dsh-plugin-voice: A Voice Broadcast and Smart Call Plugin for DeepSeek Harness
doer1296/dsh-plugin-voice
DSH plugin for voice output: cloud TTS via Volcano seed-tts (fallback to SAPI offline), desktop notifications, and scene-based WAV sounds, native on Windows with zero Python dependency.
dsh-plugin-voice is the voice and notification outlet for DeepSeek Harness, letting agents proactively reach you with sound. It integrates Volcano seed-tts cloud synthesis for high-quality speech, automatically falls back to Windows offline voice when network or key fails, and comes with scene-based WAV chimes and a smart calling mechanism—round-end reminders, a 60-second confirmation window, and a voice call to bring you back if you leave. All configuration happens in the DSH settings panel, with zero Python dependencies and native Windows support.
How to Install
dsh plugin --profile web add dsh-plugin-voice- Category
- Tools & Capabilities
- Platform
- DSH Plugin
- Author
- doer1296
- Distribution
- Plugin
dsh-plugin-voice Key Features
dsh-plugin-voice Repo Summary
What Does It Do?
dsh-plugin-voice is a DSH plugin for DeepSeek Harness that adds voice and notification capabilities, letting agents proactively contact users via cloud TTS, desktop notifications, and sound effects. It solves the problem of agents being unable to speak up, allowing you to receive voice broadcasts, scene-based chimes, and smart callbacks directly in the harness without manual effort. Core capabilities include Volcano seed-tts cloud synthesis, automatic fallback to offline speech, scene-specific WAV sounds, and automatic calling when the agent is waiting for input.
Core Features
- High-quality cloud TTS: Volcano seed-tts large-model streaming synthesis with 6 emotion acoustic mappings.
- Smart fallback: Automatically switches to Windows SAPI offline speech when Volcano fails (network issues, quota, or key invalidation).
- Scene-based chimes: 5 scenes each with a built-in WAV sound, followed immediately by voice.
- Smart calling: Round-end notification → 60-second confirmation window → if you leave and time out, voice calls you back; also auto-calls when the agent waits too long for input.
- Long-text optimization: Splits text by sentence for parallel synthesis and cleans text (skips code blocks, URLs, Markdown markers).
- Settings panel: Engine, voice, volume, speed, key, scene templates, and sounds, all effective in real time with preview support.
How to Use This Plugin?
After enabling it in DSH, simply chat to trigger model-initiated broadcasts, for example, say "Check the project status and call me by voice when done." You can also use the /voice command for manual broadcast or notifications, such as /voice Task complete or /voice --scene=task_complete --emotion=happy Done. All common configurations are done in the DSH Settings → Voice panel; if no key is configured, it automatically uses SAPI offline speech.
This page is an independent rewrite of the plugin's official README — for authoritative documentation and the latest changes, refer to the source: doer1296/dsh-plugin-voice. The plugin is third-party code that runs on your machine once installed; inclusion does not imply endorsement — please review the source before installing.