stormydayy/halo-overlay

选用指南

AI 根据 README 整理 · 2026/10/28 02:11:25

使用方式

通过 npm start 启动应用，使用图形界面进行配置和操作。

查看 README 依据

- **Profiles**: interview, meeting, sales, negotiation, presentation, study. - **Customizable shortcuts**, opacity, font size, and click-through mode. - **One-click connection test** for both the model and speech recognition. ## Quick start You'll need [Node.js](https://nodejs.org) 20 or newer and Python 3.11+. ```bash git clone https://github.com/stormydayy/halo-overlay cd halo-overlay npm install npm start ```

部署要求

需要 Node.js 20 或更高版本以及 Python 3.11 或更高版本。

查看 README 依据

- **Profiles**: interview, meeting, sales, negotiation, presentation, study. - **Customizable shortcuts**, opacity, font size, and click-through mode. - **One-click connection test** for both the model and speech recognition. ## Quick start You'll need [Node.js](https://nodejs.org) 20 or newer and Python 3.11+. ```bash git clone https://github.com/stormydayy/halo-overlay cd halo-overlay npm install npm start ```

已知限制

SystemAudioDump 仅适用于 Apple Silicon，Intel Mac 需自行编译；语音识别按停顿分块处理，延迟约1-3秒；问题检测为启发式，可能误报。

查看 README 依据

- `SystemAudioDump` in this repo is built for Apple Silicon (arm64). System audio won't work on Intel Macs until you build the helper yourself from the [source](https://github.com/Mohammed-Yasin-Mulla/Sound). - Speech recognition works in chunks split at pauses (typically a 1–3 second delay), not as a continuous stream. That's the price of being provider-independent. - The question detector is a heuristic, so false positives happen. That's why auto-answer is off by default. ## License and credits