dsh-voice

Typing out long prompts character by character and scrolling through wall-to-wall AI replies is exhausting for both fingers and eyes. dsh-voice wires speech interaction directly into DeepSeek Harness: a microphone button appears on the input row (tap-to-toggle or press-and-hold, much like sending a voice note), a global hotkey defaults to Ctrl+Shift+Space and can be re-recorded in Settings, and every completed AI answer gains a read-aloud button that synthesizes speech through Microsoft Edge neural TTS. Six Chinese voices plus English Jenny and Guy are available with instant preview, and none of them require an API key-install, restart DSH, and you are ready.

The core problem it solves is simple: when your hands are occupied, your eyes need a rest, or you simply do not want to type, how do you complete a full round of conversation with an AI at minimal friction? Voice input covers the say-it part, read-aloud covers the hear-it part, and both channels run entirely in the browser. The plugin touches and stores no keys of its own.

Who it is for: developers, content creators, and students who use DSH daily in Chrome or Edge, especially those who frequently type long prompts or prefer a quick audio recap of lengthy AI replies. As long as you grant microphone access and keep a network path to the Microsoft speech service, dsh-voice will quietly add a pair of hands and a pair of ears to your workflow.