Kokoro TTS Local GUI - pySide6

Overview

This application is a user-friendly, Python-based Text-to-Speech (TTS) demo that integrates Kokoro TTS for high-quality, efficient speech synthesis. The GUI is built using PySide6 (Qt for Python) and allows editing text, adjusting parameters, and playing synthesized audio.

Key Features:

• Automatic phoneme conversion using Misaki G2P and fallback eSpeak-ng.
• Phoneme editing (planned) for fine control of TTS output.
• Multiple voices (downloaded as .pt voice models) that can be switched at runtime.
• Adjustable speed parameter.
• Audio playback, pause, resume, stop, and waveform preview.
• Ability to save generated audio in WAV or MP3.
• Automatic cleanup of temporary audio files older than a certain threshold.

Demo & Usage

Here’s a quick look at Kokoro TTS Local GUI in action:

The following animation shows how to adjust parameters and synthesize speech:

Requirements

Below is the recommended environment configuration to run this application:

• Python version: 3.8 to 3.11
  • We recommend Python 3.11 specifically if you plan on using CUDA. At the time of writing, Python 3.12 is not officially supported by PyTorch/CUDA. If you require GPU support, use 3.8 through 3.11.
• PyTorch with CUDA (optional): If you have a GPU and want acceleration.
• The required dependencies are listed in requirements.txt. Install them via:

pip install -r requirements.txt

For CUDA support, ensure you have the appropriate NVIDIA drivers and CUDA Toolkit installed, and that your PyTorch build supports CUDA.

Installation & Setup

1. Clone or download this repository.
2. Install Python (3.8 - 3.11 recommended).
3. (Recommended) Create a dedicated environment (virtualenv or conda) to avoid dependency conflicts:

   # Using venv
   python -m venv kokoro_gui
   source kokoro_gui/bin/activate  # Linux/Mac
   kokoro_gui\Scripts\activate    # Windows
   # or using conda
   conda create -n kokoro_gui python=3.11
   conda activate kokoro_gui

4. Install dependencies in that environment:

   pip install -r requirements.txt

5. (Optional) If you have a GPU:

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

   (Adjust cuda version to your system, e.g. cu117, cu118, etc.)
6. Download/Check voice models:
   • The application automatically downloads missing voice models on first run.
   • By default, they are stored in a local voices folder.

Usage

To launch the GUI:

python main.py

Command Line Arguments

• --config <path>: Specify a config YAML file for TTS settings.

File Structure

• main.py: Entry point. Parses arguments, then starts the GUI.
• ui_main.py: Implements the PySide6 user interface.
• tts_wrapper.py: Manages Kokoro TTS pipeline, including loading voices and generating audio.
• models.py: Handles voice downloading, caching, and loading for Kokoro TTS.
• persistence.py: Loads/saves metadata about generated audio, preserving session data.
• temp_cleanup.py: Automatically removes stale temporary files.
• requirements.txt: Required dependencies.
• error_handler.py: Simple GUI error dialog.

Workflow

1. Enter Text in the GUI.
2. Select Voice and Speed.
3. Click Synthesize.
4. Playback generated audio chunks or combined output.
5. Save any chunk or the combined file in WAV or MP3.
6. Clear old results or temporary files as needed.

Known Issues & Troubleshooting

• Voice Download Failures: If a voice fails to download, ensure you have a stable internet connection and that huggingface-hub is installed.
• CUDA not detected: Double-check that you installed the correct PyTorch wheel with CUDA support and that your graphics drivers are up to date.
• Python 3.12: At this time, some dependencies (like CUDA or certain libraries) may not fully support Python 3.12. Using Python 3.11 or lower is recommended.
• eSpeak-ng Configuration on Windows: If you need eSpeak-ng for non-English phonemization, you may need to set environment variables manually so that phonemizer can locate the libespeak-ng.dll:

  [System.Environment]::SetEnvironmentVariable("PHONEMIZER_ESPEAK_PATH", "C:\\Path\\to\\espeak-ng", [System.EnvironmentVariableTarget]::Machine)
  [System.Environment]::SetEnvironmentVariable("PHONEMIZER_ESPEAK_LIBRARY", "C:\\Path\\to\\espeak-ng\\libespeak-ng.dll", [System.EnvironmentVariableTarget]::Machine)

  Make sure to open a new terminal or log out/in so the changes can take effect.
• Voice Listing Bug: On some systems, the newly downloaded voices may not appear until you close and relaunch the application.

Fun Facts

You Don't need CUDA - Kokoro Is super fast!

If you generate text only on English - Ignore eSpeak-ng warrning.

License

This project is distributed under the MIT License (or insert your own license details here).

Contributing

Feel free to open issues, submit PRs, or suggest improvements!