Voice Design: Create Any Voice from a Description

Inventing a Voice from Scratch

Voice Cloning creates a replica of an existing voice. Voice Design does something different: it invents a new voice based on a text description you provide. Describe the age, gender, accent, and character of the voice you want, and ElevenLabs generates it for you. No audio sample required.

This is useful when you need a voice that does not exist: a fictional character for a game, a brand voice with a specific personality, or a narrator with a very specific combination of traits that you cannot find in the Voice Library.

How Voice Design Works

The Voice Design tool presents you with a text box. Write a description of the voice you want — think of it as writing a character brief for a casting director. The more specific your description, the more targeted the result. After submitting, ElevenLabs generates multiple voice candidates. You listen to each one, pick the best match, and save it to your Voice Library.

Important: each generation produces a random set of candidate voices. You cannot make incremental adjustments to a specific candidate — you can only generate new batches or accept what you have. If you get a result you love, save it immediately. There is no way to recover a candidate you did not save.

Writing Effective Voice Design Prompts

A good Voice Design prompt covers four dimensions:

• Demographics — Age range, gender presentation, nationality or regional background. Example: "30s, female, American Midwest."
• Vocal character — Tone, warmth, energy level. Example: "warm and measured, with a slight huskiness."
• Delivery style — How the voice communicates. Example: "authoritative but approachable, like a senior journalist."
• Use context — What the voice will be used for. Example: "for audiobook narration of literary fiction." This helps the model weight the right characteristics.

Example of a well-written prompt: A 45-year-old British male voice, calm and thoughtful, with a slight Edinburgh accent. Deep but not booming. Sounds like someone explaining a complex topic with genuine patience. For documentary narration.

Voice Design vs. Voice Library vs. Voice Cloning

Building a Voice Library for a Project

If you are producing a game, podcast, or any project with multiple recurring voices, Voice Design is a fast way to populate a cast. Design each character separately with a distinct brief, generate three to five candidates per character, and save the best one. Keep a notes file mapping character names to voice IDs — you will need those IDs when calling the API or building dialogue scripts.

Try This

Write a Voice Design prompt for a fictional character you care about — a narrator for a podcast you would like to make, a character in a story, or a voice assistant persona for a hypothetical app. Be specific on all four dimensions: demographics, vocal character, delivery style, and use context. Generate a batch of candidates and listen to each one. If none are right, rewrite your prompt and try again. Notice which parts of your description the model seems to weigh most heavily.