The VRAM expression can tell you how much VRAM is being used. The next build will introduce features that allow you to choose which parts of the application are loaded in to VRAM and when, which will help reduce your usage a lot.
As mentioned, images are fully decompressed (to alpha-red-green-blue format, 4 bytes per pixel) and placed on a texture a size of a power-of-two, eg. 512x512 for a 300x300 image.
Large animations with lots of frames are an extremely inefficient way to use VRAM. You can end up using tens of megabytes from one object. A much faster and more conservative way is to use static images and move them about to create an animation. The end result is actually smoother than you can achieve with frame-by-frame animations, too. If you play with the bone movement, you should see the effects you can get with six unanimated sprites, which would use hardly any VRAM at all. The same goes for explosions and so on: importing a really high resolution explosion animation with loads of frames is a huge waste of memory. With a few fading, moving sprites, a couple of particle effects (which use basically nothing, if you use say a 32x32 particle texture), you can actually create a much more interesting explosion, and it'll use hardly any VRAM at all.
If you're worried about VRAM, change your technique to one which doesn't revolve around basically loading entire video files on the GPU.