The Speech-02 series from MiniMax are text-to-speech models that let you create natural-sounding voices with emotional expression. The models have support for more than 30 languages.

According to the Artificial Analysis Speech Arena, Speech-02-HD is the best text-to-speech model available today, while Speech-02-Turbo comes in third.

With Replicate, you can run these models with one line of code.

Listen to MiniMax Speech-02

Here’s a sample of the Speech-02-HD model reading an adapted version of this blog post, and the prediction that generated it.

Listen to this blog post

Try MiniMax Speech-02

You can choose between two models: Speech-02-HD for high-quality voiceovers and audiobooks, and Speech-02-Turbo, a cheaper model that’s faster and best suited for real-time applications.

Both models can be used with a cloned voice. Voice cloning needs at least 10 seconds of audio and takes about 30 seconds to train. Each voice can be adjusted for pitch, speed, and volume to make it sound natural.

Try the models in our playground:

• Speech-02-HD - For high-quality voiceovers and audiobooks
• Speech-02-Turbo - For real-time applications
• Voice Cloning - For creating custom voices

What you can build

These models can help you create:

• Virtual assistants that sound natural
• Audiobooks and voiceovers with studio-quality sound
• Language learning tools with native pronunciation
• Customer service bots that speak multiple languages
• Content that’s accessible to people who prefer audio

Emotion control

MiniMax’s emotion control system has two ways to add feeling to voices. The auto-detect mode figures out the emotional tone from your text, while manual controls let you set the exact emotion you want. This helps your voices sound natural and engaging, whether you’re making content for entertainment, education, or business.

Language support

The models work with more than 30 languages and accents. You can use different English variants (US, UK, Australian, and Indian), Asian languages (Mandarin, Cantonese, Japanese, Korean, Vietnamese, and Indonesian), and European languages (French, German, Spanish, Portuguese, Turkish, Russian, and Ukrainian).

Voice cloning and text-to-speech with JavaScript

You can run the models with our JavaScript client. First, install the Node.js client library:

Set your API token as an environment variable:

(You can get an API token from your account. Keep it private.)

Import and set up the client:

First, clone a voice. You’ll need an audio file in MP3, M4A, or WAV format. The file should be between 10 seconds and 5 minutes long and less than 20MB in size:

Now use the cloned voice for text-to-speech. You can add pauses between words using <#x#> where x is the pause duration in seconds (0.01-99.99):

Voice cloning and text-to-speech with Python

You can run the models with our Python client. First, install the client and set your API token:

Here’s how to clone a voice and use it for text-to-speech:

Pricing

The text-to-speech models charge based on input and output tokens. The turbo model costs $30 per million characters, while the HD model costs $50 per million characters. One token is a single character.

Voice cloning costs $3 per voice.

Keep up to speed

Connect with our community by following us on X and joining our Discord for updates and discussions.

Happy hacking! 🎙️