Preface

Prerequisites

• Python 3.11.x
• CUDA-enabled device 11.8

Installation

Follow these steps for installation:

1. Ensure that CUDA is installed

2. Clone the repository: git clone https://github.com/Haurrus/xtts-trainer-no-ui-auto

3. Navigate into the directory: cd xtts-trainer-no-ui-auto

4. Create a virtual environment: python -m venv venv

5. Activate the virtual environment:

   • On Windows use : venv\scripts\activate
   • On linux use : source venv\bin\activate

6. Install PyTorch and torchaudio with pip command :

   pip install torch==2.1.1+cu118 torchaudio==2.1.1+cu118 --index-url https://download.pytorch.org/whl/cu118

7. Install all dependencies from requirements.txt :

   pip install -r requirements.txt

XTTS Fine-Tuning Project for xTTSv2 xtts_finetune_no_ui_auto.py

Overview

This is a Python script for fine-tuning a text-to-speech (TTS) model for xTTSv2. The script utilizes custom datasets and use CUDA for accelerated training.

Usage

To use the script, you need to specify two JSON files: args.json and datasets.json.

args.json

This file should contain the following key parameters:

• num_epochs: Number of epochs for training, if set to 0 it will auto calculate it.
• batch_size: Batch size for training.
• grad_acumm: Gradient accumulation steps.
• max_audio_length: max audio duration of wavs used to train.
• language: language used to train the model.
• version: by default main from xTTSv2
• json_file: by default main from xTTSv2
• custom_model: by default main from xTTSv2

datasets.json

This file should list the datasets to be used with paths and activation flags.

Finetune_models folder

To train models you need a dataset, there's an exemple dataset in the finetune_models, it's a FemaleDarkElf voice from Skyrim

Running the Script

Execute the script with the following command:

python xtts_finetune_no_ui_auto.py --args_json path/to/args.json --datasets_json path/to/datasets.json

Features

• Custom model training and fine-tuning.
• Support for multiple datasets.

Audio Dataset Preprocessing xtts_generate_dataset.py

Overview

This script processes audio files to create training and evaluation datasets using the Whisper model. It has been updated to include several new features and improvements.

Usage

To use the script, provide the path to a JSON configuration file and the Whisper model version as command-line arguments:

python xtts_generate_dataset.py --config path/to/config.json --whisper_version large-v3

The JSON configuration file should contain the audio paths, target language, activation flag, and name for each dataset.

JSON Configuration Format

The configuration file should follow this format:

[
    {
    "name": "dataset_name",
        "audio_path": "path/to/audio/files",
        "language": "en",
        "activate": true
    }
]

Replace path/to/audio/files with the actual path to your audio files and dataset_name with a preferred name for your output subdirectory.

Features

1. Processing Entire Audio Files: The script has been modified to process entire audio files without splitting them into segments. Each audio file is transcribed as a whole, and the corresponding transcription is stored.

2. Output Directory Customization: The output directory is now named output_datasets and is created in the root directory where the script is executed. Inside this directory, subdirectories are created based on the name provided in the JSON configuration file.

3. Language Configuration: The script writes the target language to a lang.txt file in the output directory, ensuring consistent language settings across the dataset.

4. Audio File Copying: All processed audio files are copied into a wavs folder located in their respective output subdirectories.

5. Error Handling and Logging: The script includes error handling and logging mechanisms to provide clear feedback in case of any issues during the processing.

6. Configurable Through JSON: The entire preprocessing can be configured using a JSON file, making it easy to adjust settings like the target language, audio paths, and output names.

Contributing

Contributions are welcome. Please fork the repository and submit pull requests with your changes.

Credit

Thanks to the author daswer123 for the repository xtts-webui , My project is based on his work.