What is ElevenLabs Voice Design V3?
Voice Design V3 is ElevenLabs' current text-to-voice model, built on its Eleven v3 speech model. You describe a voice, such as its language, age, accent, tone and pacing, add a line of preview text, and it returns three voices to compare. Saving one puts it in your voice slots, ready for Eleven v3 with its audio tags or any older ElevenLabs model.
ElevenLabs still calls the feature experimental and says results vary with the prompt. For the most consistent quality it points people to Professional Voice Clones instead, which need a real recording and currently cannot use the v3 model.
What changed from Voice Design v2?
More languages and a new starting point: V3 covers 70+ languages and can work from reference audio.
- Languages
- 70+ on eleven_ttv_v3, against 29 on the older eleven_multilingual_ttv_v2.
- Reference audio
- On the API, V3 can start from a recording weighted against the prompt; v2 cannot.
- Audio tags
- Voices made with V3 work with Eleven v3's audio tags for more expressive speech.
- API default
- The design endpoint still defaults to v2, so set the model ID to eleven_ttv_v3.
Who is ElevenLabs Voice Design V3 best for?
ElevenLabs Voice Design V3 is best for creators and developers who need a specific voice, such as a game character or narrator, that they cannot find in ElevenLabs' voice library.
Characters for games, audiobooks and animation
Describe age, accent, tone and pacing in a prompt and pick from three voices that match.
Voices the library lacks
When none of the 10,000+ voices in ElevenLabs' library fits, it builds a new one instead of settling for a near match.
Apps that create voices on demand
An API generates previews and saves the chosen one, and the v3 model can also start from a reference recording.
Voices in many languages
The v3 voice design model covers 70+ languages, and ElevenLabs advises naming the language and regional accent at the start of the prompt.
What's included with ElevenLabs Voice Design V3?
ElevenLabs Voice Design V3 gives three previews per run, prompt and loudness controls, reference audio on the API, auto-written preview text and later remixing.
- Three previews per run
- Each generation returns three voices from one prompt, and you pay once for the preview text.
- Prompt controls
- A guidance scale sets how strictly it follows the prompt, and a loudness setting fixes the saved voice's volume.
- Reference audio
- On the API, the v3 model can start from a recording and weigh it against the prompt with a prompt strength setting.
- Auto-written preview text
- The app can write a preview script that suits the prompt, with audio tags included.
- Voice Remix
- Change a saved voice's gender, accent, pacing or audio quality later with another prompt.
How does ElevenLabs Voice Design V3 work?
Describe the voice
Start with the language and regional accent, then add age, gender, tone, pacing and audio quality.
Add preview text
Write a few sentences that suit the character, or let the app write them for you.
Compare three voices
Generate, listen to the three previews, and adjust the prompt or guidance scale if none fits.
Save and use it
Save the best one to a voice slot and use it for speech in the app or through the API.
How much does ElevenLabs Voice Design V3 cost?
ElevenLabs Voice Design V3 comes with every ElevenLabs plan, including the free one with 10,000 credits a month, and each design costs credits only for its preview text; paid plans start at $6 a month.
- 10,000 credits a month
- 3 custom voice slots
- No commercial licence
- 30,000 credits a month
- 10 custom voice slots
- Commercial licence and instant voice cloning
- 121,000 credits a month
- 30 custom voice slots
- Professional voice cloning
- 600,000 credits a month
- 160 custom voice slots
- 44.1 kHz PCM audio through the API
Scale costs $299 a month (1.8M credits, 660 voice slots, 3 seats) and Business $990 a month (6M credits, 2,200 voice slots, 10 seats); Enterprise is priced on request. Paying yearly costs ten months' worth: Starter works out to $5 a month, Creator $18.33 and Pro $82.50. Each Voice Design run returns three voices and charges only for the characters in the preview text (100 to 1,000 characters), once for all three; saving a voice uses one of your voice slots. Credits are shared with every other ElevenLabs tool, and paid plans carry unused credits over for up to two months. Full pricing on elevenlabs.io. We checked these prices on .
How do readers rate ElevenLabs Voice Design V3?
No one has rated ElevenLabs Voice Design V3 yet. Be the first.
Sign in to rate. One rating per account.
ElevenLabs Voice Design V3 pros and cons
It is cheap to try and covers 70+ languages, but results vary with the prompt and many users get better quality from cloning a real recording.
Pros
- Cheap to try: each run costs only the characters in the preview text, charged once for all three voices, and it works on the Free plan.
