
Open media
Transform your environment into a cognitive, zero-latency computer vision platform. No complex installations, no cloud subscriptions, just a lightweight ZIP file that runs pure, local edge intelligence on any PC, Mac, Linux, and Raspberry Pi

Traditional Network Video Recorders (NVRs) are often heavy, standalone systems isolated from the rest of your digital ecosystem. With HomeGenie, we completely flipped this paradigm since the starting point is an intelligent and programmable system not just a recorder.
VisionCore is the ultimate practical demonstration of this. To achieve it, we didn't build a massive, monolithic application. We simply added two new "neurons" to the existing automation engine: the Video Input and Video Recorder programs.
Together, these two lightweight modules create a state-of-the-art Computer Vision interface seamlessly integrated with HomeGenie's AI capabilities. By leveraging the internal YOLO AI engine, VisionCore doesn't just record your environment; it analyzes it in real-time.
This unlocks a whole new world of possibilities that extends far beyond a traditional security NVR. Whether you are building advanced smart home routines, developing autonomous robotics, or optimizing industrial automation, VisionCore allows you to trigger complex logic, feed real-time environmental context to Local LLMs, and generate deep analytical statistics based purely on visual data.
VisionCore centralizes everything you need into a single, highly responsive player interface. For each video input, you have total control at a glance:
This is where VisionCore turns raw video into actionable insights. Instead of wasting disk space on hours of empty footage, HomeGenie records and reacts only to what matters.
person, car, bicycle, backpack, truck, and more.VisionCore is an engineering powerhouse designed for extreme efficiency, bridging the gap between simplicity and professional power.
Traditional web-based NVRs rely on HTTP polling, which hits a hard wall due to browser connection limits, crashing your framerate if you open too many streams. We threw that out the window. VisionCore utilizes a multiplexed binary transport via WebSockets and MessagePack. The result is total latency elimination, allowing you to stream multiple cameras at buttery-smooth framerates through a single, highly optimized communication pipe.
VisionCore transforms cameras into high-precision data sensors. With native support for the latest YOLO26 models and advanced Pose Estimation (skeletal tracking), HomeGenie seamlessly merges live video with AI telemetry. The dashboard renders real-time analytical charts, empowering you to track exact object counts, human movement, and environmental trends instantly.
Deploying advanced Computer Vision shouldn't require massive Docker containers or complex dependencies. HomeGenie is a self-contained, portable ZIP file of less than 200MB. Just install FFmpeg, extract the archive, and run it on Windows, Mac, Linux, or ARM edge devices like the Raspberry Pi.
Configuring a new video input requires only a URL and your desired resolution/FPS. It accepts virtually any source:
rtsp://192.168.1.50/stream1http://192.168.1.51/videostream.mjpeg.mp4 files directly, providing a perfect sandbox for testing and fine-tuning your AI algorithms.Because VisionCore is built natively using HomeGenie's C# Automation Programs, it provides an unparalleled level of transparency and control.
Customize Your View: From the player's settings menu, resize the viewer, change the camera name, or open the Data Monitor to log and chart any real-time parameter.
Integrated IDE: You can view, edit, and extend the C# NVR algorithms directly in the HomeGenie Web UI. Prototyping complex logic requires no external IDE.
