Prompt to Voice - Describe a Voice, Hear It Free
Prompt to voice: describe age, gender, timbre and character, and AI designs a new voice. Try it free, no sign-up. OpenAI, Gemini and ElevenLabs compared.
Last updated: 2026-10-22
Prompt to voice turns a sentence into a brand-new voice: describe the age, gender, timbre, pace and character you want, and AI designs a voice that never existed before. Try it below without signing up: write a description, and Gemini 3.8 designs the voice and improvises a few lines in character. When you find one you like, save it to your voice library in the VoiceSmiths workbench and use it for any voiceover.
Design a voice with Gemini 3.8: free, no sign-up
2 free designs a day. Describe age, accent, timbre and attitude in a sentence or two. Gemini answers with a short line it improvises in character, in the language you write the description in.
Runs on Google's free tier: Google may use what you type here to improve its products, so don't enter anything private.
Turn your story into a full-cast audio drama with these voicesPaste a chapter or script: every character gets a voice and every line is acted. Finding the cast is free.Renders from $3.90: one-time, no subscription.Hear a voice made from one sentence
The prompt: "The classic Hollywood 'In a world…' trailer voice: ultra-deep, gravelly, epic." Gemini answered with an epic opening about the history of mankind:
How to write a voice prompt
Formula: age + gender + timbre + pace + manner or role. The more specific the description, the steadier the voice.
| Voice you want | Prompt you can copy |
|---|---|
| Late-night radio host | A man in his thirties, low and gentle, unhurried, like a late-night radio host telling a story |
| Bright young narrator | A woman in her early twenties, clear and bright, quick and playful, with a smile in her voice |
| Documentary narrator | A man in his forties, deep and steady, precise diction, measured pace, made for nature documentaries |
| Calm British butler | A refined British butler in crisp RP: calm, dry, impeccably courteous |
| Warm storyteller | A woman in her sixties with a warm, slightly husky voice, slow and cosy, like reading a bedtime story |
Four rules that save retries:
- Describe fixed traits, not a passing mood. Leave "whispering" or "furious" for audio tags when you voice a script.
- Write in the language you want to hear. Gemini's preview follows the language of the description.
- Describe the sound, not the threat. "Low, slow, cold, crisp diction" passes; "menacing, every word a blade" was blocked by Gemini's safety policy in our test.
- Name no one. Leave out celebrities, streamers and characters; describe the voice itself.
Prompt to voice: OpenAI vs Gemini vs ElevenLabs
| OpenAI prompt-to-voice | Gemini voice design | ElevenLabs voice design (VoiceSmiths workbench) | |
|---|---|---|---|
| Available since | Oct 16, 2026, all API customers | Sep 23, 2026, with Gemini 3.8 TTS | Live in the VoiceSmiths workbench |
| How you use it | API or OpenAI's voice Playground | Gemini API, Google AI Studio, or free on this page | Type a description on the web and click generate |
| Where the voice works | Only in GPT-Live sessions; not the Text to Speech API | Gemini API | VoiceSmiths text-to-speech, multi-speaker dialogue and the audio drama studio |
| Price | Creating is free; speech billed as GPT-Live sessions | 2 free a day here; no separate API price | 20 credits per voice |
| Limits | 100 voices per organization; adult voices only; no named people or fictional characters | 200 voices per project, kept one year; some prompts blocked | Description of at least 20 characters |
| Languages | Not listed by OpenAI | Follows your description's language | English, Chinese and many more |
To hear what a sentence can become, the free demo above is enough. To use the voice in a video, a podcast or an audiobook, design it in the VoiceSmiths workbench: the voice appears in your library, no code needed.
More tests and the API walkthrough are on Gemini voice design.
After you design a voice
- Open the voice workbench → Voices, name the voice and write the description (at least 20 characters), then click generate.
- The voice appears in your voice list; only you can see and use it.
- Pick it in text to speech to read any script, or cast it in the audio drama studio: every character's voice menu has a "My voices" group.
Want a real person's voice, such as your own? That is voice cloning, which starts from a recording.
Pricing
- Demo on this page: free, no sign-up, two voices a day from a shared site-wide allowance.
- Workbench voice design: 20 credits per voice. The $3.90 trial pack has 150 credits (one-time, no subscription, credits never expire), enough for seven voices; speech is billed separately at 7 credits per 100 characters. See pricing.
FAQ
What is prompt to voice? Prompt to voice (also called voice design) creates a brand-new synthetic voice from a text description, such as "a calm woman in her forties with a warm, low voice and a slight British accent". No recording is needed. OpenAI, Google Gemini and ElevenLabs all offer it.
Is there a free prompt-to-voice tool? Yes. This page lets you design two voices a day for free, no sign-up, using Google Gemini 3.8 voice design (a site-wide free allowance; if it runs out, come back tomorrow). To keep a voice and use it for voiceovers, design it in the VoiceSmiths workbench for 20 credits per voice.
When did OpenAI release prompt-to-voice? On October 16, 2026, for all API customers. You create the voice through the API or the voice Playground at no charge, but according to OpenAI's documentation the voice only works in GPT-Live sessions; the Text to Speech, Realtime and Chat Completions APIs do not support it.
Can I make a voice that sounds like a celebrity or a cartoon character? No. OpenAI does not allow imitating a named person, public figure or fictional character, and Gemini blocks some descriptions. Describe the voice itself (age, timbre, pace, manner) instead of naming anyone.
Which languages work? Gemini voice design follows the language of your description: write it in Mandarin and the preview is in Mandarin. ElevenLabs voice design, which the VoiceSmiths workbench uses, handles English, Chinese and many other languages. OpenAI has not listed the languages prompt-to-voice supports.
Start with one sentence
Write the voice you want in the box above and hear what comes back, free. When one fits, design it in the workbench and it is yours.