Coz_y_not Posted April 6 Share Posted April 6 If you take Crag test geometry. Go to your global animation options, uncheck integer frame values (have the step at 0.1). After Crag moves to grab the hammer,scrub the the timeline at every subframe (right arrow), be patient it's coming ... carry on, not yet, there, right there do you see it? geometry going crazy in between frames, not once many times, even the hammer takes the opportunity to flip around 360 degrees in between frames, coz why not? no one is looking, right? not really, the solvers can be a bit allergic to this "immature recalcitrance" in between frames. What I've tried to do is first to kill the subframe data that came with Crag animation by default, coz it's trashy. then attempted to recreate subframe data several ways for half a day (stealing the kitchenware from the deprecated timeblend sop), got some less than optimal results, then time to ask: how do I generate smooth subframe data? I mean as smooth as the original except some craziness in between frames? 1 Quote Link to comment Share on other sites More sharing options...
David Tom Posted April 7 Share Posted April 7 What you’re seeing is actually pretty common with rigged/constraint-driven animation. The “crazy” subframes usually come from how transforms are evaluated between frames, not from the geometry itself. A couple of practical things you can try: 1. Check where the motion is coming from (object vs SOP level) If the animation is happening at object level (bones, constraints, etc.), Houdini is interpolating transforms, not geometry. That’s where flips can happen. Try baking the motion down to point-level first (e.g. using a ROP Geometry or Object CHOP → Bake workflow). 2. Use TimeBlend + TimeWarp properly TimeBlend alone won’t fix bad interpolation, it just interpolates what it’s given. Make sure you have consistent topology Try increasing sampling (e.g. resample to higher FPS first, then retime) Sometimes switching interpolation mode (if available) helps reduce flipping 3. Check quaternion vs Euler interpolation Those 360 flips are classic Euler issues. If this is coming from rotations, try: converting to quaternions before interpolation or using CHOPs to filter/smooth rotation channels 4. Cache and stabilize Before doing any subframe work: cache the animation to disk then apply TimeBlend / retime on the cached geo This avoids re-evaluation quirks from rigs/solvers. 5. Accept that some “in-between” data is undefined In some setups (especially constraints or IK), subframes are not really “authored”, just interpolated. So unless you bake or resimulate at higher FPS, you’ll always get weirdness. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.