voice-input-src

voice-input-src documents and exposes the generated source workflow for a macOS menu-bar speech input application. Holding the Fn key starts recording, and releasing it transcribes and pastes text into the currently focused field. It uses Apple's streaming Speech Recognition framework and requires macOS 14 or newer. The application supports English, Simplified Chinese, Traditional Chinese, Japanese, and Korean, with the chosen language stored locally. A floating capsule displays a live waveform and expanding transcription text while recording. Clipboard restoration and temporary input-method switching make text injection safer for CJK users. An optional OpenAI-compatible model can conservatively correct obvious recognition errors before the final text is inserted.