Sure.
First, the runtimes have two loaders, one for JSON and one for binary. Both do the same thing, which is to load the skeleton data and animations from files. The skeleton data contains the bones and slots for the setup pose, and also the skins which contain all the attachments. You create a skeleton from the skeleton data. The skeleton holds state for one on-screen instance, such as the current pose of the bones and which attachments are in which slots. Any number of skeletons can use the same skeleton data.
To animate your skeleton, you pass it to the animation along with a time that the skeleton has been in that animation state. Eg, from the libgdx SkeletonTest example:
Code: Select all
animation.apply(skeleton, time, true);
skeleton.updateWorldTransform();
skeleton.draw(batch);
The apply method poses the skeleton at the specific time (the boolean true is for whether the animation should loop if time exceeds the animation's duration). Posing the skeleton just modifies the bones' local position, rotation, and scale. The updateWorldTransform method updates the world position, rotation, and scale for each bone. The draw method uses the world positions and draws all the attachments in the correct place. You can also access the local and world positions of the bones, eg to position particle effects.
Animation mixing works like this, from the libgdx MixTest example:
Code: Select all
walkAnimation.apply(skeleton, time, true);
jumpAnimation.mix(skeleton, time - beforeJump, false, 1 - (time - blendOutStart) / blendOut);
The apply method poses the bones with the walk animation. The mix method blends the current pose (which will be walking) with the jump animation. The mix method takes the same parameters as apply, with an extra one which is how much to interpolate between the current pose and the new pose (0 to 1). The math to compute the time and blending amount looks tricky, but with a proper state machine to manage the animation states like you'd have in a game, it is straightforward. This animation mixing lets you transition smoothly from one animation to another.
When we get to it, you'll be able to define bounding boxes in Spine, then can do hit detection using built-in method. For now, the runtimes consist of loading the data, providing a nice object model for the bones, slots, skins, attachments, aniamtions, etc and manipulating the skeleton either directly (eg attaching a specific item or setting a bone's rotation) or by applying animations. Drawing is also handled.
