As you know, video replay of the most recent motion detected by the doorbell takes quite a while to process before we can see it. The result is that we hear the alert “motion detected at your front door” but, by the time we can review what’s going on, it’s long gone.
To solve this, store a small buffer of snapshots - just jpegs - that record only the most recent few seconds. When “motion detected” occurs, display one or more of these images so we don’t have to wait for your site to build us a .mpeg file. A rapid auto-scroll through these images would make an acceptable pseudo-movie.
The buffer would cycle such that you’re only storing the last five or ten seconds. Storage requirements would be modest. Encoding .jpegs is fast, and if image quality is kept low, file size could be kept down to ensure rapid transmission.