Right now, the implementation can only move entire images/cels based on their position on the canvas. Could it be that these frames are at the same (or very close) positions? Blending pixels is not possible at this point.
There’s also an improvement for the inbetween frames that wasn’t released yet - it’s better in correctly calculating where the images should be placed when moving them. You can download the unreleased build here, if you’re interested.
So I am seeing it move on a simpler canvas/objects. But I notice it's not quite interpolating as I expected. Like, when I rotate a things, it can't correctly predict it. I'm assuming it's not supposed to handle more complex stuff?
