How long AI video generation takes
A short AI video takes a few minutes to generate, and the wait scales with the length of the clip and the model you picked — not with how fast your phone is, because none of the work happens on your phone.
What the wait is made of
Almost all of it is the model generating frames on a GPU in a datacentre. A clip is generated as a whole rather than streamed, so there is no partial result to show you at the halfway mark.
Length is close to linear: a clip twice as long takes roughly twice as long. Model choice is the other multiplier, and the gap between the cheapest and the most expensive model is large.
Why you can close the app
The job runs on the server and finishes whether or not the app is open. Leaving the screen, switching apps or locking the phone does not cancel it — the finished video appears in your library when it lands.
This is also why a poor connection slows the upload and the download but not the generation itself.
Ways to wait less
Pick a Fast variant of a model. They exist specifically to trade a little fidelity for a much shorter render.
Test on a short clip. Get the photo and the template right on something cheap and quick before committing to a long premium render.