This approach is indeed feasible, essentially dividing the player’s movement into multiple modules. However, the granularity of dynamic loading is ABP (or Layer) rather than individual animations
With this method, it seems that basic locomotion remains resident in memory. For example, resources like start running, stopping, running, etc., are still loaded all at once rather than loaded when needed. Only some less commonly used features (like wall climbing) can be separated out
Previously, I used a self-developed animation system that would only load specific animation resources when needed. During asynchronous loading, it would maintain the pose from the previous frame to prevent T-Pose from appearing. Although there’s theoretically latency, it’s not noticeable in practice
It seems that UE itself doesn’t have asynchronous loading at the granularity of individual animation resources?