Hitorigoto

Talk to yourself in English, and AI reviews what you said. Press Start, speak, press Stop. You get up to 5 suggestions (phrasing, vocabulary, grammar), a corrected version for reading aloud, and a note on what went well. - No API key needed: runs on Chrome's built-in AI (Gemini Nano), so your words never leave your device - The built-in AI (Gemini Nano) is a lightweight model, so suggestions have accuracy limits. Optionally add your own Gemini API key for higher accuracy. Accuracy will keep improving with updates. - Tracks your recurring mistakes and shows them before each practice - With your own Gemini API key, the recorded audio itself is sent to Gemini for transcription and feedback in one pass (v1.1.0+, can be turned off in Settings). Fewer "corrections" of things you never said due to speech-recognition errors, and words that didn't come across are shown separately - With on-device AI (text only), there is no pronunciation feedback: your audio never reaches the AI, and guessing pronunciation from text would mean making things up - Likely speech-recognition errors are shown separately, not counted as your mistakes - Add phrases you want to learn: a few come up each day as "today's phrases", and the app checks whether you used them. Phrases you got right come back after a while; three hits graduate them (v1.2.0+) - Stuck for a word? Say it in Japanese. The coach returns two or three natural ways to say it in English, with pronunciation (IPA) (v1.2.0+, with your own Gemini API key) Built-in AI requirements: Chrome 138+, 16 GB RAM recommended, ~4 GB one-time model download. On unsupported devices, use a Gemini API key instead.