Audio Generation Playground part of the Demo Playground, and the UltimaX Intelligence project.

This Space runs the Pocket TTS model from Kyutai.

A lightweight text-to-speech (TTS) application designed to run efficiently on CPUs. Forget about the hassle of using GPUs and web APIs serving TTS models.

Additionally, this Space uses a custom Docker image to maximize model performance and is optimized for the constraints of Hugging Face Spaces.

Like this project? You can support me by buying a coffee.

Generated Audio Output

Connecting to server... Establishing WebSocket connection

Waiting in queue... Position: calculating... 00:00

Generating audio... 00:00

No audio generated yet. Enter text below and click Generate, or try an example prompt.

Voice Selection

Select Voice

Voice Source

Model Parameters

Temperature 0.7

EOS Threshold -4.0

LSD Decode Steps 1

Frames After EOS Auto

Text Input

0 / 1000

Example Prompts