dsh-voice-input

Typing out long mixed Chinese-English passages in the DeepSeek Harness Web composer can be slow and tiring. dsh-voice-input adds native voice input directly to the Web draft field: tap the 🎤 icon to start continuous recognition, tap it again to stop, and the transcript lands right in your draft. The whole flow rides on the browser built-in Web Speech API, so there is no backend, no API key, and no local speech model to install or configure. A few practical notes: the locale is locked to zh-CN, so accuracy for Chinese, English, and mixed-language dictation depends on your browser engine; Chrome or Edge with Web Speech API support is required; and the audio may be routed through your browser vendor online service, so review the privacy policy before relying on it. If you spend a lot of time in the DeepSeek Harness Web writing bilingual docs, notes, or prompts and prefer dictation over typing, this lightweight plugin removes that small but constant friction and lets you focus on the content rather than your fingers.