ElevenLabs


Prefix	`elevenlabs`
Default model	`eleven_multilingual_v2`
Env var	`ELEVENLABS_API_KEY`
Official docs	elevenlabs.io/docs

Models

Model	Streaming	Audio Tags	Notes
`eleven_v3`	Yes	Yes (passthrough)	Most expressive; supports all bracket tags
`eleven_multilingual_v2`	Yes	No	Default; stable multilingual
`eleven_flash_v2_5`	Yes	No	Low-latency, multilingual
`eleven_flash_v2`	Yes	No	Low-latency, English

Usage

import { generateSpeech } from "@speech-sdk/core"

const result = await generateSpeech({
  model: "elevenlabs/eleven_v3",
  text: "Hello from SpeechSDK!",
  voice: "EXAVITQu4vr4xnSDxMaL",
})

The voice is an ElevenLabs voice ID — find IDs in the ElevenLabs voice library.

Audio Tags

eleven_v3 supports SpeechSDK's standardized audio tags natively via bracket syntax:

await generateSpeech({
  model: "elevenlabs/eleven_v3",
  text: "[laugh] That's hilarious! [sigh] But really though.",
  voice: "EXAVITQu4vr4xnSDxMaL",
})

Tags are passed straight through to the API. Other ElevenLabs models strip tags and return warnings.

Provider Options

await generateSpeech({
  model: "elevenlabs/eleven_v3",
  text: "Hello!",
  voice: "EXAVITQu4vr4xnSDxMaL",
  providerOptions: {
    output_format: "mp3_44100_128",
    enable_logging: false,
    optimize_streaming_latency: 3,
    voice_settings: {
      stability: 0.5,
      similarity_boost: 0.8,
      style: 0.2,
      use_speaker_boost: true,
    },
  },
})

output_format, enable_logging, and optimize_streaming_latency are sent as query parameters. voice_settings is sent in the JSON body.

Custom Configuration

import { generateSpeech } from "@speech-sdk/core"
import { createElevenLabs } from "@speech-sdk/core/providers"

const elevenlabs = createElevenLabs({
  apiKey: process.env.ELEVENLABS_API_KEY,
})

const result = await generateSpeech({
  model: elevenlabs("eleven_v3"),
  text: "[whispers] Hello from SpeechSDK.",
  voice: "EXAVITQu4vr4xnSDxMaL",
})