---
title: Voice Design for custom AI voices | Telnyx
canonical: https://telnyx.com/products/voice-design-lab
---
# Voice Design for custom AI voices | Telnyx

[View as page](https://telnyx.com/products/voice-design-lab)

**TELNYX VOICE AI**

# Design a custom voice for your brand in minutes

Design a fresh voice from a prompt, clone your own voice from audio, or dub an existing voice into another language. Then ship it across AI Assistants, Call Control, and TTS.

[Design a voice](https://telnyx.com/sign-up)[Explore Text-to-Speech](https://telnyx.com/products/text-to-speech-api)

WHY VOICE DESIGN

## Your voice, your brand, every surface

Stock voices sound like everyone else. Voice Design lets you create one that sounds like you, then use it everywhere Telnyx handles speech.

* ### Designed for your brand  
Describe the voice that fits your product and generate it. Iterate on tone, accent, and energy until it sounds right. No studio, no dataset, no waiting.
* ### One voice across every channel  
Use your custom voice in AI Assistants, Call Control, IVRs, and the TTS WebSocket. Same voice ID, same quality, no re-recording.
* ### From design to live in one platform  
Create, preview, save, and deploy a voice without wiring together separate providers. The same API that generates the voice carries it into production.

FEATURES

## Built for control, not just generation

Voice Design gives you the tools to create, refine, and ship a voice that is exclusively yours.

* Minutes to a custom voice  
From a text description to a production-ready voice in under a minute. Brand-fit voice creation at the speed of a prompt.
* Iterate until it sounds right  
Generate as many samples as you need. Refine the description, try again. You commit only when the voice lands.
* Quality that fits the use case  
Default for speed, Ultra for flagship content, Minimax for longer source recordings. Match the model to the job.
* A voice exclusive to your product  
A designed or cloned voice serves only your product. Your IVR, your agents, your TTS, all sounding like you.
* Voice infrastructure included  
Create, store, and deploy from the same platform that carries your calls and runs your agents. Your voice lives where your speech happens.
* Built for live voice  
Voice Design outputs work in real-time AI Assistants, Call Control, and the TTS WebSocket. The voice is ready for live conversation, batch synthesis, and everything between.

HOW IT WORKS

## From idea to production voice

Design a voice from a text prompt or clone one from audio, then use it across AI Assistants, Call Control, and the TTS WebSocket.

[EXPLORE DEV DOCS](https://developers.telnyx.com/docs/voice/voice-design-lab)

cURL

```
# Step 1: Design a voice from a text prompt
curl -X POST "https://api.telnyx.com/v2/voice_designs" \
  -H "Authorization: Bearer $TELNYX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Friendly Receptionist",
    "prompt": "Female, mid-thirties. Warm and full, slightly husky.",
    "text": "Hello, thank you for calling. How can I help you today?",
    "language": "en",
    "provider": "telnyx"
  }'

# Step 2: Save the design as a usable voice clone
curl -X POST "https://api.telnyx.com/v2/voice_clones" \
  -H "Authorization: Bearer $TELNYX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Friendly Receptionist",
    "voice_design_id": "DESIGN_ID",
    "version": 1,
    "language": "en",
    "gender": "female"
  }'
```

PRODUCTS

## Your entire voice stack on one platform

One account, one API, one bill.

[See all products](https://telnyx.com/products)

* ### Inference  
Run frontier open-source models on GPUs we own.  
[Learn more](https://telnyx.com/products/inference)
* ### Voice AI Agents  
Build and deploy low-latency voice AI agents.  
[Learn more](https://telnyx.com/products/voice-ai)
* ### Text-to-Speech  
Use your custom voices in production through the TTS API and WebSocket.  
[Learn more](https://telnyx.com/products/text-to-speech-api)
* ### Call Control  
Programmable voice for calls, IVRs, and SIP trunks.  
[Learn more](https://telnyx.com/products/voice-api)

## Build a voice worth listening to.

Create a Telnyx account, design or clone your first voice, and ship it into production today.

[Design a voice](https://telnyx.com/sign-up)[TALK TO US](https://telnyx.com/contact-us)

## Build a voice worth listening to.

Create a Telnyx account, design or clone your first voice, and ship it into production today.

[Design a voice](https://telnyx.com/sign-up)[TALK TO US](https://telnyx.com/contact-us)

## FAQ

### What is Voice Design?

Voice Design is the Telnyx tool for creating custom voices. Design a new voice from a prompt, clone an existing voice from audio, or dub a voice into another language. Then use it across AI Assistants, Call Control, and the TTS WebSocket.

### Can I create a voice from a prompt?

Yes. Describe the voice that fits your brand and generate it. You can iterate on the prompt and regenerate samples until the voice sounds right, then save it for production use.

### Can I clone a voice from audio?

Yes. Upload a short recording of the voice you want to replicate. The clone captures the speaker identity and is ready to use in seconds, depending on the provider you choose.

### What's the difference between a voice design and a voice clone?

A voice design is a draft. You generate samples from a prompt, listen to them, and iterate. When you save a design as a voice clone, it becomes a production-ready voice you can use across AI Assistants, Call Control, and the TTS WebSocket.

### Which providers can I choose from?

Telnyx and Minimax. Telnyx (Qwen3TTS) follows prompts closely and supports generation parameters like temperature and top\_p for fine control. Minimax interprets the same prompt differently, useful when you want a second take on a voice description. An Ultra model is also available for higher-quality cloning from audio.

### What audio formats and lengths work for cloning?

WAV, MP3, FLAC, OGG, and M4A. For Telnyx Qwen3TTS, 5 to 10 seconds of clean speech is optimal. Longer audio is auto-trimmed to 10 seconds, and the file size limit is 5 MB. For Minimax, 1 to 2 minutes of varied speech gives more vocal range, with a 20 MB limit. The Ultra model supports up to 10 seconds.

### What languages are supported?

Auto, Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian. Auto detects the language from the input text or audio.

### Can I preview a voice before saving it?

Yes. After you create a voice design, you can listen to the generated sample audio before deciding to save it as a clone. You can regenerate samples with a refined prompt if the first pass is not right.

### How long does voice cloning take?

Qwen3TTS and Minimax clones are ready immediately. The API returns the clone in the same response. Ultra clones are processed asynchronously: the API returns a pending status, and you poll the clone until its status becomes active.

### Where can I use a custom voice?

Custom voices work across AI Assistants, Call Control, and the TTS WebSocket. In AI Assistants, select the voice in the assistant voice settings. In Call Control, pass the voice ID in the speak command. In the TTS WebSocket, pass it as the voice query parameter.

### Can I dub my voice into another language?

Yes. You can create a voice in one language and use it to generate speech in another while keeping the speaker identity. This is useful for localized content that needs to sound like the same person across markets.

[Start building](https://telnyx.com/sign-up)