Chatterbox Turbo is Live on Models

Siddharth Dhulipalla

23 Dec 2025

2 min

We just added Chatterbox Turbo to Models.

Chatterbox Turbo is one of the fastest open-source text-to-speech models available today—and the first to combine real-time performance with paralinguistic prompting. That means speech that doesn’t just sound realistic, but behaves like a human voice: laughs, sighs, hesitations, and all.

On its own, Chatterbox Turbo is already impressive.

On Models, it gets even better.

Expressive Speech, Not Just Fast Speech

Real conversations aren’t just words—they’re reactions.

Chatterbox Turbo lets you embed non-verbal cues directly into your prompts:

[sigh] I don’t know if that’s a good idea. [chuckle]

The model performs those moments naturally, in the same cloned voice, with support for tags like:

[laugh], [chuckle], [sigh]
[gasp], [sniff], [clear throat]
and more

Combined with zero-shot voice cloning from as little as 5 seconds of audio, this makes it possible to build voice agents that feel present, expressive, and human—without brittle post-processing or hacks.

Why It’s Better on Models

Chatterbox Turbo is optimized for speed at the model level—up to 6× faster than real-time on a GPU.

Models pushes that further.

We deploy Chatterbox Turbo across 14+ regions globally, routing each request to the closest available GPU over our private backbone network. The result is lower end-to-end latency than you can get running the model in a single region—or even most managed inference platforms.

You get:

Global, region-local inference out of the box
Consistently low latency for real-time voice apps
One unified API to move from prototype to production

This is the difference between a fast model and a fast experience.

Open, Production-Ready, and Built for Trust

Chatterbox Turbo is MIT-licensed and production-ready, making it a strong foundation for teams that want transparency and control.

Each generated audio file also includes an imperceptible watermark for traceability—an important step toward responsible, compliant voice AI as regulations evolve.

Try It Now

You can test Chatterbox Turbo directly in the Models UI today—then bring the same model into your own voice AI app with the exact same API.

👉 https://models.hathora.dev/model/resemble-ai-chatterbox-turbo

👉 Deploy it globally in your voice AI application

👉 Mix and match models with the same API on Models

We’re excited to see what you build when expressive, real-time speech meets globally distributed infrastructure.