About TTSLab
TTSLab is an open-source tool for testing and comparing text-to-speech (TTS) and speech-to-text (STT) models directly in your browser. No servers, no data collection, no API keys required.
How It Works
WebGPU Powered
Models run directly in your browser using WebGPU acceleration. No server-side inference means instant results with no queue times.
Complete Privacy
Your text and audio never leave your device. All processing happens locally in the browser. No data is sent to any server.
No Setup Required
No Python, no dependencies, no API keys. Just open your browser and start testing models. Models are downloaded once and cached locally.
Methodology
TTSLab provides a standardized environment for evaluating voice AI models. By running models in the browser with WebGPU, we ensure consistent and reproducible results across different hardware.
Each model is loaded using its ONNX-optimized variant when available, ensuring the best balance between quality and performance for client-side inference.
Comparisons are done side-by-side with shared inputs, allowing you to directly hear and evaluate the differences between models on identical text or audio.
Open Source
TTSLab is fully open source under the MIT License. We believe that tools for evaluating AI models should be transparent and community-driven.
The entire codebase, including the model integration layer, UI components, and database schema, is available on GitHub.
Contributing
We welcome contributions of all kinds. Here is how you can help:
- Add a model — Integrate a new TTS or STT model with an ONNX export.
- Improve the UI — Fix bugs, improve accessibility, or add new features.
- Write documentation — Help others get started with clear guides.
- Report issues — Found a bug or have a suggestion? Open an issue on GitHub.