TTSLab lets you run and compare TTS & STT models directly in your browser using WebGPU and WASM — no server, no API keys, no data collection. Your text and audio never leave your device. Run Kokoro 82M, Whisper, Moonshine, and more. Models are downloaded once and cached locally. Compare voices side by side, run benchmarks, or chat with a Voice Agent — all on-device. Built for developers, researchers, and product teams. Fully open source (MIT).

Hey Product Hunt! 👋 I built TTSLab because I kept running into the same problem: evaluating TTS models is painful. Every service has different APIs, pricing, and quality — and you never really know how they sound until you're deep in integration. So I built a place where you can just open your browser and test them. No signups, no API keys, no sending your text to a server. Models run entirely on your device via WebGPU and WASM. TTSLab supports both TTS (text-to-speech) and STT (speech-to-text) models — including Kokoro 82M, Whisper, Moonshine, and more. You can compare voices side by side, run benchmarks, and even try a Voice Agent demo powered entirely on-device. It's fully open source under MIT. Would love your feedback on which models to add next!

Since certain parts of the generated audio might be suboptimal, what do you think about adding a small audio editor as well such that you will need to cut out and replace just the badly generated section?

TTSLab

Test TTS & STT models in your browser. No server required.

Test TTS & STT models in your browser. No server required.