I tried duplicating a segment containing the same animation.
Each segment had the same morph targets and identical keyframes.
When I played the animations, I compared the CPU load between having one segment and having two.
The difference in CPU bottleneck was quite noticeable.
This is especially problematic in interaction-heavy animations between characters with complex physics, since most processing has to be done on a single thread, leading to thread overload.
Although I don’t have in-depth knowledge of the code, after spending three hours investigating the issue with ChatGPT, here’s the hypothesis we came up with: CPU resources are consumed not only by the keyframes of the currently playing animation, but by all keyframes present in the timeline being called every frame.
I’m not exactly sure what that means technically, but if only the keyframes of the currently playing segment or animation clip consumed CPU resources, I believe it could greatly reduce CPU bottlenecks regardless of timeline size.
Of course, I wonder if there was a reason it was designed that way.
I tried to optimize the Embody plugin through ChatGPT by creating a lightweight version and adjusting the update interval in hopes of improving performance—but in the end, the original plugin itself was already extremely well optimized.
Even without understanding the code, I could tell—at least indirectly—that its creator, AcidBubbles, is truly a remarkable developer."
Each segment had the same morph targets and identical keyframes.
When I played the animations, I compared the CPU load between having one segment and having two.
The difference in CPU bottleneck was quite noticeable.
This is especially problematic in interaction-heavy animations between characters with complex physics, since most processing has to be done on a single thread, leading to thread overload.
Although I don’t have in-depth knowledge of the code, after spending three hours investigating the issue with ChatGPT, here’s the hypothesis we came up with: CPU resources are consumed not only by the keyframes of the currently playing animation, but by all keyframes present in the timeline being called every frame.
I’m not exactly sure what that means technically, but if only the keyframes of the currently playing segment or animation clip consumed CPU resources, I believe it could greatly reduce CPU bottlenecks regardless of timeline size.
Of course, I wonder if there was a reason it was designed that way.
I tried to optimize the Embody plugin through ChatGPT by creating a lightweight version and adjusting the update interval in hopes of improving performance—but in the end, the original plugin itself was already extremely well optimized.
Even without understanding the code, I could tell—at least indirectly—that its creator, AcidBubbles, is truly a remarkable developer."
