ElevenLabs Voice Design — Generate Voices from Prompts

Quick Use

1. POST /v1/voice-generation/generate-voice with description + preview text
2. POST /v1/voice-generation/create-voice to save to library
3. Use the saved voice_id in standard TTS calls

Intro

ElevenLabs Voice Design generates entirely new synthesized voices from a text prompt — describe the age, accent, gender, energy, and timbre, get back a unique voice you own and can reuse across the TTS API. No source audio required, no clone consent issues. Best for: characters in audio fiction, podcast hosts, app personas where you don't have a real reference voice, multilingual voices in languages with thin voice libraries. Works with: ElevenLabs Voice Lab dashboard or API. Setup time: 5 minutes.

Generate a voice via API

import requests, json

resp = requests.post(
    "https://api.elevenlabs.io/v1/voice-generation/generate-voice",
    headers={"xi-api-key": os.environ["ELEVENLABS_API_KEY"]},
    json={
        "voice_description": "An elderly Scottish man with a warm, smoky voice — like a retired sea captain telling stories by a fireplace",
        "text": "Aye, that was the night the lighthouse went dark, and we were three miles off the rocks.",
    },
)
generated_voice_id = resp.headers["voice_id"]
with open("preview.mp3", "wb") as f:
    f.write(resp.content)

Save the design to your voice library

requests.post(
    "https://api.elevenlabs.io/v1/voice-generation/create-voice",
    headers={"xi-api-key": API_KEY},
    json={
        "voice_name": "Captain Hamish",
        "voice_description": "Elderly Scottish sea captain, warm and smoky",
        "generated_voice_id": generated_voice_id,
        "labels": {"accent": "Scottish", "age": "elderly", "use": "narration"},
    },
)

Use the saved voice in TTS

from elevenlabs.client import ElevenLabs
from elevenlabs import play

client = ElevenLabs()
audio = client.text_to_speech.convert(
    voice_id="Captain Hamish",        # name or ID from your library
    model_id="eleven_turbo_v2_5",
    text="Ah, you wouldn't believe what we saw that morning on Skye.",
)
play(audio)

Prompt patterns that work

Cost

Voice Design generation costs ~1,000 credits per call (preview); saving a voice is free; TTS usage is the standard per-character rate. Starter plan ($5/mo) covers ~50 designs.

FAQ

Q: Are these voices royalty-free for commercial use? A: Yes — voices you generate via Voice Design are yours under the ElevenLabs commercial license tied to your plan. Read the current Terms — Creator plan and above explicitly allow commercial publishing. Free plan is non-commercial only.

Q: Can I tweak a voice after generating? A: Yes — saved voices have stability and similarity sliders. For more dramatic changes, re-prompt Voice Design with refined description text. The platform doesn't let you literally re-mix two designed voices, but iterating on prompts gets close.

Q: How does this compare to cloning a real voice? A: Voice Cloning needs 1-3 minutes of source audio and consent; outputs sound very close to the source. Voice Design needs only a prompt; outputs are novel synthetic voices. Use Design when you don't have a source or for ethically clean characters.

Source & Thanks

Built by ElevenLabs. Voice Design docs at elevenlabs.io/docs/voices/voice-lab.

elevenlabs/elevenlabs-python — official SDK