---
language:
- ja
license: apache-2.0
model_type: transformer
---

# Kotoba-Speech-v0.1

Kotoba-Speech v0.1 is a 1.2B Transformer-based speech generative model. It supports the following properties:
1. Fluent text-to-speech generation in Japanese
2. One-shot voice cloning through speech prompt

![logo](./logo.webp)


## Usage
Plesae check out our HF Spaces [demo](https://huggingface.co/spaces/kotoba-tech/Kotoba-Speech?logs=build). 


## Model Details

* **Model type**: Our model is end-to-end transformers.
* **Language(s)**: Japanese 
* **Library**: We'll releasde our training code soon. Inference and model code are largely adopted from [metavoice](https://github.com/metavoiceio/metavoice-src). 


## Acknowledgements
- We thank meta-voice for opensourcing their code.

## License
Apache License Version 2.0, January 2004