> For the complete documentation index, see [llms.txt](https://langtech-bsc.gitbook.io/applied-intelligence/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://langtech-bsc.gitbook.io/applied-intelligence/models/speech-models/matxa-tts-multiaccent.md).

# matxa-tts-v2-ca-multiaccent-graphemes

Catalan text-to-speech with 16 voices across five regional accents, for research and non-commercial use.

<a href="https://huggingface.co/BSC-LT/matxa-tts-v2-ca-multiaccent-graphemes" class="button primary" data-icon="arrow-up-right-from-square">View on Hugging Face</a> <a href="https://colab.research.google.com/drive/12Wki0uuWWipAcQA-lxtnvAGKey9SVrXi" class="button secondary" data-icon="play">Try it in Colab</a>

**matxa-tts-v2-ca-multiaccent-graphemes** speaks Catalan in **five regional accents with 16 different voices**. Like the [Central](https://langtech-bsc.gitbook.io/applied-intelligence/models/speech-models/matxa-tts-central.md) model, it reads text directly (graphemes) with no phonemizer. It was trained from scratch on studio-quality recordings from the LaFresCat, ReFresCat and EnVeuAlta datasets.

## Key facts

|                  |                                                                                           |
| ---------------- | ----------------------------------------------------------------------------------------- |
| **Task**         | Text-to-speech (acoustic model: text → mel spectrogram)                                   |
| **Architecture** | Matcha-TTS v2 (OT-CFM flow matching)                                                      |
| **Voices**       | 16 (speaker IDs 0–15)                                                                     |
| **Accents**      | Balearic, Central, North-Western, Valencian, Northern                                     |
| **Files**        | PyTorch checkpoint (213 MB) and an end-to-end ONNX model with a built-in vocoder (273 MB) |
| **License**      | Non-commercial: RAIL-style open license for research and non-profit use                   |

## Voices

Choose a voice by its speaker ID. The names are labels, not the real speakers' names.

| Accent            | Voices (speaker ID)                         |
| ----------------- | ------------------------------------------- |
| **Balearic**      | Olga (0) · Quim (1) · BM (10)               |
| **Central**       | Elia (2) · Grau (3) · CF (11) · CM (12)     |
| **North-Western** | Emma (4) · Pere (5) · Estel (13)            |
| **Valencian**     | Gina (6) · Lluc (7) · Arnau (8) · Berta (9) |
| **Northern**      | Laura (14) · Jordi (15)                     |

## Evaluation

Same 350-sentence test set as the Central model. Here the baseline is the same architecture trained on phonemes:

| Model                                     | ScoreQ ↑ (LaFresCat / ReFresCat) | WER ↓           | CER ↓           |
| ----------------------------------------- | -------------------------------- | --------------- | --------------- |
| **matxa-tts-v2-ca-multiaccent-graphemes** | 3.29 / 3.30                      | **9.44 / 9.15** | **3.04 / 3.02** |
| Matxa-TTS v2 (phonemes)                   | **3.37 / 3.46**                  | 12.40 / 11.53   | 4.07 / 3.79     |

The grapheme model is clearly **more intelligible** (20–24% lower WER), with naturalness close to the phoneme version.

## Quickstart

Try it in the [Colab notebook](https://colab.research.google.com/drive/12Wki0uuWWipAcQA-lxtnvAGKey9SVrXi), or install BSC's Matcha-TTS fork locally:

```bash
git clone -b v2-graphemes --single-branch https://github.com/langtech-bsc/Matcha-TTS.git
cd Matcha-TTS
pip install -e .
```

{% hint style="warning" %}
**Normalize numbers first.** Input text containing digits must be normalized into words before synthesis.
{% endhint %}

## License

{% hint style="danger" %}
**Non-commercial use only.** The model is free for academic, research and other non-profit purposes. Commercial use requires a licence from the voice artists: contact <bsc-lt@bsc.es> and <lafrescaproduccions@gmail.com>. The license also forbids using the model to imitate a real person's voice without their informed consent.
{% endhint %}

For commercial projects in Central Catalan, use [matxa-tts-v2-ca-central-graphemes](https://langtech-bsc.gitbook.io/applied-intelligence/models/speech-models/matxa-tts-central.md) (GPL-3.0) with the WaveNext vocoder.

<details>

<summary>Funding and citation</summary>

Funded by the Ministerio para la Transformación Digital y de la Función Pública and by the EU (NextGenerationEU), within the framework of the project ILENIA (ref. 2022/TL22/00215337). Training ran on MareNostrum 5.

```bibtex
@misc{peirolilja2026matxamultiaccent,
  title={Grapheme-Based Catalan Multi-Accent Text-to-Speech},
  author={Alex Peiró-Lilja and Jose Giraldo},
  organization={Barcelona Supercomputing Center},
  url={https://huggingface.co/BSC-LT/matxa-tts-v2-ca-multiaccent-graphemes},
  year={2026}
}
```

</details>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://langtech-bsc.gitbook.io/applied-intelligence/models/speech-models/matxa-tts-multiaccent.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
