中文版

clearcam（GitHub：https://github.com/roryclear/clearcam ，2025-01-04 创建，已 2,120★，GPL-3.0）是一个开源的智能 NVR：给任意一台支持 RTSP 的安防摄像头加上目标检测、多目标跟踪、手机通知、AI 事件摘要和内容搜索，自托管运行，录像和数据都留在自己手里。项目非常活跃，1,600 多次提交，至今仍在高频更新。

它能干什么

• 目标检测与跟踪：YOLOv9 做检测、OC-SORT 做多目标跟踪，有人、车等目标出现才触发事件与录像片段，不用傻录全天再回放找画面
• AI 通知摘要：较新的功能，用 Qwen3 VL 为事件生成一句话摘要再推送到手机，不点开视频就知道发生了什么
• 内容搜索：可以按内容检索历史事件画面
• 通知自己接管：通知发送地址可以自定义，能接 Home Assistant、Pushover、n8n 或任何自己的服务，告警直接打进自己的工作流
• 手机 App：iOS / Android 客户端同样开源，iOS 版零第三方依赖、iOS 15 以上可用，已上架 App Store 与 Google Play

技术与运行

• 推理框架用的是 tinygrad（不是 PyTorch），主体用 Python 编写，支持 GPU 加速（DEV=AMD / DEV=NV）
• 运行要求：Mac 或 Linux（Windows 还差一个 ffmpeg 修复）、ffmpeg、Python 3.11 以上
• 三步跑起来：pip install -r requirements.txt → python3 clearcam.py → 浏览器打开 localhost:8080
• 没有 RTSP 摄像头也没关系，README 里提供了一个公开测试视频流，可以先拿来体验

收费模式

• 自托管核心开源免费；Clearcam Premium（clearcam.org）提供远程查看直播与事件片段、通知中继，全程端到端加密
• 纯本地使用、通知接自己的服务，基本不用花钱

English Version

clearcam (GitHub: https://github.com/roryclear/clearcam , created 2025-01-04, now 2,120★, GPL-3.0) is an open-source smart NVR that adds object detection, multi-object tracking, mobile notifications, AI event summaries, and content search to any RTSP security camera. It runs self-hosted, so your footage and data stay in your own hands. The project is very active — over 1,600 commits and still updated frequently.

What it does

• Detection and tracking: YOLOv9 detection with OC-SORT multi-object tracking; events and clips are triggered only when people, vehicles, or other objects actually appear, instead of recording all day and scrubbing through footage later
• AI notification summaries: a newer feature that uses Qwen3 VL to generate a one-line summary of each event and push it to your phone — you know what happened without opening the video
• Content search: search past event footage by content
• Bring your own notifications: the notification endpoint is configurable, so alerts can go to Home Assistant, Pushover, n8n, or any service you run, straight into your own workflow
• Mobile apps: the iOS and Android clients are open source too; the iOS app has zero third-party dependencies, runs on iOS 15+, and is live on the App Store and Google Play

Tech and setup

• Inference runs on tinygrad (not PyTorch), the core is written in Python, and GPU acceleration is supported (DEV=AMD / DEV=NV)
• Requirements: Mac or Linux (Windows still needs an ffmpeg fix), ffmpeg, and Python 3.11+
• Up and running in three steps: pip install -r requirements.txt → python3 clearcam.py → open localhost:8080 in your browser
• No RTSP camera yet? The README links a public test feed you can try it with

Pricing

• The self-hosted core is free and open source; Clearcam Premium (clearcam.org) adds remote live viewing, remote event clips, and notification relay, with end-to-end encryption throughout
• Running it purely locally with your own notification service costs essentially nothing