kyutai-labs/pocket-tts

A TTS that fits in your CPU (and pocket)

What it solves

Pocket TTS is a lightweight text-to-speech (TTS) application that removes the need for expensive GPUs or external web APIs. It allows users to generate high-quality audio from text directly on their local CPU, providing a fast, private, and efficient way to produce speech.

How it works

The project uses a small model (100M parameters) optimized for CPU execution. It supports audio streaming and voice cloning, where a short audio sample (wav file) can be used to create a voice embedding (voice state) that the model uses to generate speech in that specific voice. It supports multiple languages including English, French, German, Portuguese, Italian, and Spanish.

Who it’s for

It is designed for developers who want to integrate TTS capabilities into their applications without requiring GPU hardware, as well as users who want a simple CLI or web interface for local speech generation.

Highlights

  • CPU-Optimized: Runs efficiently on CPUs, achieving up to 6x real-time speed on a MacBook Air M4 using only 2 CPU cores.
  • Low Latency: Provides first audio chunks in approximately 200ms.
  • Voice Cloning: Ability to clone voices from a plain wav file or pre-computed safetensors embeddings.
  • Multi-language Support: Supports six major European languages.
  • Flexible Deployment: Offers a Python API, a CLI, and a local HTTP server, with community ports to WebAssembly, Rust, C++, and C#.

Related

  • Project
    samuel-vitorino/sopro

    A lightweight 120M-parameter text-to-speech model family that enables high-quality voice cloning and streaming audio generation on laptop CPUs or in the browser.

  • Project
    KittenML/KittenTTS

    A lightweight, ONNX-based text-to-speech library that provides high-quality voice synthesis on CPU with models as small as 25MB.

  • Project
    tronghieuit/v-tts

    A lightweight Vietnamese text-to-speech system featuring multi-speaker synthesis and zero-shot voice cloning that runs efficiently on CPU.

  • Project
    KoljaB/RealtimeTTS

    A Python library for low-latency text-to-speech that converts strings and LLM token streams into audio using a wide variety of local and cloud engines.

  • Project
    OpenMOSS/MOSS-TTS-Nano

    A multilingual tiny speech generation model with 0.1B parameters designed for real-time, CPU-based inference and high-fidelity audio output.