Kokoro Voice Blender GUI

A PyQt5-based graphical user interface (GUI) for blending multiple voices using the Kokoro ONNX text-to-speech (TTS) model. This tool allows users to mix various voices with customizable weights, adjust playback speed, and save configurations for later use. It complements the Kokoro TTS GUI by sharing the same configuration directory (/home/pg/Dokumente/Kokoro-82M/configs/ by default), enabling seamless voice mix exchanges between the two applications.

Features

The Kokoro Voice Blender GUI offers a rich set of features for voice blending and audio synthesis:

1. Text-to-Speech Input

• Text Input: Enter any text to be synthesized into speech via a text area.

2. Voice Blending with Sliders

• Multiple Voices: Choose from a predefined list of 50+ voices (e.g., af_alloy, im_nicola, zf_xiaoyi).
• Customizable Weights: Adjust the contribution of each voice using sliders (range: 0.00 to 1.00).
• Normalization Option:
  • Enabled: Automatically adjusts slider values to sum to 1.00, ensuring balanced blending.
  • Disabled: Allows raw weights, with internal scaling for audio output to maintain full intensity.
• Responsive Sliders: Click anywhere on a slider to jump to a specific value.

3. Random Voice Mixing

• Randomize Button: Select 1 to 20 voices and assign random weights.
  • With normalization: Weights sum to 1.00 (using Dirichlet distribution).
  • Without normalization: Weights range from 0.01 to 1.00.
• Refresh Button: Re-randomizes weights for currently active voices and plays the new blend immediately.

4. Audio Playback and Saving

• Preview Blend: Synthesize and play the blended voice mix in real-time.
• Synthesize and Save: Save the synthesized audio as output_blended.wav.
• Auto-Loop Preview:
  • Automatically replays the blend after changes or continuously if enabled.
  • Controlled via "Auto-Loop Preview" and "Continuous Loop" checkboxes.

5. Configuration Management

• Save Config: Save voice weights, normalization settings, slider layout, and speed to a JSON file in the configs/ directory (default: /home/pg/Dokumente/Kokoro-82M/configs/).
• Load Config: Load previously saved configurations.
• Last Config: Automatically saves the current state on exit and loads it on startup.
• Shared Configs: Uses the same configs/ directory as Kokoro TTS GUI for interoperability.

6. Customization Options

• Sliders per Row: Adjust the GUI layout (1 to 5 sliders per row) for better usability.
• Speed Control: Modify playback speed (0.1x to 3.0x) using a spin box.
• Reset Sliders: Set all sliders to 0.00 to start fresh.

Screenshot

Installation

Prerequisites

• Python 3.8 or higher
• Kokoro ONNX model files (kokoro.onnx, voices-v1.0.bin) in a directory accessible to the application (e.g., /home/pg/Dokumente/Kokoro-82M/)
• A compatible audio backend (e.g., pygame for playback)

Steps

1. Clone the repository:

   git clone https://github.com/Patrick-Ric/kokoro-voice-blender-gui.git
   cd kokoro-voice-blender-gui

2. Create and activate a virtual environment:

    python -m venv venv
    source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Ensure the Kokoro model files (kokoro.onnx, voices-v1.0.bin) are in the expected directory, in the same directory as the kokoro-voice-blender-gui. It is advisable to also have the “kokoro-tts-gui” in the same directory so that the mixed voices in the “/config” subdirectory can be easily swapped between the two programs (default: /home/pg/Dokumente/Kokoro-82M/). Update the paths in kokoro_voice_blender_gui.py (model_path, voices_path, config_dir) if your setup differs.
5. Run the application:

   python kokoro_voice_blender_gui.py