MiniMax Voice Design

Overview

MiniMax Voice Design creates a synthesized voice profile from a text prompt. It returns a voice_id that works with any MiniMax HD or Turbo speech endpoint. The model operates on a per-call basis with a flat charge for each designed voice. On Kyma, the model runs through the standard OpenAI-compatible base URL using a single API key. Every request includes automatic failover if a serving path degrades, and responses report exact costs in usage.cost alongside the active model in the X-Kyma-Model header. Prompt caching applies to repeated prefixes at a 10% rate, and new accounts include a $0.50 signup credit. The model accepts text input and outputs audio metadata rather than continuous speech streams. It does not support reasoning, vision, or structured outputs, and operates with a 1,000-token context window. It is optimized for voice design rather than direct text-to-speech generation.

Specs

Field	Value
Model ID	`minimax-voice-design`
Best for	Branding when no voice talent available, fictional characters, persona voices from text alone
Context window	1,000
Input modalities	Text
Output modalities	Audio
Tool calling	Yes
Structured outputs	Yes
Prompt caching	Yes
Speed	Medium
Cost band	Premium
Release stage	Stable

Pricing

	Per 1M tokens
Input	$0.20
Output	$0.60

Use this when

Brand Voice Creation — Generate consistent synthetic voices for corporate branding without hiring voice actors.
Fictional Character Voices — Design unique vocal profiles for game or narrative characters using text prompts.
Persona Voice Generation — Create custom speaker identities from descriptive text when reference audio is unavailable.
Rapid Audio Prototyping — Test different vocal styles during product development using a flat per-call workflow.

Not ideal for

Do not use this model to clone existing voices from audio samples or to generate continuous long-form speech directly.

Pick something else when

You need to clone a real voice from audio samples: use eleven-turbo-v2-5 or minimax-voice-clone.
You need direct text-to-speech generation: use minimax-speech-hd or minimax-speech-turbo.
You need multilingual speech with existing voice IDs: use eleven-multilingual-v2 or eleven-v3.

Example

from openai import OpenAI

client = OpenAI(base_url="https://kymaapi.com/v1", api_key="ky-...")

response = client.chat.completions.create(
    model="minimax-voice-design",
    messages=[{"role": "user", "content": "..."}],
)

Agent query example

Ask the API which models fit, instead of hardcoding an id:

curl "https://kymaapi.com/v1/models?recommended_for=generate-speech&tools=true&quality_tier=frontier-open"

FAQ

How do I use the generated voice profile? The model returns a voice_id that you pass to the /v1/audio/speech endpoint with any MiniMax HD or Turbo SKU. Does Kyma cache requests for this model? Yes, prompt caching is supported and bills repeated prompt prefixes at this model’s cached input rate. What happens if the serving path degrades? Kyma handles automatic failover, so degraded paths are rerouted transparently without requiring client-side retries.

​Overview

​Specs

​Pricing

​Use this when

​Not ideal for

​Pick something else when

​Example

​Agent query example

​FAQ