Dofer TTSv1.0.0 (Stable)
100% offline Text-to-Speech generator utilizing the state-of-the-art Kokoro-82M ONNX model directly on your local CPU.
Product Overview
Dofer TTS is built to provide high-fidelity, human-like voice synthesis without relying on cloud services or external APIs. By caching the model weights locally, it guarantees total privacy, zero latency spikes, and offline usability. The app features a responsive multithreaded architecture separating speech synthesis from the user interface to ensure high performance.
Features Deep-Dive
Kokoro ONNX Inference Engine
Runs the compact 82-million parameter model natively via ONNX Runtime, offering top-tier speech synthesis quality equivalent to cloud APIs.
Responsive Multithreaded GUI
Utilizes PyQts QThread implementation to offload heavy audio processing tasks, keeping the UI fluent and responsive during generation.
Integrated Audio Player
Custom controls powered by QMediaPlayer and QAudioOutput, supporting play, pause, seek timelines, and volume sliders.
Speech Configuration Settings
Allows selecting between multiple American and British voice profiles (male & female) with adjustable speed scales (0.5x to 2.0x).
Local Caching & History
Automatically saves all generated audio as high-definition WAV files in your ~/Music/DoferTTS/ directory and loads them into a history sidepanel.
Technical Architecture
Dofer TTS uses a multi-tier setup. The front-end is written in PyQt6 with custom QSS styling sheets for a modern look. When a user requests speech generation, a worker thread launches the KPipeline/Kokoro engine. It converts input text into phonemes, synthesizes the waveform with the ONNX model, and saves the output to disk.