dsh-voice-plugin

Typing long passages with a keyboard alone can feel slow and mechanical, especially when you need to capture ideas on the fly or simply prefer a dictation-style workflow. dsh-voice-plugin drops a microphone button straight into the DeepSeek Harness composer so you can speak naturally and watch your words appear as text in real time, without switching to a separate dictation app.

The plugin ships with two speech-recognition back-ends out of the box: Doubao Voice for Chinese input and OpenAI for English and other languages. Setup is deliberately minimal—head to Settings, pick the model that matches your language, paste in the API Key, and you are done. If the key is still missing, tapping the mic simply redirects you to the settings panel rather than throwing an error, keeping the workflow interruption-free.

Whether you dictate meeting notes in Chinese, draft bilingual content, or simply want a hands-free way to compose in DeepSeek Harness, this lightweight client-side plugin removes the friction of pure text input and turns the editor into a genuine multi-modal writing space.