How to Design a Custom Voice from a Text Prompt in ElevenLabs

Use Voice Design to generate a brand-new synthetic voice by describing age, accent, and tone in plain words.

7 minBeginner

Visit ElevenLabsAffiliate link. See our disclosure.

Voice Design creates a unique voice that does not belong to any real person, generated entirely from a written description. It is ideal when you want a consistent character voice for a brand, game, or audiobook and do not want to clone anyone. This guide designs a voice and saves it for reuse.

What you need

An ElevenLabs account (Voice Design is available on most plans)
A clear idea of the character: age, gender, accent, and mood
A short sentence the preview will speak

Step 1: Open Voice Design

Go to Voices, click Add a new voice, and pick Voice Design.

Step 2: Describe the voice

Write a description focused on traits: age band, gender, accent, pacing, and emotional tone. Be specific. A vague prompt produces a generic voice, while a detailed one gives you something distinctive.

voice-prompt.txt

A warm middle-aged British woman with a calm, reassuring tone. Speaks slowly and clearly, like a documentary narrator. Slight smile in the voice, no harshness.

ElevenLabs - Voice Design

Describe your voice:

[ warm middle-aged British woman, calm narrator... ]

Preview text:

[ The river wound slowly through the green valley. ]

[ Generate voices ]

Candidates: ( o 1 ) ( 2 ) ( 3 ) [ Save selected ]

A prompt plus a preview line generates several candidate voices.

Step 3: Add a preview line and generate

Enter a sentence for the preview to speak, then generate. ElevenLabs returns several candidate voices that all match the prompt differently. Play each one and compare.

Use a representative preview line

Pick preview text close to what the voice will actually say. A narration line previews very differently from an excited ad read, so test in context.

Step 4: Save your favourite

Select the candidate you like best, give it a name, and save it to your Voices library. From then on it behaves like any other voice and shows up in Text to Speech and Studio.

Result: a one-of-a-kind narrator voice you fully control, with no real-person likeness and no licensing worries.

Ready to try it yourself? Get set up withElevenLabsin a few minutes.(affiliate link. We may earn a commission at no extra cost. Disclosure)