Blender is a magnificent free open-source 3D modelling software. Latest Blender versions include a working X3D exporter, so you can export your content from Blender out-of-the-box and open it with our engine. We advice testing the X3D models by opening them with view3dscene.
We have our own X3D exporter version, which is based on the Blender X3D exporter
and adds some small fixes. In time, we hope to have them submitted to official Blender code.
Install this by copying over (overwriting) the original Blender exporter,
which is in
The Blender X3D exporter cannot handle animations for now.
To solve this, use the exporter below to export animations to
a Castle Animation Frames format. The
.castle-anim-frames files can be read by our engine
and will play animations.
Install it like every other Blender addon:
Actions are Blender containers for animations. Each Blender object may have many actions, like walk, run, die... A new action is automatically created on an object (if needed) when you insert a keyframe.
In Blender, it matters to which object you attach an action. Action describes the complete animation of a given Blender object. If you try to reuse the same action on two different objects, you will find that they animate (move, rotate...) the same way. If you want every object to animate in a different way, you will usually use an action only on a single object. I explain this, to make it clear that Blender actions do not span multiple objects, which is a little different than what we need (we want to export a series of animations, and each animation should just apply to the whole scene).
When exporting the animation, you can select an object to export all the actions of this object. If your scene includes an armature with some actions, we automatically select it as the object from which to take actions (you can deselect it, if desired). The range of exported frames is determined by the minimum and maximum keyframe set in this action (that's how Blender calculates
action.frame_range in Python).
When you don't select any such object then we export the whole animation (from Start to End frames that you set on the Timeline). The resulting animation will be called just "
animation" in this case (this is useful if you run animations from code, using the PlayAnimation method). This is perfectly reasonable in many situations:
Thanks to the simplicity of the
this format exports every kind of Blender animation to our engine:
castle-anim-frames animations render smoothly when the models are "structurally equal", which means that you should only animate things that can be interpolated. You should not change a topology (e.g. create, destroy faces or whole objects) from one frame to another, at least you should not change it too often. In particular:
Ignoring these advices will make the animation "jump" at certain frames, since the engine will not be able to interpolate between the frames you provided. Sometimes this is OK (e.g. when you really change one object to something completely different), but sometimes this is a bad artifact. Use the view3dscene with
--debug-log command-line option, and watch for warnings about the model not being "structurally equal", if you want to eliminate such issues.
castle-anim-frames exporter uses the X3D exporter (the original one, provided with Blender, or our custom one if installed) to export the static data. So if something doesn't look like you want after exporting, consult the advices above about using the X3D exporter. You can always export, as a test, a single frame of your animation to X3D, to check does it look OK.
TODO: Right now the interpolation is not done using the proper X3D interpolators at runtime, like
PositionInterpolator. Instead we interpolate (create intermediate frames) at the load time, then put the resulting frames inside a
Switch node, animated using the
IntegerSequencer. The nodes inside all
Switch nodes are shared smartly, but still the memory usage is much higher than it could be.
For the sake of the collision detection, each animation is treated as it's bounding box. Calculating detailed collision trees for every animation frame would be too time-consuming. TODO: We can improve it one day to collide as a different bounding box, corresponding to the current animation frame, not to the whole animation.
You can render a set of six images that can be used as a skybox (
Background in X3D, see documentation of the Environmental effects component) or a cube map texture (
ComposedCubeMapTexture in X3D, see documentation of the Cube map environmental texturing component). We have a small Python script for Blender that renders the images to the appropriate names following the X3D conventions (front, back, top,....), and a simple X3D test scenes that allow to test that the resulting images indeed work as a skybox or a cubemap texture (e.g. you can open them in view3dscene).