TTSLab

TTSLab offers on-device speech processing by running text-to-speech (TTS) and speech-to-text (STT) models directly in your browser. Key features include: - Runs models locally using WebGPU and WASM - No server interaction, API keys, or data collection - Supports models like Kokoro 82M, Whisper, and Moonshine - Local caching of models for instant reuse - Side-by-side voice comparison and performance benchmarking This platform eliminates privacy concerns by ensuring your text and audio never leave your device. It provides a secure environment for processing sensitive information, making it ideal for applications requiring strict data confidentiality. The WebGPU acceleration removes network latency, enabling real-time applications such as voice agents and live captioning directly within the browser. TTSLab is built for developers, researchers, and product teams needing a fast, private, and standardized environment to evaluate speech models. It's fully open-source (MIT) and allows users to compare different voices, run benchmarks, or interact with a voice agent, all on their local machine. This browser-native approach simplifies model testing and integration workflows.