What is MiniMax Audio?
MiniMax Audio is a text-to-speech and voice cloning service from MiniMax, formerly called Hailuo AI Audio. It runs on the MiniMax Speech models, supports more than 40 languages, and offers a web app for creators plus an API for developers. Speech 2.6 targets voice agents with end-to-end latency under 250 milliseconds.
Many text-to-speech tools sound flat, stumble on numbers and URLs, or need long, clean recordings to clone a voice. MiniMax Audio handles URLs, email addresses, phone numbers, dates and currency without preprocessing, and its Fluent LoRA approach can clone a voice from imperfect or accented samples. The same models power voice agents and hardware products, so they are built for both narration and real-time use.
Main Features
- Text to speech in more than 40 languages with automatic language detection.
- Voice cloning that works from short or accented source recordings.
- Voice Design to create a new voice from a text description.
- End-to-end latency under 250 milliseconds with Speech 2.6 for live voice agents.
- Automatic handling of URLs, emails, phone numbers, dates and money amounts.
- A web app at minimax.io/audio and an API for developers.
How to Get Started with MiniMax Audio
- Open minimax.io/audio on a computer and sign in.
- Paste your text, pick a language and a voice from the library.
- Adjust speed, emotion or other settings and generate the audio.
- Clone a voice from a sample or move to the API for production use.
MiniMax Audio Pricing
The web app offers a free starting tier with monthly credits and paid subscription tiers for higher volume. API use, including voice cloning, is billed separately on usage.
Who Should Use It?
- Video creators and YouTubers adding voiceovers
- Podcasters and audiobook producers
- Developers building real-time voice agents
- Localization teams producing speech in many languages
- Educators recording narrated lessons
MiniMax Audio Alternatives
- Fish Audio: large community voice library of over 200,000 voices
- Meta's Audiobox: Meta research model, not a production service
- Sonix AI: transcribes speech to text rather than generating it
Frequently Asked Questions
What happened to Hailuo AI Audio?
Hailuo Audio is now called MiniMax Audio. It is the same speech service, now at minimax.io/audio, running newer Speech models such as Speech 2.6 and Speech 2.8 alongside Voice Design.
How many languages does MiniMax Audio support?
MiniMax says Speech 2.6 supports more than 40 languages, with accent control and switching between languages in the same text. Automatic language detection is also available.
Does MiniMax Audio have an API?
Yes. Developers can use MiniMax speech and voice cloning through the MiniMax API platform. MiniMax says its speech model is used by voice agent frameworks such as LiveKit and Pipecat.