I took a little time to see what Qwen does. Wow. Unbelievable. Here's a general video: , Apparently it's good at OCR. So instead of trying to extract a license plate from three or four frame grabs from a four second video, as BI/Codeproject AI is doing, Qwen can look at every frame. It also strikes me that even though it needs more processing power than the typical BI machine offers, it can do its processing intermittently, say once an hour, it looks at the clips BI extracted, and analyze them, check for errors and tell the database which reads need to be corrected. Maybe if it were doing the primary analysis of the footage, it could extract plate reads the CPAI can't process because it only has three or four images to work from.