Mamorukun

Real-time speech-to-text with AI cleanup. Recognizes 11 languages. Add a Gemini API key to auto-polish your transcript.

Mamorukun turns your voice into clean, readable text right in your browser. ■ Two ways to clean up 📝 Transcription only (zero setup) Real-time speech-to-text in the side panel. Copy or clear with one click. ✨ AI cleanup — choose your engine - Chrome's built-in AI (Gemini Nano): no API key needed, and the text never leaves your device. Requires Chrome 138+, 16 GB RAM recommended, ~4 GB one-time model download. - Gemini API (your own API key): works on any machine and gives higher accuracy. Switch engines anytime in Settings. The built-in AI is a lightweight model, so cleanup accuracy has limits — it will keep improving with updates. When you stop recording, the selected engine cleans up the transcript — removing fillers, adding punctuation, and tidying spoken language into something you can paste anywhere. ■ Speaks your language Recognizes 11 languages: Japanese, English (US/UK), Spanish (Spain/Mexico), Chinese (Simplified/Traditional), Korean, French, German, and Portuguese (Brazil). You can keep the interface in one language and switch only the recognition language — useful for language learners who want to practice speaking and read back what they actually said. ■ About English practice The self-talk English practice mode has moved to its own extension, "Hitorigoto". Mamorukun now focuses on transcription and cleanup. Your past self-talk history is not deleted and can be imported into Hitorigoto. Get Hitorigoto here: https://chromewebstore.google.com/detail/aleiapbacffelpgabepjjmnbfclcodpm ■ What you need A microphone (the one built into your laptop works fine — no external mic needed), an internet connection (used for speech recognition), and microphone permission on first launch. A Gemini API key is optional (only if built-in AI is unavailable on your machine, or you want higher cleanup accuracy). Note: The API key must belong to a Google Cloud project with billing set up — keys from free-tier-only projects won't work for cleanup. ■ Privacy Speech is processed through the Web Speech API (Google's speech recognition service). Cleanup with the built-in AI (Gemini Nano) happens entirely on your device — the text is never sent anywhere. Only if you use a Gemini API key, the transcribed text is sent to the Gemini API for cleanup. Your API key is stored only in your own Chrome and is never sent to any external server.