Countenance overlays live facial-expression cues on your calls and interviews — 100% on-device ML, nothing leaves your browser. 察言观色 (chá yán guān sè) — an old Chinese idiom for reading a person's words and expression together to understand what they really mean. That's what Countenance does for your calls, interviews, and meetings. Countenance overlays each visible face on screen with its detected expression, live — happy, neutral, sad, angry, surprised, disgusted, or fearful — so you can read the room as the conversation actually moves, not just replay it afterward in your head. 100% on-device. Two machine-learning models — face detection and expression classification — run entirely inside your browser using ONNX Runtime Web. No video, audio, or image is ever uploaded; there is no server. No account, no sign-up, no tracking. Session history, optional and yours. Turn on logging to keep a local timeline of expression changes for each call — stored only in your browser, deletable anytime, with an auto-expiry you control. Works wherever you meet. Any browser-based video call, interview platform, or meeting tool — Countenance isn't tied to one service.