Show HN: TTSLab – A voice AI agent and TTS lab running in the browser via WebGPU (ttslab.dev)

A new web-based tool called TTSLab has been launched, allowing users to test text-to-speech (TTS) and speech-to-text (STT) models directly in their browsers without requiring server-side processing or data collection. Powered by WebGPU and WebAssembly, TTSLab offers a variety of models, including the fast VITS-based Piper, Microsoft’s SpeechT5, and OpenAI’s Whisper, all of which generate high-quality, natural-sounding speech in real time. This innovation caters to developers, researchers, and product teams by providing a standardized environment for evaluating TTS options privately and efficiently. The significance of TTSLab lies in its focus on on-device speech AI, which addresses privacy concerns associated with cloud-based services. Since user data never leaves the browser, sensitive content remains secure from external servers. Moreover, the use of WebGPU helps eliminate latency, enabling practical real-time applications like voice assistants and live captioning without reliance on a backend. As an open-source project licensed under MIT, TTSLab invites community contributions, making it a valuable resource for advancing accessible and user-empowered AI tools within the AI/ML landscape.