YT YouTube2Text

YouTube2Text → Transcripts → IK Rig: Procedural Pose Animation

IK Rig: Procedural Pose Animation — Transcript

by GDC Festival of Gaming · 10,329 words · 1,448 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00I am Alexander Marisnick
0:07of Ubisoft and today I'm going to talk
0:09about the tech I developed at ubisoft
0:11that I've named the IKEA I was asked to
0:14start with some info on the speaker and
0:16I did a little research I thirteen years
0:20in industry worked on a bunch of
0:21projects in a bunch of studios past
0:24three years I'm technical art director
0:26of Ubisoft Toronto now some of you have
0:28seen the original reveal talk from
0:30Vienna conference last year some of you
0:32have not so I will do a short recap
0:34before I proceed to the new cool stuff I
0:38started working at Ubisoft on the title
0:40that I can't disclose because we just
0:43didn't denounce it yet then I move to
0:45another title which is under NDA for now
0:47and to troll you guys even more I won't
0:51tell you about my two current titles and
0:53if you think that's a very boring folio
0:56just wait till the announcement days for
0:58these guys because it won't be as boring
1:00when you see what's behind the black
1:01boxes so for now we're going to have a
1:04friend dai give you hints you speculate
1:06we wait to help you speculate I just
1:09want to remind you what Ubisoft has
1:11recently shipped or announced and I was
1:14just to give you some itty-bitty ideas
1:16some disclaimers before we begin so I'm
1:19talking about the tech that's quite
1:20fresh I came up with this idea about a
1:22year ago and it was for one of the
1:26unannounced project and right now it's
1:27sort of a kind of forum spreading
1:29through several projects IKEA has passed
1:32proof of concept made it to run time
1:35implementation now now we're looking
1:38where else to apply it
1:39I'm not going to disclose anything
1:41that's secret which is like 90% of what
1:44we do and I will not mention any project
1:47names or details I can't show you the
1:49footage from unannounced stuff
1:51so today's examples are not related to
1:54specific titles I'm not here to announce
1:57one game I'm here to talk about the tech
1:59that's going to be working in a number
2:01of games so today we're talking generic
2:04and I prepared a bunch of examples where
2:07actual assets are hidden or replaced by
2:09placeholders so today we talk about the
2:12principles used
2:13effects on design effects on the
2:15pipeline and we talked about how you
2:18guys can try this at home
2:19disclaimer number two you will see a
2:22number of examples which are
2:24intentionally exaggerated so for example
2:26when I say we can take human motion and
2:28turn it into octopus riding a bicycle I
2:30do it I show it it illustrates a certain
2:33solution doesn't mean that whenever you
2:35need to have a creature like this moving
2:37you have to use the IKEA no you're
2:40welcome to go on and mocap a t-rex doing
2:42just that that's up to you I use these
2:44examples to talk about possible use
2:46cases to demonstrate possible
2:49applications do's and don'ts don't sew
2:52rabbits will run and octopuses will ride
2:55what is the IKEA IKEA rig is a systemic
3:00animation approach it's a principle
3:02attack developed to let us do more with
3:05animations and first thing is unbinding
3:08animation from rigs so a unified
3:13language sort of a thing a system that
3:15will bring all animation data of similar
3:18characters to one shared form
3:21secondly we move away from animation in
3:23keyframes we start animating their
3:25animations and we control the way our
3:28characters behave at runtime in the
3:30engine we create behavioral rules that
3:33can be mixed together to produce a wide
3:35variety of animations all in all it's
3:38retargeting 2.0 done on the fly with
3:40direct and immediate reaction to changes
3:44in environment character stats whatever
3:46happens at this moment in the game this
3:49is the most important slide incidentally
3:51it's the most boring slide you guys will
3:53want me to rewind to this couple minutes
3:55from now I promise you so IP rig is the
3:58text solution for converting the
3:59animation of any rig into a satify to
4:03change application of context-aware
4:06modifications to those jeans and
4:07implication of the result to any other
4:10rig where the runtime are offline
4:12so you have your source your motion
4:14capture rig you map it to the ikea
4:16unified structure using is mapping all
4:18the animation can be converted without
4:21any loss of data within the idea we can
4:24CC to modify the parameters of what
4:25happens with limbs and character
4:27as a whole we create sets of rules to
4:30explain what to do at any particular
4:32situation and then we can mix and match
4:35we can apply these changes on top of
4:37other changes so for example if one of
4:39the rules teaches the character walking
4:41character to crouch and other rule
4:43teaches the character to aim a gun
4:45third rule transforms male walk into
4:47female walk the combination will take
4:49male walk into a crouching aim in female
4:53this is actually elementary rocket
4:55science and I love it not because it's
4:57complicated but I love it because it's
4:58simple let's take a look at something
5:01simple the leg so our usual leg made of
5:04three bones plus all the small secondary
5:06things some of the problems we face here
5:09are first of all secondary animations
5:11for muscles and stuff like this if we
5:13bake them they will not really be
5:15working at the game during the runtime
5:17if we have a rag doll for example
5:19secondly we have a bunch of transforms
5:21but they have little meaning we can
5:23blend them but those position rotation
5:26of each of the bones in the chain don't
5:27really tell us much about what's going
5:29on finally most importantly our rigs and
5:32animations are bound together so you
5:35can't change the rig because the number
5:37of bones their names their hierarchy
5:39their sizes they must be content for
5:42this animation to play properly if you
5:45really want to change the radio can of
5:47course we target andrey import this way
5:49the update involves the offline loop a
5:52plus you need to maintain consistency
5:54between all results of all of your
5:56targets simple fact is we just need a
5:58couple of things to move this wag
6:00assuming it's planar we need the start
6:02position when in the target position
