Show HN: Hushkey – Silent dictation hotkey for macOS

Category: ai-ml

Tags: silent-speech, macos, computer-vision, dictation, lora, privacy

Score: 6.8/10 (Innovation: 8, Technical: 7, Documentation: 8, Utility: 4)

Hushkey is a macOS proof-of-concept that enables silent dictation by reading lip, jaw, and throat motion from the webcam instead of audio. It combines a frozen lip-reading model with LoRA adapters, a kinematic branch for throat motion, and personalization training on local clips, all running locally. The project is technically ambitious, well-documented, and cites clear prior art, but it is an early-stage prototype with tight hardware and platform requirements.

Target audience: macOS users with accessibility needs, ML researchers, and developers interested in privacy-preserving human-computer interaction

Repository: https://github.com/amirrsyed/hushkey · Python · MIT

View on Hacker News