YOLO v26

YoloNDI: I Taught a Broadcast Camera to See

Because apparently just looking at video wasn't enough anymore.

YoloNDI takes any NDI video source on your network, runs YOLO26 AI object detection on every single frame, and spits the result back out as a new NDI stream. Yes, in real time. No, your CPU didn't sign up for this.

Three modes:

Detect slaps bounding boxes on everything it recognises, complete with motion trails and velocity arrows because why just find objects when you can also tell them how fast they're moving.

Pose draws a skeleton of any person detected, also on black — compositing-ready with no input video behind it.

Segment cuts detected objects out onto a black background.

So What Can You Actually Do With This

- Live production — instant object isolation into your video mixer, no green screen, no gaffer tape, no excuses

- Sports analysis — real-time velocity vectors per player, for when spreadsheets just aren't nerdy enough

- Interactive installations — feed the skeleton stream into a generative visualiser and tell everyone it's art

Built in Python, compiles to an exe, works with anything that outputs NDI. Setup takes ten minutes. Questioning your life choices is optional but likely.