Animation Rotation Question

[nopjia]

Under Animation section:

The animation's parameters define the inputs to the animation: a set of floating point scalar values representing time (the "TIME" input); and sets of three-component floating-point vectors representing translation, rotation, and scale. Each of these inputs is stored in a buffer and accessed via an accessor.

Under sampler.interpolation:

When an animation targets a node's rotation, and the animation's interpolation is "LINEAR", spherical linear interpolation (slerp) should be used to interpolate quaternions.

A couple questions:

• Providing a 3-component Euler angles alone is ambiguous. We're missing the rotation order. Without it you cannot correctly construct the rotation since different engines have different default rotation orders.
• Is it valid to directly store 4-component quaternions for rotation instead? This would solve the ambiguity plus avoid the conversion at runtime. If so, how do we distinguish it from the 3-component Euler rotation? Would it be "VEC4" under accessor.type?

[pjcozzi]

Hi @nopjia, I'm excited that you are checking out glTF! Let me know what you think.

In glTF, animated rotations are defined by quaternions (x, y, z, w). They were defined by axis-angle in and older version of the spec (#144), and this is just a typo. I fixed it in the new 1.0-final branch. Let me know if you find any other errors in the spec.

For an example glTF model with animated rotations, see https://github.com/KhronosGroup/glTF/tree/master/sampleModels/CesiumMilkTruck/glTF

[nopjia]

Thanks for the clarification @pjcozzi!

We've been using glTF 0.8 for a couple months now, and it has served us very well. We've been privately maintaining an Assimp exporter plugin in-house, which we hope to open-source at some point.

Will definitely give you feedback when I come across them, and hope to make more contributions in the future. Looking forward to the 1.0 spec!

[pjcozzi]

Wow, @nopjia that exporter would be great. You might want to coordinate with @kimkulling: #456 (comment)

[pjcozzi]

Nothing left to do here.