Skip to content
eWiki The practical AI reference

AI Tool Questions

ElevenLabs Voice Design: How to Describe, Audition, Refine, and Save a Synthetic Voice

Black studio microphone with a round pop filter on an adjustable arm above a white desk, with a cup, blue notebook, and padded wall panels.

ElevenLabs Voice Design creates a synthetic voice from a written description. It is intended for cases where the Voice Library lacks a suitable option, with a workflow for describing a voice, generating three previews, auditioning the candidates, refining the prompt, and saving one result.

What ElevenLabs Voice Design Is and When to Use It

Voice Design is ElevenLabs’ prompt-based tool for creating a synthetic voice. It is available in the app and through an API, and it can help when the Voice Library does not offer a suitable voice for a character, narrator, or project. The product page documents controls for qualities such as tone, regional speech characteristics, age, pacing, delivery, emotion, and audio quality.

ElevenLabs positions the tool for quick exploration and iteration. Results can vary depending on the prompt and use case, so a generated voice should be treated as a candidate to audition rather than an assured final result. When the priority is the platform’s most consistent, production-ready quality, ElevenLabs recommends choosing a suitable Professional Voice Clone when one is available in its library.

Close-up of a black studio microphone in a shock mount beside a black audio interface, set against geometric acoustic panels.

How to Describe and Refine the Voice Prompt

Start the prompt with the intended native language and regional variant. ElevenLabs recommends following that information with gender, age range, quality level, persona, emotion, and a concise description of timbre, pacing, and delivery. This creates a structured brief that separates the voice’s identity from how it should perform the preview text.

Granular descriptions generally produce more accurate and nuanced results, while a short prompt may be sufficient for a neutral or broadly usable voice. Add detail deliberately: a quality descriptor may improve clarity but reduce adherence to an unusually specific or niche voice concept.

Terminology also matters. ElevenLabs advises stating the language and regional variant in the first sentence to reduce language or dialect drift. Use “accent” only when you mean a regional or dialect characteristic; using it to request intonation can produce an unintended dialect change. If the first set of candidates misses the target, adjust one or two attributes—such as emotion, timbre, pacing, or delivery—before generating another set. This makes it easier to understand how the revised description changes the candidates.

Person wearing black headphones and adjusting a control knob at a wooden desk with a speaker, mug, paper, pen, lamp, shelves, and large windows.

How to Generate, Audition, and Save a Voice

In the ElevenLabs app, go to Voices → My Voices → Add a new voice → Voice Design. Enter the text you want the candidates to speak and provide the structured voice description. The supplied text can be previewed before generation.

Each generation returns three voice options. Listen to all three instead of assuming the first is the closest match. If none fits, revise the description and generate another set. Once a candidate is suitable, select it and save it to the voice library.

Preview generation charges credits once for the supplied text even though it produces three samples. The interface displays the character deduction in the Text to preview box. Saving the selected generation occupies one voice slot, so both credits and available slots matter when iterating.

ElevenLabs also documents separate API operations for generating previews from a prompt and saving a selected preview. The official pricing page observed on October 23, 2026 listed Voice Design in the Free plan, which included 10,000 monthly credits. Prices and access can change, and the listed prices exclude taxes, levies, and duties.

Conclusion

ElevenLabs Voice Design is a practical starting point when an existing library voice does not match a project and rapid exploration is more important than the most consistent production-ready result. Evaluate every candidate, refine the description when necessary, and keep preview credits and voice slots in view.

After a voice is saved and used for Text to Speech, ElevenLabs says voice selection has the greatest influence on the output, followed by model selection and then model settings. Draft a prompt beginning with the intended language and regional variant, generate and audition all three previews, and save only a candidate that suits the project.

Disclosures and limitations

  • This article was prepared with AI assistance using cited ElevenLabs documentation, product, and pricing materials.
  • Access, plan allowances, and pricing details were observed on October 23, 2026 and may change. Readers should confirm current terms before making a purchase or plan decision; any future affiliate relationship should be disclosed separately.

Sources

Improve this guide

Report a problem or suggest the next question

Submissions enter a private editorial queue. They are reviewed and deduplicated before any public update or new page is created.