curl -X POST 'https://api.elevenlabs.io/v1/text-to-voice/design?output_format=mp3_22050_32' \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
  "voice_description": "string",
  "model_id": "eleven_multilingual_ttv_v2",
  "text": "string",
  "auto_generate_text": false,
  "loudness": 0.5,
  "seed": 0,
  "guidance_scale": 5,
  "stream_previews": false,
  "should_enhance": false,
  "remixing_session_id": "string",
  "remixing_session_iteration_id": "string",
  "quality": 0,
  "reference_audio_base64": "string",
  "prompt_strength": 0
}'

# POST /v1/text-to-voice/design (ElevenLabs API)
Base URL: https://api.elevenlabs.io
Auth: xi-api-key: <ELEVENLABS_API_KEY>
Last verified: 2026-09-03

> Design A Voice.

## Critical Gotchas
* Authentication uses the non-standard `xi-api-key` header, not `Authorization: Bearer`. Sending a Bearer token returns HTTP 401.
* Text-to-speech responses are raw binary audio, not JSON. Write the body to a file (`--output speech.mp3`) rather than parsing it.

## cURL
curl -X POST 'https://api.elevenlabs.io/v1/text-to-voice/design?output_format=mp3_22050_32' \
  -H "xi-api-key: $ELEVENLABS_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"voice_description":"string","model_id":"eleven_multilingual_ttv_v2"}'

## Parameters
| Name | In | Type | Required | Description |
| :--- | :--- | :--- | :--- | :--- |
| voice_description | body | string | Yes | Description to use for the created vo... |
| output_format | query | string | No | Output format of the generated audio.... |
| model_id | body | string | No | Model to use for the voice generation... |

## Response (200 OK)
{"previews":[{}],"text":"string"}

---
Need an API key? Get one at https://elevenlabs.io/app/sign-up

POST

/v1/text-to-voice/design

ElevenLabs API

Design A Voice.

Base URL: https://api.elevenlabs.io
Auth: xi-api-key: <ELEVENLABS_API_KEY>
Last verified: 2026-09-03 · upstream hash matched

Actions

Agents: curl -H "Accept: text/markdown" this URL
→ 214 tokens · Vary: Accept

Critical gotchas

Authentication uses the non-standard xi-api-key header, not Authorization: Bearer. Sending a Bearer token returns HTTP 401.

Text-to-speech responses are raw binary audio, not JSON. Write the body to a file (--output speech.mp3) rather than parsing it.

cURL

curl -X POST 'https://api.elevenlabs.io/v1/text-to-voice/design?output_format=mp3_22050_32' \  -H "xi-api-key: $ELEVENLABS_API_KEY" \  -H 'Content-Type: application/json' \  -d '{  "voice_description": "string",  "model_id": "eleven_multilingual_ttv_v2",  "text": "string",  "auto_generate_text": false,  "loudness": 0.5,  "seed": 0,  "guidance_scale": 5,  "stream_previews": false,  "should_enhance": false,  "remixing_session_id": "string",  "remixing_session_iteration_id": "string",  "quality": 0,  "reference_audio_base64": "string",  "prompt_strength": 0}'

Get a free ElevenLabs API key → sponsored

Parameters

Name	In	Type	Required	Description
output_format	query	string	No	Output format of the generated audio. Formatted as codec_sample_rate_bitrate. So an mp3 with 22.05kHz sample rate at 32kbs is represented as mp3_22050_32. MP3 with 192kbps bitrate requires you to be subscribed to Creator tier or above. PCM with 44.1kHz sample rate requires you to be subscribed to Pro tier or above. Note that the μ-law format (sometimes written mu-law, often approximated as u-law) is commonly used for Twilio audio inputs.
voice_description	body	string	Yes	Description to use for the created voice.
model_id	body	string	No	Model to use for the voice generation. Possible values: eleven_multilingual_ttv_v2, eleven_ttv_v3.
text	body	string	No	Text to generate, text length has to be between 100 and 1000.
auto_generate_text	body	boolean	No	Whether to automatically generate a text suitable for the voice description.
loudness	body	number	No	Controls the volume level of the generated voice. -1 is quietest, 1 is loudest, 0 corresponds to roughly -24 LUFS.
seed	body	integer	No	Random number that controls the voice generation. Same seed with same inputs produces same voice.
guidance_scale	body	number	No	Controls how closely the AI follows the prompt. Lower numbers give the AI more freedom to be creative, while higher numbers force it to stick more to the prompt. High numbers can cause voice to sound artificial or robotic. We recommend to use longer, more detailed prompts at lower Guidance Scale.
stream_previews	body	boolean	No	Determines whether the Text to Voice previews should be included in the response. If true, only the generated IDs will be returned which can then be streamed via the /v1/text-to-voice/:generated_voice_id/stream endpoint.
should_enhance	body	boolean	No	Whether to enhance the voice description using AI to add more detail and improve voice generation quality. When enabled, the system will automatically expand simple prompts into more detailed voice descriptions. Defaults to False
remixing_session_id	body	string	No	The remixing session id.
remixing_session_iteration_id	body	string	No	The id of the remixing session iteration where these generations should be attached to. If not provided, a new iteration will be created.
quality	body	number	No	Higher quality results in better voice output but less variety.
reference_audio_base64	body	string	No	Reference audio to use for the voice generation. The audio should be base64 encoded. Only supported when using the eleven_ttv_v3 model.
prompt_strength	body	number	No	Controls the balance of prompt versus reference audio when generating voice samples. 0 means almost no prompt influence, 1 means almost no reference audio influence. Only supported when using the eleven_ttv_v3 model.

Response 200 OK

{
  "previews": [
    {
      "audio_base_64": "string",
      "generated_voice_id": "string",
      "media_type": "string",
      "duration_secs": 0,
      "language": "string"
    }
  ],
  "text": "string"
}