> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kymaapi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# MiniMax Voice Design

> Generate a synthesized voice from a natural-language description. No reference audio needed.

## Overview

`minimax-voice-design` is a text-to-voice synthesis service. Describe a voice in plain English ("warm female narrator with British accent, mid-30s") and MiniMax synthesizes a complete voice profile. The returned `voice_id` works the same way [voice clone](/models/minimax-voice-clone)'s output does — usable across all MiniMax HD/Turbo TTS calls.

Right for branding when no voice talent is available, fictional characters, persona voices for chatbots and agents, and any case where a written brief is all you have.

## Specs

| Field           | Value                                                              |
| --------------- | ------------------------------------------------------------------ |
| Model ID        | `minimax-voice-design`                                             |
| Creator         | MiniMax                                                            |
| Best for        | Persona voices from text, fictional characters, no-talent branding |
| Max description | 1000 characters                                                    |
| Optional hints  | `gender`, `age_group`, `preview_text`                              |
| Output          | `voice_id` (kyma-namespaced)                                       |
| Pricing mode    | Per call (flat)                                                    |

## Pricing

|                    |          Cost          |
| ------------------ | :--------------------: |
| Per designed voice | \$4.20 flat (one-time) |

Voice design is \~2× the cost of voice clone because synthesizing timbre from a text description is strictly more compute-intensive than reproducing a captured one.

## Use this when

* You don't have voice talent and want a polished voice anyway.
* You're prototyping characters and want to iterate on voice descriptions cheaply.
* You need a chatbot or agent with a distinctive but synthesized persona.

## Pick something else when

* You have reference audio → [`minimax-voice-clone`](/models/minimax-voice-clone) (\$2.10/voice).
* Stock voices are sufficient → use [`GET /v1/audio/voices`](/api-reference/audio-voices) directly.

## Example

```bash theme={null}
# Step 1: design the voice
curl -X POST https://kymaapi.com/v1/audio/voice-design \
  -H "Authorization: Bearer $KYMA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "description": "Warm female narrator with British accent, mid-30s, calm cadence",
    "gender": "female",
    "age_group": "young"
  }'
# → { "voice_id": "kyma_a91f4d2e7c8b5301", ... }

# Step 2: use the voice
curl -X POST https://kymaapi.com/v1/audio/speech \
  -H "Authorization: Bearer $KYMA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-speech-hd",
    "input": "And so the journey began.",
    "voice_id": "kyma_a91f4d2e7c8b5301"
  }' \
  --output intro.mp3
```

## See also

* [`POST /v1/audio/voice-design`](/api-reference/audio-voice-design) — endpoint reference
* [`POST /v1/audio/voice-clone`](/api-reference/audio-voice-clone) — clone from reference audio instead