Supertonic TTS Studio

Supertonic TTS Studio is a professional text-to-speech application that supports almost real-time continuous streaming and full audio generation while using CPU. This project leverages the Supertonic repository for high-quality TTS synthesis using ONNX models and multiple voice styles.

Features

• Real-time continuous streaming of generated speech.
• Full audio file generation with download capability.
• Multiple voice styles with descriptions.
• Adjustable speaking speed.
• Character and word count tracking for input text.
• Professional Gradio-based user interface.

Installation

1. Clone this repository:

git clone https://github.com/aritrodium/Supertonic-TTS.git
cd Supertonic-TTS

2. Ensure you have the required dependencies installed:

pip install -r requirements.txt
pip install gradio numpy soundfile

3. Clone the Supertonic repository and pull LFS files:

git lfs install
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/Supertone/supertonic supertonic
cd supertonic
git lfs pull

Usage

Run the application using Python:

python app.py

The Gradio interface will launch at http://0.0.0.0:7860 and can be accessed via your browser or shared using the share=True option.

Input Controls

• Text Input: Enter the text to synthesize. Supports multiple paragraphs.
• Voice Style: Select a preferred voice style from the available options.
• Speaking Speed: Adjust the rate of speech generation.

Output

• Streaming Output: Continuous playback of generated speech chunks.
• Download Output: Complete audio file generation and download.

Project Structure

Supertonic-TTS-/
├── app.py                 # Main Gradio application
├── supertonic/            # Cloned Supertonic repository
│   ├── onnx/              # Pretrained ONNX models
│   ├── voice_styles/      # JSON files for voice styles
│   └── tts_model.py       # TTS model functions
└── README.md

Functions

• run_tts_stream(text, speed, style_name) – Generates speech in streaming mode, yielding audio chunks and status updates.
• run_tts_full(text, speed, style_name) – Generates a complete audio file for download.
• list_voice_styles() – Lists all available voice style JSON files.
• load_style_by_name(filename) – Loads a specific voice style.

Contributing

Contributions are welcome. Please fork the repository, make your changes, and submit a pull request with a clear description of the changes.

License

This project is licensed under the MIT License.