03 Nov 2016 · Animation Compression
One of the primary advantage of storing bone transforms in local space as opposed to object space is that it reduces dramatically the amount of data that changes. For example, if we have two bones connected, and they have their local rotation animated, we end up with two animated tracks in local space, but three tracks in object space. This is true because the rotation changes from the root bone will cause a translation offset on the second bone.
Read the full post
01 Nov 2016 · Animation Compression
Every compression algorithm covered herein is lossy in nature, so how we measure accuracy is critically important. Measured deviation from source data needs to be representative of the visual differences observed.
Read the full post
27 Oct 2016 · Animation Compression
Animation clips more or less always boil down to the same thing: a series of animated tracks. There might also be other pre-processed data (such as the total root displacement, etc.) but we won’t concern ourselves with this since their impact on compression is usually limited. Instead, we will focus on bone transform tracks: rotation, translation, and scale.
Read the full post
26 Oct 2016 · Animation Compression
Central to character animation is our animation sequence. These are also commonly called animation clips and I will use both terms interchangeably.
Read the full post
25 Oct 2016 · Animation Compression
This post is meant as a preface to explain the overall context of what is character animation, how we use it, and why it needs special consideration.
Read the full post