kbd-audio is a collection of command-line and graphical tools for capturing, visualizing, and analyzing keyboard audio. Its Keytap utilities explore whether pressed keys can be identified from sounds recorded through a nearby microphone. The original Keytap uses training recordings, while Keytap2 relies on English letter and n-gram frequency statistics. Keytap3 improves the algorithms and automates the recovery of unknown typed text without manual intervention. Additional tools record audio, play captured samples, detect individual keystrokes, and display waveforms for inspection. The project supports Linux, FreeBSD, macOS, and Windows through CMake, SDL2, and optional FFTW3 components. It is intended for acoustic security research and authorized vulnerability testing.
Features
- Microphone-based keystroke detection
- Training-based acoustic classification
- Automated n-gram text recovery
- Audio recording and playback tools
- Graphical waveform visualization
- Cross-platform C and C++ builds