---
license: cc-by-nc-4.0
datasets:
- mesolitica/Malaysian-Emilia
language:
- ms
- en
base_model:
- SWivid/F5-TTS
new_version: mesolitica/Malaysian-F5-TTS-v3
---

# Full Parameter Finetuning Malaysian Emilia F5-TTS v2

Continue training from [SWivid/F5-TTS](https://huggingface.co/SWivid/F5-TTS) `F5TTS_Base` checkpoint on [Malaysian-Emilia](https://huggingface.co/datasets/mesolitica/Malaysian-Emilia),

with total 8472 hours included 600 hours Mandarin sampled from [amphion/Emilia-Dataset](https://huggingface.co/datasets/amphion/Emilia-Dataset).

## Features

1. This model should be able to zero-shot voice conversion any Malaysian and Singaporean speakers.
2. This model able to generate minimal filler sounds such as `erm`, `huh`, for example below,

<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/5e73316106936008a9ee6523/qoFxdAgPkxTxsNxrbdjs_.wav"></audio>

`Isu sekarangnya, erm, kita harus jadi yang terbaik untuk rakyat Malaysia, dan kita, uh, kena makan nasi lemak yang sedap lagi lazat, hah, penat nak kena cakap.`

## Checkpoints

We uploaded full checkpoints with optimizer states at [checkpoints](checkpoints).

## Dataset

We train on postfilter [Malaysian-Emilia](https://huggingface.co/datasets/mesolitica/Malaysian-Emilia) called [Malaysian-Voice-Conversion](https://huggingface.co/datasets/mesolitica/Malaysian-Voice-Conversion)

## Source code

All source code at https://github.com/mesolitica/malaya-speech/tree/master/session/f5-tts