Matcha-TTS-PL v2

Matcha-TTS-PL v2

Speaks Polish text aloud in your browser with two synthetic voices. The ONNX models run on your own device with WebGPU or WebAssembly, and nothing is sent to a server.

Made by machinekind on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This is a browser playground for Matcha-TTS-PL v2, a Polish text-to-speech system. It offers two synthetic voices, named male_voxcpm_2 and female_voxcpm_2, which were designed with VoxCPM2 and do not copy real people. All the speech work runs on your own device, so no text is sent to a server.

Who it suits: It suits people who want to hear Polish speech made in the browser, and developers who want to study or reuse an in-browser speech setup.

Industry | Who uses it | How they use it | Example
Media | Content creators | Turn Polish text into spoken audio without an account | A creator tries a line of narration with each voice
Software | Web developers | Study how ONNX models run in a page with WebGPU or WebAssembly | A developer reads the JavaScript files in the repository
Education | Language teachers | Let learners hear written Polish read aloud | A teacher plays a sentence in each voice
Research | Speech researchers | Compare the Studio and Edge setups on one text | A researcher switches setups and listens for differences

You can skip it if you need speech in a language other than Polish.

Where to use it: You use it in these places.

• A web browser, on the Space's page on Hugging Face
• A local copy of the page, opened in a browser on Windows, macOS or Linux

Use case | Who | What they get back
Hear Polish text | Anyone | Spoken audio from one of two synthetic voices
Compare setups | Researchers | Studio output next to Edge output for the same text
Study browser speech | Developers | A working example of ONNX Runtime Web with espeak-ng

How to set it up: Online use needs only a browser. A local copy needs Python for the download step.

• A web browser, with WebGPU if possible, since WebAssembly is the fallback
• Python with pip, only to fetch a local copy

1. Open the Space's page on Hugging Face to use it online, with nothing to install.
2. For a local copy, install the Hugging Face Hub package.
3. Download the Space's files into a folder and move into it.
4. Load index.html in your browser.

pip install -U huggingface_hub
hf download machinekind/matcha-tts-pl-v2 --repo-type space --local-dir matcha-tts-pl-v2
cd matcha-tts-pl-v2

How to use it: Open the page, pick a voice and type Polish text. Picking a voice triggers a one-time download of its model, which your browser keeps. Then play the speech. An advanced switch swaps the random duration predictor for the fixed one.

Tool | What it does
Studio | Uses the chosen voice's ONNX graph, an acoustic model with its own HiFi-GAN vocoder
Edge | Uses an encoder, a decoder and a small vocoder, and streams audio in 0.5 second chunks
Advanced switch | Turns on the deterministic duration predictor in place of the random one

Example: You pick the female voice and type a short Polish sentence. The page downloads that voice's model the first time, then plays the sentence as speech, all inside your browser.

For newcomers and experts: Text-to-speech means a program reads written text aloud. Here the whole program runs in your browser tab. Start by opening the Space's page, picking one voice and typing a short sentence.

Under the hood, the models run through ONNX Runtime Web, with WebGPU when available and WebAssembly otherwise. Phonemization uses an espeak-ng build for WebAssembly, covering Polish and English words it detects, and the effects chain is plain JavaScript. Durations are random by default, from a flow-matching predictor that draws once per phoneme, so repeated runs can differ.

You type Polish text
espeak-ng turns it into phonemes
An ONNX model makes the speech
The effects chain shapes the audio
You hear the result

Good to know: Each voice needs a model download the first time you use it.

• License: the playground code is MIT, the model weights are CC BY-SA 4.0, and espeak-ng is GPL-3.0.
• Cost to run: none named.
• Data it reaches: your text stays on your device, and the only fetch is each voice's model file, which your browser keeps after the first download.
• Not for: full English speech, since English is handled only for words it detects inside Polish text.

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: static
• License: mit

Links

• Space: https://huggingface.co/spaces/machinekind/matcha-tts-pl-v2
• Files: https://huggingface.co/spaces/machinekind/matcha-tts-pl-v2/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.