---
license: mit
language:
- en
base_model:
- coqui/XTTS-v1
---

# Fine-Tuned Xtts Model

This project fine-tunes a TTS (Text-to-Speech) model using an mp3 file extracted from a YouTube video. The training was conducted on a Hugging Face Space running locally via Docker. A GPU is recommended for faster training.

### Training Data
- **Source Video**: [YouTube Video](https://www.youtube.com/watch?v=u6J20_Aem3Y)
- **Training Audio**: The mp3 file used for training is included in the `files` directory.

### Hugging Face Space
The fine-tuning process is based on the Hugging Face Space found here:  
[FineTune Xtts Space](https://huggingface.co/spaces/drewThomasson/FineTune_Xtts)

### Docker Setup

#### With GPU
To run the training with GPU support:
```bash
docker run -it -p 7860:7860 --gpus all --pull always --platform=linux/amd64 registry.hf.space/drewthomasson-finetune-xtts:latest python app.py
```

#### Without GPU
To run without GPU support:
```bash
docker run -it -p 7860:7860 --pull always --platform=linux/amd64 registry.hf.space/drewthomasson-finetune-xtts:latest python app.py
```

### Notes
- Ensure you have a GPU available for optimal performance during training.
- The Docker image pulls the latest version each time it's run.