I think the main issue is that animation should be restricted to 1. where the user is currently looking and 2. where they
should be looking [1]. When you're reading a text, you're looking at the text and ingesting it, so animated images will be irritating. If you really need to do an animation to illustrate your point, it should be click-to-play, so that it only moves when the user is actually purposefully looking at it.
[1] An example for this is how an app's symbol jumps on the macOS dock when there's activity in that app. (Although I would say that this particular instance is just a bit too annoying. When it jumps repeatedly, I frequently reply in my head "I KNOW that you have something to show me, but I need to finish this other thing first!".)