Pocket TTS

Name: Pocket TTS
Author: Pocket TTS

Free plan APILinuxMacSelf-hostedWebWindows

What Pocket TTS is

Pocket TTS is a lightweight text-to-speech application built to run efficiently on CPUs. Its model has 100 million parameters, supports streaming audio and takes about 200 milliseconds to produce the first audio chunk. The documentation reports around six times real-time speed on a MacBook Air M4 CPU using two cores. It can process very long text inputs, supports six languages and exports WAV audio.

Users can access Pocket TTS through a Python API, a command-line interface or a local HTTP server. A browser demo can generate speech without installation, and the maker says the model can run in a browser, although browser implementations are not officially supported. Installation is documented for Linux, macOS and Windows. The package supports Python 3.10 through 3.14 and requires PyTorch 2.5 or later; a GPU build of PyTorch is not required. On Linux, the default PyTorch installation may download several gigabytes of CUDA runtime wheels unless users select the CPU-only index. Pocket TTS is free, permits commercial use and supports voice cloning, subject to restrictions against impersonation without explicit lawful consent and other prohibited uses. The model files are listed under CC-BY-4.0, and access requires agreeing to share contact information.

Who it suits

Pocket TTS suits developers and other users who want CPU-based speech generation through a Python API, command line or local server. It can also suit commercial projects and voice-cloning workflows, provided they follow the stated use restrictions.

What it does well

Free to use, with commercial use permitted.
Runs on CPUs without requiring GPU PyTorch.
Offers a Python API, command-line interface and local HTTP server.
Streams audio with a first chunk in about 200 milliseconds.
Handles long inputs and supports six languages.
Provides a browser demo without installation.

What to know first

Browser implementations are not officially supported.
Linux’s default PyTorch installation may download several gigabytes of CUDA runtime wheels.
Model file access requires agreeing to share contact information.
Text input cannot add silence to create pauses.

Limits

Adding silence in text input to create pauses is unsupported. The browser implementation is not officially supported, and model file access requires agreeing to share contact information.

Plans

Plan	Billed and limits	Price
Pocket TTS	Open source · CPU-based text-to-speech · Python API and CLI	Free

As kyutai-labs.github.io lists them, read 12 Oct 2026. The maker's own page has the final word.

Getting started

Open the Pocket TTS website at https://kyutai-labs.github.io/pocket-tts/.
For speech generation without installation, open the linked browser demo.
For local use, install the package on Linux, macOS or Windows.
Use Python 3.10 through 3.14 and PyTorch 2.5 or later.
On Linux, select the CPU-only PyTorch index to avoid the default CUDA runtime wheel downloads.
Agree to share contact information to access the model files.

Verdict

Choose Pocket TTS if you need a free, CPU-oriented text-to-speech tool with API, command-line and local server access. It supports commercial use and voice cloning under its stated restrictions. Look elsewhere if you need browser use with official support or pause insertion through silence in the text.

Questions

Is Pocket TTS free?

Yes. Its listed plan is free, and there is no free trial.

Can I use Pocket TTS commercially?

Yes, commercial use is permitted. Voice impersonation or cloning requires explicit lawful consent, and other illegal, harmful, deceptive or privacy-invasive uses are prohibited.

Which platforms does Pocket TTS support?

Its listed platforms are API, Linux, macOS, self-hosted, web and Windows. Installation is documented for Linux, macOS and Windows; browser implementations are not officially supported.

What interfaces and audio format does it offer?

It provides a Python API, command-line interface and local HTTP server. WAV is the listed export format.

What does it require to run locally?

The package supports Python 3.10 through 3.14 and requires PyTorch 2.5 or newer. It does not require the GPU version of PyTorch.

What license applies to the model files?

The model card lists CC-BY-4.0. Access requires agreeing to share contact information.

Facts from Pocket TTS's own pages

API access: Yes · kyutai-labs.github.io, 20 Sept 2026
Browser availability: The maker offers a browser demo and says the model can run client-side in a browser, while noting that browser implementations are not officially supported. · kyutai-labs.github.io, 11 Oct 2026
Browser demo: The documentation links to a Kyutai website demo that can generate speech in the browser without installation. · kyutai-labs.github.io, 12 Oct 2026
Commercial use: Yes · kyutai-labs.github.io, 20 Sept 2026
CPU performance: The project reports about six times real-time generation on a MacBook Air M4 CPU using two CPU cores. · kyutai-labs.github.io, 12 Oct 2026
Export formats: WAV · kyutai-labs.github.io, 20 Sept 2026
Free plan: Yes · kyutai-labs.github.io, 20 Sept 2026
Headquarters: Paris, France · kyutai-labs.github.io, 28 Sept 2026
Integrations: The project lists community projects including a Wyoming protocol Docker container for Home Assistant Voice and a Unity 6 integration. · kyutai-labs.github.io, 12 Oct 2026
Interfaces: It offers a Python library, command-line interface, and local HTTP server with a web interface at localhost:8000. · kyutai-labs.github.io, 12 Oct 2026
Languages: 6 · kyutai-labs.github.io, 20 Sept 2026
Languages: The documented pretrained language models cover English, French, German, Portuguese, Italian, and Spanish. · kyutai-labs.github.io, 12 Oct 2026
License and access: The Hugging Face model card lists the model license as CC-BY-4.0 and says users must agree to share contact information to access the model files. · huggingface.co, 11 Oct 2026
Long inputs: Pocket TTS can handle infinitely long text inputs. · kyutai-labs.github.io, 12 Oct 2026
Long text: The model can handle infinitely long text inputs. · kyutai-labs.github.io, 11 Oct 2026
Model size: The model has 100 million parameters. · kyutai-labs.github.io, 12 Oct 2026
Platforms: windows,macos,linux,api,self_hosted · kyutai-labs.github.io, 20 Sept 2026
Platform setup: The maker documents installation on Linux, macOS, and Windows; on Linux, its default PyTorch installation may download several gigabytes of CUDA runtime wheels unless users choose the CPU-only index. · kyutai-labs.github.io, 11 Oct 2026
Purpose: Pocket TTS is a lightweight text-to-speech application designed to run efficiently on CPUs. · kyutai-labs.github.io, 11 Oct 2026
Requirements: The package supports Python 3.10 through 3.14 and requires PyTorch 2.5 or later. · github.com, 11 Oct 2026
Runtime requirements: The package supports Python 3.10 through 3.14 and requires PyTorch 2.5 or newer; it does not require the GPU version of PyTorch. · kyutai-labs.github.io, 12 Oct 2026
Security and use policy: The project prohibits voice impersonation or cloning without explicit and lawful consent, along with deceptive, harmful, or privacy-invasive uses. · kyutai-labs.github.io, 12 Oct 2026
Streaming and latency: It supports audio streaming and reports about 200 ms to produce the first audio chunk. · kyutai-labs.github.io, 12 Oct 2026
Support: The maker invites users to open issues or pull requests on GitHub and accepts contributions. · kyutai-labs.github.io, 11 Oct 2026

Where Pocket TTS stands on Freedom251

Text-to-Speech Software#53 of 223
Text-to-speech tools#82 of 232