Elevenlabs Voice

ElevenLabs Voice AI Skill

Ultra-low latency text-to-speech and AI sound effects generation via ElevenLabs API.

Capabilities

Authentication

All requests require the xi-api-key header:

export ELEVENLABS_API_KEY="your-api-key"

Flash v2.5 Text-to-Speech (Streaming)

Endpoint: POST https://api.elevenlabs.io/v1/text-to-speech/{voice_id}/stream

Best for real-time applications with ~75ms latency.

Basic Usage

curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM/stream" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello! How can I help you today?",
    "model_id": "eleven_flash_v2_5",
    "voice_settings": {
      "stability": 0.5,
      "similarity_boost": 0.75
    }
  }' \
  --output speech.mp3

With Output Format Options

# MP3 (default, best compatibility)
curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM/stream?output_format=mp3_44100_128" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "This is high quality MP3 audio.",
    "model_id": "eleven_flash_v2_5"
  }' \
  --output speech.mp3

# PCM (raw audio for real-time processing)
curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM/stream?output_format=pcm_24000" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Raw PCM for Twilio streaming.",
    "model_id": "eleven_flash_v2_5"
  }' \
  --output speech.pcm

# μ-law (for telephony/Twilio)
curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM/stream?output_format=ulaw_8000" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Optimized for phone calls.",
    "model_id": "eleven_flash_v2_5"
  }' \
  --output speech.ulaw

Request Body Parameters

Output Formats

Popular Voice IDs

List All Voices

curl -X GET "https://api.elevenlabs.io/v1/voices" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" | jq '.voices[] | {name, voice_id}'

AI Sound Effects Generation

Endpoint: POST https://api.elevenlabs.io/v1/sound-generation

Generate sound effects from text descriptions.

Basic Usage

curl -X POST "https://api.elevenlabs.io/v1/sound-generation" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "A thunderstorm with heavy rain and distant thunder",
    "duration_seconds": 10
  }' \
  --output thunder.mp3

With All Options

curl -X POST "https://api.elevenlabs.io/v1/sound-generation" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Futuristic sci-fi door opening with hydraulic hiss",
    "duration_seconds": 5,
    "prompt_influence": 0.5
  }' \
  --output scifi_door.mp3

Request Body Parameters

Sound Effect Prompt Examples

# Notification sounds
"A pleasant notification chime, bright and short"
"Soft bubble pop notification sound"
"Futuristic UI confirmation beep"

# Environmental
"Forest ambience with birds chirping and wind in leaves"
"Busy coffee shop background noise with chatter"
"Ocean waves gently crashing on a sandy beach"

# Actions
"Footsteps on wooden floor, slow walking pace"
"Car engine starting and idling"
"Glass shattering on concrete floor"

# Game/UI
"8-bit retro game jump sound"
"Level up fanfare, triumphant and short"
"Health pickup sound, magical sparkle"

Twilio Integration Pattern

For phone conversations with ElevenLabs + Twilio:

import httpx
import base64

async def generate_speech_for_twilio(text: str, voice_id: str = "21m00Tcm4TlvDq8ikWAM") -> bytes:
    """Generate μ-law audio for Twilio streaming."""
    async with httpx.AsyncClient() as client:
        response = await client.post(
            f"https://api.elevenlabs.io/v1/text-to-speech/{voice_id}/stream",
            params={"output_format": "ulaw_8000"},
            headers={
                "xi-api-key": os.environ["ELEVENLABS_API_KEY"],
                "Content-Type": "application/json"
            },
            json={
                "text": text,
                "model_id": "eleven_flash_v2_5",
                "voice_settings": {
                    "stability": 0.5,
                    "similarity_boost": 0.75
                }
            }
        )
        return response.content

# For Twilio Media Streams, base64 encode the audio
audio_bytes = await generate_speech_for_twilio("Hello, how can I help?")
audio_base64 = base64.b64encode(audio_bytes).decode()

Error Handling

Cost Optimization

• Flash v2.5: Cheaper per character than Multilingual v2
• Caching: Cache generated audio for repeated phrases
• Text chunking: For long text, chunk at natural sentence breaks
• Voice settings: Default settings are optimized for most use cases

Quick Test

# Test TTS (saves to /tmp/elevenlabs_test.mp3)
curl -X POST "https://api.elevenlabs.io/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM/stream" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "ElevenLabs Flash is working!", "model_id": "eleven_flash_v2_5"}' \
  --output /tmp/elevenlabs_test.mp3 && echo "Success! Audio saved to /tmp/elevenlabs_test.mp3"

# Test Sound Effects (saves to /tmp/elevenlabs_sfx.mp3)
curl -X POST "https://api.elevenlabs.io/v1/sound-generation" \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "A happy notification chime", "duration_seconds": 2}' \
  --output /tmp/elevenlabs_sfx.mp3 && echo "Success! Sound effect saved to /tmp/elevenlabs_sfx.mp3"