MiniMax API Text-to-Speech (TTS): Voices & Samples

3 min read • December 30, 2024 (June 15, 2026)

Introduction
Cloned Voices Samples
Standard Voices Samples
Current API: Mureka speech
Conclusion

These samples come from the original MiniMax TTS model and are kept as a record of it. MiniMax speech is current again and now lives at POST speech/create, with ten models up to Speech 2.8 — see How to Generate Text-to-Speech Audio with the MiniMax API for what it can do today.

Introduction

The MiniMax API v1 was a third-party API for the MiniMax speech AI model, deployed at www.minimax.io/audio.

The samples below were produced in December 2024. MiniMax speech has moved on since, and today the API provides:

Concurrent generations are limited per account and the ceiling depends on the plan — a free account allows three. Beyond it a request returns 429.
You can connect as many accounts as you need.
Average response time for live streaming is about four seconds to the first audio.
Creating an MP3 from text takes about 5 seconds for 500 characters and about 22 at the 3,000 ceiling.
600+ pre-built voices available.
Ability to clone voices.
Supported Languages: 40 as of August 2026, from English, Chinese (Mandarin and Cantonese), Japanese and Korean through to Thai, Hindi, Tamil and Afrikaans. MiniMax keeps adding to the list.
Supported Emotions: happy, sad, angry, fearful, disgusted, surprised, neutral, fluent.
Supported Accents: EN-US (General), EN-Australian, EN-British, EN-Indian, CN-Northern, CN-Southern.
Supported Ages: Child, Young, Middle-aged, Elderly.
Supported Genders: Male, Female.