Add Your Own AI Models: Want to integrate a custom ML.NET predictive analysis or a new neural network? With just two lines of code, you can declare your new algorithm to the system, and VisionCore will automatically add a native toggle button for it right into the player's interface:
// Let the UI know about your custom AI algorithm
Program
.Emit("NVR.Vision.Algorithm", "ML.MyCustomAI")
.Emit("NVR.Vision.Algorithm.Icon", "smart_toy");graph TD
subgraph "VIDEO SOURCE"
Camera["📹 IP Camera / MP4 File"]
end
subgraph "LIVE PROCESSING (VideoInput Program)"
VideoInput["VideoInput Program"]
FFmpeg_Live["⚙️ FFmpeg (Live Transcoder)"]
Camera --> VideoInput
VideoInput -- "Starts & Manages" --> FFmpeg_Live
FFmpeg_Live -- "Writes to disk" --> HLS["🌐 Live HLS Stream"]
FFmpeg_Live -- "Writes to disk" --> Snapshot["🖼️ Live Snapshot"]
VideoInput -- "Emits Parameter" --> NVR_Path["HG Parameter: NVR.StreamPath"]
end
subgraph "USER INTERFACE"
UI["🖥️ Dashboard & Player"]
HLS --> UI
Snapshot --> UI
UI -- "Manual Commands" --> Scheduler
end
subgraph "TRIGGERS"
AI["🧠 AI Engine (YOLO26)"]
Scheduler["📅 Scheduler / User"]
AI -- "Emits Event" --> SensorDetect["HG Parameter: Sensor.ObjectDetect"]
end
subgraph "RECORDING & STORAGE (VideoRecorder Program)"
VideoRecorder["VideoRecorder Program"]
FFmpeg_Rec["⚙️ FFmpeg (Recorder)"]
Recording["💾 MP4 Recording"]
NVR_Path --> VideoRecorder
SensorDetect --> VideoRecorder
Scheduler --> VideoRecorder
VideoRecorder -- "Starts & Manages" --> FFmpeg_Rec
HLS -- "Input (Stream Copy)" --> FFmpeg_Rec
FFmpeg_Rec -- "Writes to disk" --> Recording
end
Looking at the architecture above, the true power of VisionCore is that none of this flow is locked away in a compiled black box. Both the VideoInput and VideoRecorder modules are fully exposed C# Automation Programs. You can open, edit, and rewrite their logic directly from the HomeGenie Web UI to fit your exact needs.
But it goes much deeper than just tweaking capture settings. The VideoRecorder is explicitly architected to support live frame post-processing before the footage is ever written to disk.
This capability completely transforms the system from a passive NVR into a powerful AI video generation pipeline. You can inject your own custom machine learning algorithms directly into the recording flow. Want to feed a local .mp4 file through a custom neural network, draw real-time analytical overlays on every single frame, and export a brand-new, AI-processed batch video?
With VisionCore, you aren't just recording cameras, you have a complete, edge-computing studio ready to bring your most advanced Computer Vision experiments to life.
[Download HomeGenie Today] and take absolute control over your environment with the ultimate local edge-intelligence orchestrator.
Because VisionCore is built on HomeGenie’s native unified bus, every function is exposed as a standardized API. These endpoints can be called locally via HTTP REST, streamed in real-time via WebSockets, or triggered remotely through MQTT using JSON payloads.
Below are the core REST API endpoints with practical examples.
Media.Video/)These endpoints are handled by the Video Input module and are routed using the camera's unique address.
Camera.GetPicture
GET http://<homegenie_address>/api/Media.Video/1/Camera.GetPicture(Retrieves the instantaneous live image frame for Video Input "1".)Camera.MotionDetected
Sensor.MotionDetect event for the specified video input. Perfect for bridging external physical sensors (like a Zigbee PIR) to wake up the video processing or trigger a recording.HomeAutomation.HomeGenie/NVR/)These endpoints manage the storage, retrieval, and metadata of your recordings. They are structured to accept the target camera's Domain and Address to locate the correct footage.
Recordings.Days
TotalSize (bytes), TotalTime (seconds), and SessionCount for instant dashboard rendering.GET http://<homegenie_address>/api/HomeAutomation.HomeGenie/NVR/Recordings.Days/Media.Video/1(Returns the daily summary of all archived recordings for Video Input "1".)Recordings.List
ffprobe (e.g., resolution, fps, video/audio codecs, and exact timestamps).GET http://<homegenie_address>/api/HomeAutomation.HomeGenie/NVR/Recordings.List/Media.Video/1/20260228(Fetches the detailed metadata JSON for all clips recorded by Video Input "1" on February 28, 2026.)Recordings.Thumbnail
GET http://<homegenie_address>/api/HomeAutomation.HomeGenie/NVR/Recordings.Thumbnail/Media.Video/1/20260228/143000/15.5(Extracts and returns the video frame exactly 15.5 seconds into the recording session that started at 14:30:00 on Feb 28, 2026.)Recordings.Start / Recordings.Stop
Recordings.Check