dsh-voice-by-edge

Conversations in the DSH web app often read like static text: useful, but quiet. dsh-voice-by-edge adds a Voice Practice mode that makes the exchange feel more like a spoken drill. Once enabled, the model responds in short, conversational sentences, the browser reads the assistant turns aloud, and a microphone button lets you dictate your reply using the Web Speech API. The result is a simple listening, pronunciation, and response-speed practice loop without leaving the web interface.

For audio, the plugin prefers Microsoft Edge neural voices through edge-tts, using suitable English and Chinese voices and automatically splitting bilingual content so each segment is read with the right voice. If the Edge service cannot be reached, it falls back to the browser's local speechSynthesis engine, so the reading feature remains available. The server-side prompt also encourages short spoken English or Chinese, avoids dense markdown, and can add translations and gentle corrections, which makes the replies more useful as practice material. Speech recognition supports real-time typing into the composer and optional sending after dictation, with useful feedback if permissions, browser support, or network access are the issue.

This plugin is aimed at DSH web users who want to improve listening and speaking, rehearse daily conversation, practice pronunciation, or use model chat as a low-pressure language partner. It fits beginners who need short, easy turns as well as more advanced users who want fast, realistic speaking practice in a browser-only workflow.