I know motion matching is still experimental at this stage because it relies upon the trajectory plug-in which is still listed as experimental, but I was wondering if someone could maybe assist me in debugging the issue I’ve encountered.
I’ve reviewed several YouTube videos on the subject of motion matching by using the Lyra animations - I’ve seen that most of these videos used the 5.3 release of Lyra starter project, where as I’ve used the 5.4 release.
With implementations for all their settings, their animations are very smooth, and it seems there’s no issue with continual input.
However, in my case, the animation constantly interrupts the movement input, causing the character model to frequently stop and stutter. I’ve deleted all animations, pose search schemas, normalization sets and databases, and re-migrated them and set them all up again - and I’m getting the same issue.
To ensure that the correct animations are being imported for the correct states (i.e. crouching, walking, running), I’ve set up a Chooser Table using the ‘Beta’ plug-in for this.
If I swap my animation blueprint to another vanilla blueprint (i.e. the post process one), the character movement input is fine again and I can ‘run’ around in the idle animation state. If I swap back to my custom animation blueprint, the stutter returns - I’m not doing anything particularly special or taxing though:
The bit at the bottom is just to blend in the movement of the head when entering “look” mode. It doesn’t affect the input. This can be entirely bypassed and just have the post history added to the output pose and the input stutter still occurs.
Does anyone have any Pose Search Schema details on what might be required to get a better trajectory response as the only thing I can think of here is that the animation schema isn’t able to calculate what the best animation is to use based on the schema data, and therefore it just throws out the idle animation state and that roots me to the spot?