From raw footage to a scene you can chat with.
Three steps turn a perspective camera clip into calibrated 2D trajectories and a natural-language assistant grounded in exactly what happened in the video.
Upload your video
Drag in any CCTV, traffic camera, or drone recording — MP4, MKV, MOV, or WebM, up to 24 hours long. No special hardware or smart cameras required; existing footage works.
AI analyzes it
Dedicated GPU workers detect every person, vehicle, and bicycle, calibrate the camera view against a real-world map using homography, and reconstruct full entrance-to-exit trajectories at up to 94% precision.
Chat with your scene
Ask plain-language questions — "who loitered by the entrance?", "what was the peak hour?" — and get answers grounded in real counts, timestamps, and anonymous IDs you can click to highlight on the map.
What makes the analysis trustworthy
Every answer you get back is traceable to a detection in the original footage.
Homography calibration
Camera pixels are mapped to real ground coordinates, so distances and headings are metric, not guesswork.
Zero hallucination answers
The chat assistant only reports numbers it computed from the actual trajectory data — no invented figures.
Anonymous by default
Each target is a sequential ID (#1, #2, #12) — no facial recognition or license-plate extraction, ever.
Try it with your own footage
Upload a clip now, or check pricing first to see what's included.