Skip to content
  • Home
  • Plugin manual
  • dsh-voice-mic

Zachary7456/dsh-voice-mic

30Last commit Aug 15, 2026

dsh-voice-mic DSH plugin

dsh-voice-mic adds real-time speech-to-text to the DSH Web interface. It offers three recognition engines: browser built-in (Web Speech API), local offline backend (sherpa-onnx with SenseVoice/Paraformer), and cloud API (OpenAI-compatible). Users can record with a mic button or hotkey (Alt+V) and see results appended to the input box without auto-sending.

browser-webui-productivity

How to install the dsh-voice-mic DSH plugin

dsh plugin --profile web add github:Zachary7456/dsh-voice-mic

Copying does not run this command. Review the repository and version before installing the dsh-voice-mic DSH plugin.

dsh-voice-mic DSH plugin data source

dsh-voice-mic DSH plugin snapshot date: Aug 16, 2026

discovered

What the dsh-voice-mic DSH plugin can do

  • Real-time transcription written into the input box during recording
  • Three recognition engines: browser built-in, local offline backend, cloud API
  • One-click deployment of local backend (Python + model download with resume)
  • Supports Chinese, English, Japanese, Korean, Cantonese (SenseVoice models)
  • Hotkey Alt+V (configurable) to toggle recording
  • Cloud API compatible with OpenAI, Groq, SiliconFlow, etc.
  • Local backend binds to 127.0.0.1, audio never leaves the machine
  • No build step needed; client-side changes reflect on page refresh

Where the dsh-voice-mic DSH plugin fits

  • Hands-free text input in DSH Web GUI during conversations
  • Offline speech recognition with local models for privacy
  • Multi-language transcription for international users
  • Integration with existing OpenAI-compatible endpoints for custom ASR
  • Quick note-taking or command input via voice

Who the dsh-voice-mic DSH plugin is for

  • DSH Web GUI users who prefer voice input over typing
  • Developers needing to extend DSH with voice capabilities
  • Users in environments where typing is inconvenient (e.g., accessibility, mobility)

dsh-voice-mic DSH plugin limitations

  • Browser recognition requires Chrome/Edge (Web Speech API)
  • Hotkey only works when the page has focus (browser restriction)
  • Local backend requires Python 3.9+ and model download (~158–223 MB)
  • Cloud API incurs usage costs; API calls made ~1 per second during recording
  • DSH restart requires re-deploying local backend unless autoStart is configured
  • Local backend only binds to 127.0.0.1, not accessible from other hosts

dsh-voice-mic DSH plugin: from the repository README

Quoted from the Zachary7456/dsh-voice-mic README, the upstream source of the dsh-voice-mic DSH plugin. Copyright remains with the original authors.

DeepSeek Harness Web GUI 语音输入插件。录音后实时转写并写入输入框，不自动发送。 ## 安装 要求：dsh `>=0.1.0-rc.6`、Node.js `>=18`。浏览器识别需 Chrome/Edge；本地后端另需 Python `3.9+`。 ```bash dsh plugin --profile web add github:Zachary7456/dsh-voice-mic # 或（npm 发布后）：dsh plugin --profile web add dsh-voice-mic ``` 安装后重启 `dsh web`。插件带 bundle patch，会自动挂进 profile，无需手动改 `cordis.patch.yml`。 验证： ```bash dsh --profile web --dump-config | grep dsh-voice-mic curl -s http://127.0.0.1:3080/plugins/dsh-voice-mic/client.js | head -c 100 ``` ## 使用 - 输入框左侧麦克风按钮：点按切换录音，默认快捷键 `Alt+V`（设置页可改） - 录音期间识别结果实时写入输入框草稿；停止后提交，不自动发送 - 快捷键仅在页面获得焦点时生效（浏览器限制） - 实时写入采用追尾替换：只更新上次追加的尾缀，不覆盖手动输入；无结果或出错时自动回滚 ## 识别引擎 设置 → 语音输入 → 识别引擎，三选一。 ### 浏览器内置（默认） 零配置，走 Web Speech API。中文质量与延迟取决于浏览器与网络。说话停顿导致浏览器自行断开识别会话时，插件自动重启识别器继续监听，直到手动停止。 ### 本地离线后端 设置页一键部署：检测 Python → pip 安装依赖 → 下载模型（断点续传，可取消）→ 启动 `asr_server.py` → 轮询就绪。已下载的模型会缓存，切换模型免重复下载。 模型： | 模型 | 适用 | 大小 | |---|---|---| | SenseVoiceSmall int8 | 中/英/日/韩/粤，日常使用 | ~158MB | | Paraformer | 仅普通话，准确率与标点最佳 | ~223MB | - 模型目录：`~/.dsh/voice/models/<model>`（可用 `DSH_VOICE_MIC_MODEL_DIR` 覆盖）；下载缓存：`~/.dsh/voice/cache` - 服务端口默认 `7860`，只绑定 `127.0.0.1` - 后端进程由 dsh 托管：dsh 重启后需重新点部署（或配置 `autoStart: true` 自动拉起） - 转写期间音频不出本机

Read the full READMERepository license: MIT

dsh-voice-mic DSH plugin questions

How do I install dsh-voice-mic?

Run the command: `dsh plugin --profile web add github:Zachary7456/dsh-voice-mic`. After installation, restart `dsh web`. The plugin includes a bundle patch that automatically hooks into the profile. You can verify with `dsh --profile web --dump-config | grep dsh-voice-mic`.

Why doesn't the hotkey Alt+V work?

The hotkey only works when the browser page has focus. This is a browser security restriction. Also ensure you haven't remapped the shortcut in the plugin settings page. You can change the hotkey in the settings or via environment variable `DSH_VOICE_MIC_HOTKEY`.

Can I use the local backend offline?

Yes, the local backend runs entirely on your machine. It uses sherpa-onnx with SenseVoice or Paraformer models. You need Python 3.9+ and the first-time setup will download the model (~158–223 MB). Once downloaded, the model is cached and offline recognition works without internet.

How do I configure the cloud API for speech recognition?

Go to Settings → Voice Input → Recognition Engine and select 'Cloud API'. Enter the base URL, API key, and model name. The plugin includes presets for OpenAI, Groq, and SiliconFlow. Your API key is stored in browser localStorage and sent directly to the provider, not through DSH.

What browsers are supported?

The browser built-in recognition requires Chrome or Edge (Web Speech API). The local backend and cloud API work with any modern browser that supports the Web Audio API. The plugin's client side is plain JavaScript, so it should work in most cases.