dsh-voice-input-space

Typing in the DeepSeek Harness composer feels slow, or your hands are planted on the keyboard and you do not want to hunt for a mic button? dsh-voice-input-space binds voice dictation to the finger you are already resting on: long-press Space for half a second, speak, release, and the transcript lands at the caret. No model to download, no click-to-record workflow, zero external dependencies.

A few small design choices make the loop feel native. A 450 ms long-press threshold starts the session, and a floating panel scrolls the live transcript in real time so you always see what the engine is hearing. Press Esc or tap the cancel button to discard the take before it touches the document. Chinese input-method users get an isComposing guard, so the Space you use to confirm a pinyin candidate is never intercepted, while a quick tap still produces a regular space. The plugin ships with six languages — Mandarin, Traditional Chinese, Cantonese, English, Japanese, and Korean — switchable from Settings. All audio is processed through the browser Web Speech API; nothing is stored on disk.

If you are on Chrome or Edge, would rather have a hands-on-keyboard voice shortcut than a mic-button flow, and want dictation to be available the moment you install the plugin with no build step, this is built for you.