- The v3 model covers 70+ languages, while Hume's Octave voice design works in English only.
- Saved voices support Eleven v3's audio tags and also work on ElevenLabs' older and faster models, such as Flash v2.5.
- The same design and save steps are open through an API, which can also start from reference audio.
Cons
- Results vary: ElevenLabs calls the feature experimental, and several Reddit users say designed voices come out generic or inconsistent and that cloning a real recording works better.
- The API still defaults to the older v2 design model, so you have to set eleven_ttv_v3 yourself to get V3.
- Every saved voice takes a slot (3 on Free, 10 on Starter), the Free plan has no commercial licence, and designed voices cannot be shared in the Voice Library.
What are the best ElevenLabs Voice Design V3 alternatives?
Four tools on Recaply do ElevenLabs Voice Design V3's job. Qwen3-TTS is open source, so you can run it on your own server.
Hume AIFreeChoose it when you design English voices and want a cheaper start: its Starter plan costs $3 a month for 30,000 characters, against $6 for ElevenLabs Starter.
Realtime TTS-2FreemiumChoose it when the voice will speak in a live agent: Inworld includes voice design on its pay-as-you-go plan and bills speech at $25 per million characters with no monthly fee.
Sonic 3.6FreeChoose it when you have a recording you are allowed to use: Cartesia clones a voice from 10 seconds of audio on its $5 Pro plan and speaks 44 languages.
Qwen3-TTSOpen SourceChoose it when you want to design a voice from a text description with open models you host yourself.
Questions about ElevenLabs Voice Design V3
What is ElevenLabs Voice Design V3?
ElevenLabs Voice Design V3 turns a written description into a new synthetic voice, with three options per try, for creators and developers who cannot find the voice they need in ElevenLabs' library.
Who is ElevenLabs Voice Design V3 for?
ElevenLabs Voice Design V3 is best for creators and developers who need a specific voice, such as a game character or narrator, that they cannot find in ElevenLabs' voice library.
Is ElevenLabs Voice Design V3 free?
Yes. ElevenLabs Voice Design V3 has a free plan: 10,000 credits a month.
How much does ElevenLabs Voice Design V3 cost?
ElevenLabs Voice Design V3's paid plans are Starter at $6 per month, Creator at $22 per month and Pro at $99 per month.
Is there an open source alternative to ElevenLabs Voice Design V3?
Yes. Qwen3-TTS is an open source tool that does the same job, so you can run your own.
What are the best ElevenLabs Voice Design V3 alternatives?
The closest alternatives on Recaply are Hume AI, Realtime TTS-2, Sonic 3.6 and Qwen3-TTS.
Where can I try ElevenLabs Voice Design V3?
On its official website, elevenlabs.io.
How many credits does ElevenLabs Voice Design V3 use?
ElevenLabs Voice Design V3 charges only for the preview text each voice reads, 100 to 1,000 characters, and charges it once even though you get three voice options. Saving a voice costs no credits but takes a voice slot: 3 on Free, 10 on Starter.
Can I use ElevenLabs Voice Design V3 voices commercially?
Not on the Free plan, which has no commercial licence. Commercial use starts with the Starter plan at $6 a month, which also adds instant voice cloning and raises your custom voice slots from 3 to 10.
Should I design a voice or clone one in ElevenLabs?
Clone one when you own or have permission to use a recording: ElevenLabs says Professional Voice Clones give its most consistent quality. Design one when you have no recording or need a character that does not exist, and expect a few rounds of prompts.
Can I share a Voice Design V3 voice in the ElevenLabs Voice Library?
No. ElevenLabs only lets human-verified Professional Voice Clones into the Voice Library, so voices made with Voice Design, like Instant Voice Clones, cannot be shared there or earn payouts. You can still use them in your own projects.
Our verdict
Voice Design V3 is worth trying when no library voice fits, but treat each result as a draft.
It costs only the preview text per run, covers 70+ languages and returns three options at a time, yet ElevenLabs itself calls it experimental and users report generic or uneven voices.
- Skip it for
- Brand or narration work where you can record a real speaker, since a Professional Voice Clone gives more consistent quality.
- Check first
- Run a few rounds on the Free plan before paying, and count the voice slots your plan gives you.
MiMo-V2.5-TTS
Eleven v4
Qwen3-TTS CustomVoice 1.7B
Eleven v4 Turbo
Gemini 3.8 Flash TTS
MAI-Voice-2