Learn hybrid animation in unreal engine, combining traditional keyframed animation with procedural animation techniques. Check it out: udemy.com/course/hybridanima…

I had an idea for something like this where it would be like motionmatching, but more pose-matching, where the position/velocities of the feet are arranged dimensionally (I guess 4x 2d) for all frames of all animations you feed it. Then whenever a foot is grounded you'd know it's next move in the "space" because you know what the ground has done in the last update. Then IKing for correction (which is fed for the position next time too so it keeps choosing the real best pose). And it'd have authority to ground/unground the feet (both feet in air it would be using just their velocities). Then I thought it'd need some velocity of the actual blendspace position so that it 'keeps going', but the poses wouldn't be arranged temporally at all so I'm not sure that'd work or do anything useful. I imagine it'd probably get stuck/weird at the point where both feet are off the ground. Or maybe would have to be 3d on the foot velocities stored, maybe positions too, so that the in-air arcs would find good next poses. Then I got a headache and thought motion matching probably works fine.

Not looking good to be honest

Hmmm might have found a way to use this procedural state transition logic in a way that works with animations, similar to motionmatching but better. Lot of awkward workflow stuff to figure out first but I think there's something to it

Realising the grapple is viewangle based was a gamechanger in empulse

New favourite mechanic discovered, paintbangin'

#empulse 1 minute 9 seconds 'upload' speedrun. 12 cap points, 0 deaths. Just deliverin the packages like a good little postman

Empulse speedrun challenge; "upload" mode, get all 11 caps. Starting the bidding at 2:18. Whiffed a lot towards the 2nd half

You sure?

Has anyone found X to be annoyingly playing videos lately? Have I turned on some autoplay with sound option?

Too fast for aim assist #empulse

A comparison of how Motion Rig handles your two example cases (A-pose character, upperbody scaled). The main things differing: - The swing-time (time the foot is in lifted phase) adjusts; this is actually an interesting artifact of having a large upperbody but not large legs where it'd cancel out the change. - The left/right swing of the chest, and again, interestingly affects the swing of the pelvis here from the disproportionate change. - There's some extra lean forward/back from the upperbody change but it's quite subtle (maybe too subtle, now that I'm seeing it) where it's partly influenced by the upperbody size (measured from spine01 to spine05). This is actually something I'll tweak/change though because that is primarily based on the legs. So no change to the legs means no extra bobbing on moving where there typically should be. Note: for comparison's sake the arms are kept static not procedurally animated, makes it easier to see the upper swing.

Is this idea viable/has it been attempted: Meta-procedural generation. Headline version: Procedural generation of a procedural generator. It figures out the "rules" based on you approving, correcting, or discarding it's attempt to match your criteria. Here's a rough idea of the process for a level generator: - The interface is like a split-screen game. Your side on the left, the systems output on the right. - You create a box on the left side - It generalises what you did, and presents a different box on the right side of the interface - If the box it generates does not 'fit your criteria' then you modify it or delete it. EG if it generates a box too thin, you stretch it out or simply discard it's attempt - You repeat this a few times, on successes or fails or corrections it will regenerate a follow-up attempt - It continually refines it's rules based on what you accept/discard/change. - You now have a box generator that generates boxes to your liking - You then add a corridor box, it attempts it, messes it up by overlapping it with box#1, you correct it or discard, etc. It gradually learns the rule that the corridor box should share an edge with box#1. - It converges on a ruleset which matches what you want to see over time. This could apply to any proc gen. EG object placement. You place a chair in a room, it attempts it but places it floating in the air, you move it down to the ground, and it loosely infers that "chairs are on the ground" and tests that criteria with you a few times. So basically in these examples it would be figuring out the parameter space for size, position, and relative positions of multiple things. Likewise could be generating a landscape or whatever else by the same sort of system. Implementation would be complex, but might be viable as a handy tool for making procedural generation in a more user-friendly/codeless way.

Been doing a lot with custom functions for Control Rig, gradually building up a library of useful stuff. There's so many little things that could be added that would have a billion applications. If anyone from Epic is interested I'll make a writeup of a few hundred little helper functions that it could really use. Ranging from "this saves adding loads of mess" to "this isn't possible without a simple code implementation".

🥊5 months into development Motion fluidity still not quit there yet, but at least head movement, hands movement and footwork are all procedurally generated by the system in real time according to player input, and they are all works together now, as you can see in the video below when control our guy with keyboard & mouse Next, will add physics component, so this guy can finally hit something #gamedev #indiegame #boxing

Short video testing something interesting from Unreal Engine 5.8 preview: youtube.com/watch?v=U7X3A82q…

A comment on my latest course promo vid, threw down the gauntlet, 2 hours pass .. then..