_background_noise=True
)
With Voice Labels
Add metadata labels to organize your voices:
voice = client.voices.ivc.create(
name="Marcus",
description="British narrator",
files=["./voice_sample.mp3"],
labels={
"accent": "british",
"gender": "male",
"age": "middle_aged",
"language": "english"
}
)
Using the Cloned Voice
Once created, use the voice ID for text-to-speech:
from elevenlabs.client import ElevenLabs
from elevenlabs.play import play
client = ElevenLabs(api_key="YOUR_API_KEY")
# Clone the voice
voice = client.voices.ivc.create(
name="Custom Voice",
files=["./sample.mp3"]
)
# Use it for TTS
audio = client.text_to_speech.convert(
text="Hello, this is my cloned voice speaking.",
voice_id=voice.voice_id,
model_id="eleven_multilingual_v2"
)
play(audio)
Async Voice Cloning
For async operations:
import asyncio
from elevenlabs.client import AsyncElevenLabs
async def clone_voice():
client = AsyncElevenLabs(api_key="YOUR_API_KEY")
voice = await client.voices.ivc.create(
name="Async Clone",
description="Voice created asynchronously",
files=["./sample1.mp3", "./sample2.mp3"]
)
return voice.voice_id
voice_id = asyncio.run(clone_voice())
print(f"Voice ID: {voice_id}")
Best Practices
For best results:
- Provide 3-5 high-quality audio samples
- Each sample should be 30 seconds to 2 minutes long
- Use clear audio with minimal background noise
- Samples should contain varied speech patterns
- Ensure consistent audio quality across samples
Only use audio from speakers who have given explicit consent for voice cloning.
Response Object
The create method returns an AddVoiceIvcResponseModel with:
voice_id - The unique identifier for the cloned voice- Other voice metadata
Error Handling
try:
voice = client.voices.ivc.create(
name="Test Voice",
files=["./sample.mp3"]
)
print(f"Success! Voice ID: {voice.voice_id}")
except Exception as e:
print(f"Error creating voice: {e}")