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

# Full Parameter Finetuning Malaysian Emilia F5-TTS Speech Enhancement

Continue training from [SWivid/F5-TTS](https://huggingface.co/SWivid/F5-TTS) on post Speech-Enhancement [Malaysian-Emilia-annotated](https://huggingface.co/datasets/mesolitica/Malaysian-Emilia-annotated)

**This model should be able to zero-shot voice conversion any Malaysian and Singaporean speakers**.

WanDB at https://wandb.ai/huseinzol05/CFM-TTS

## Checkpoints

We uploaded full checkpoints with optimizer states for each 50k steps at [full-checkpoint](full-checkpoint).

If you found out the latest checkpoint is overfitting, we have also uploaded alternative checkpoints,

1. 150000 steps, [huseinzol05/Malaysian-F5-TTS-150000](https://huggingface.co/huseinzol05/Malaysian-F5-TTS-150000)
2. 200000 steps, [huseinzol05/Malaysian-F5-TTS-200000](https://huggingface.co/huseinzol05/Malaysian-F5-TTS-200000)
3. 250000 steps, [huseinzol05/Malaysian-F5-TTS-250000](https://huggingface.co/huseinzol05/Malaysian-F5-TTS-250000)

## Dataset

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

## Source code

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