Alternative ways to use Kokoro-TTS kokoro-onnx, Kokoro-FastAPI, kokoro

• At least Python 3.10
• Git
• (Optional) Virtual Environment for dependency isolation

1. Clone the Repository

   https://github.com/NeuralFalconYT/Kokoro-TTS-Subtitle.git
   cd Kokoro-TTS-Subtitle

2. (Optional) Create and Activate a Virtual Environment

   • Windows:

     python -m venv myenv
     myenv\Scripts\activate

   • Mac/Linux:

     python3 -m venv myenv
     source myenv/bin/activate

3. Install PyTorch

• Forcefully install Torch with CUDA. If you did not, then requirements.txt will install the CPU version [It's happening with my system.]

  • Check CUDA Version (for GPU setup):

    nvcc --version

    Find your CUDA version example 11.8

  • Visit PyTorch Get Started and install the version compatible with your CUDA setup.:
    

    • For CUDA 11.8:

    pip install torch  --index-url https://download.pytorch.org/whl/cu118
    

    • For CUDA 12.1:

    pip install torch  --index-url https://download.pytorch.org/whl/cu121
    

    • For CUDA 12.4:

    pip install torch  --index-url https://download.pytorch.org/whl/cu124
    

4. Install Dependencies
   • Windows/Mac/Linux:

     pip install -r requirements.txt

4. Install eSpeak NG

• For Windows:

  1. Download the eSpeak NG 1.50 release from the eSpeak NG GitHub Releases. [Or, You can try eSpeak NG latest Version]
  2. Locate and download the file named espeak-ng-X64.msi.
  3. Run the installer and follow the installation steps. Ensure that you install eSpeak NG in the default directory:

     C:\Program Files\eSpeak NG
     

     Note: This default path is required for the application to locate eSpeak NG properly.

• For Linux:

  1. Open your terminal.
  2. Install eSpeak NG using the following command:

     sudo apt-get -qq -y install espeak-ng > /dev/null 2>&1

     Note: This command suppresses unnecessary output for a cleaner installation process.

5. Run the Application
   If you used a virtual environment, make sure you enabled it.
   • Windows/Mac/Linux:

     python beta.py

   Or,
   

   python one_clicker.py
   

   Then, double-click on run_app.bat (Windows) to execute the script.

6. Use Kokoko TTS Gradio app as an API in your other project:
   Check out scripts/api.py Code

1. Simply delete the kokoro_v1 folder from your project directory.

1. Inside the kokoro_v1 directory, run the following command to uninstall dependencies:

   pip uninstall -r requirements.txt

2. Uninstall PyTorch:

   pip uninstall torch

• Navigate to C:\Users\<username>\.cache\huggingface\hub
• Delete the contents of the hub folder. That's it! You've successfully removed everything.

If you encounter an error while generating Japanese audio, try the following steps:

1. Uninstall existing unidic:

   pip uninstall unidic -y

2. Reinstall unidic:

   pip install unidic

3. Download the dictionary data:

   python -m unidic download

Note: This solution is likely to work, the same issue has been observed on both Colab and Hugging Face.

Kokoro model is licensed under the [Apache License 2.0]

Kokoro HuggingFace