6:04rotation and we need the direction of
6:06the knee you understand these can be
6:09easily received from mocap you just need
6:12to map bones to nodes that's it
6:14that's basic retargeting extra bones
6:17such as twists they don't have to be
6:18animated are fine when we use the bones
6:22which have soft skin geometry geometry
6:24Susskind to them and we can animate them
6:26using constraints as we play so the
6:29accurate definition is mapping of bone
6:32rig to the IKEA directly in the engine
6:35once it's mapped all the animations be
6:38throw to this bone rig will be playable
6:40on
6:41the rigs which have been mapped to the
6:43unified structure so in this example the
6:45blue leg is the default config
6:47everything else is running exactly the
6:49same animation but on different fix this
6:52is possible because each of those rigs
6:54have been has been mapped what
6:57constitutes a casual rig so we have
6:59active bones for body data hands feet we
7:01have twist bones we have small bones
7:03four fingers and face secondary
7:05constraint muscle bones different nodes
7:07pivots for root capsule proper - meant
7:11and so on
7:12in case of IKEA rib bones can be the
7:14same but we're not working with bones
7:16after we have mapped them okay we're
7:18working with their IKEA nodes so there
7:22are two chains IP chains for hands two
7:24chains four legs one for spine one for
7:26head that's it
7:28so what what we do is we create one IKEA
7:31rig per character type as in biped
7:35quadruped add bird fish we create mesh
7:39with skin two bones just like we would
7:40normally do but now we don't fit the
7:43model to the rig as we often do instead
7:46we fit to reach each rib to each model
7:49so the look and proportions of each
7:50character are driven by the concept and
7:53by the character artist and not by
7:56shared rig when we export import in the
7:58engine we have the accurate definition
8:00tool which we use we map the bones to
8:02the full body a key to hike a chains
8:05everything else all the small bones are
8:07constrained and ran directly you can the
8:09engine at runtime all the extra bones
8:12they don't have to be shared we can have
8:14different bone setups for different
8:15characters we can add or remove them as
8:18we please and we don't need to rebait
8:21any data let me just briefly go over the
8:26most common constraints because they
8:28will come in handy later on first of all
8:30we have the position constraint which is
8:31super useful when there's multiple
8:32parents for example cowboy one position
8:35parent can be his moving hands from the
8:37animation another one is his holster
8:39now if I constrain something between
8:41them I can at any moment move the hand
8:43to the holster position orientation
8:45constrain much the same way if I have I
8:48want to add a tail I can take throat
8:50Asians of upper thighs and the pelvis to
8:52rotate the starting part of the tail
8:54so far so good look at of course what we
8:56usually do to aim our guns but I also
8:59like to use it to aim the head in a
9:01conversation and again we need to do
9:03this at runtime we don't need to bake
9:05this spring and dampen for with all
9:08flapping pieces we are using it right
9:10now as isolated example I like to use
9:13this within my character add this
9:15secondary motion tremblin curve to throw
9:18in arbitrary curves that we want or
9:20maybe take a sign of time and driver
9:22driven so for example one transform of
9:24one bone effects different type of
9:26transform of different bones I want to
9:29note that for most cases where few
9:31constraints are actually required so
9:33over constrained case like this there's
9:35only 5% of what we see animated is the
9:39actual bones everything else is
9:40constrained we don't usually need this
9:41in our games such cases would really
9:45never happen but it's good to keep in
9:48mind we can LOD up and down as much as
9:49we want and again we don't need to
9:51re-import we export anything if you want
9:53to add new stuff move into the character
9:56but we want to do more than simply share
9:59the animations we want to change them so
10:02to the naked eye it may appear that it's
10:05the same animation and it is the key
10:08difference here is that the left
10:09character is running 35 bones and the
10:12right character is run in the same
10:14animation but it's the IKEA rig format
10:18they're not different they should not be
10:21different by default without any changes
10:23that we've introduced we should have
10:25exactly full quality that animator
10:27expected to have that that was invested
10:29in the start in there now now we've
10:32defined this rig as Ikea right so we can
10:35start changing in old behaviors as they
10:37play the animations the income income
10:40animations can be received from just the
10:43key frames in your 3d software in alpha
10:45in case or which is more fun we can take
10:47the result that the blend machine spits
10:49out and modify that right as we play so
10:53now we have this accurate definition we
10:55can have some fun with it but yeah our
11:01original base animation is always there
11:03so any change we have is just a rule
11:05that I threw in I explained that I want
11:07different
11:07havior from start and nose of the thing
11:10and by controlling the behavior of these
11:13targets for hands spine feet I can do
11:17whatever I want I can modify the
11:18existing style of motion I can create a
11:21new style for specific case I have
11:23control over footsteps over hand arcs
11:27more importantly yes we can throw in new
11:30nodes and start animating to those nodes
11:33to create some new behavior I call such
11:36sets of modifications the IKEA rules
11:38there may be a rule teaching a character
11:41to Crouch may be a rule teaching him to
11:43aim a gun may be a rule teaching him to
11:45cover his head whatever else the cool
11:48part is when we throw in especially with
11:50the prop works when we throw in the
11:53weight of the prop and based on that we
11:55change the way that the prop is being
11:56carried by the character or we can also
11:58factor in things like character strength
12:01or tiredness one of the goals are set
12:07for the IKEA initially was to have one
12:10rig per one character so you have 50
12:13characters and again we have 50 rigs in
12:14again but keep it simple simple share
12:17their animations and allow for
12:19modifications which are easy to manage
12:21some character in case of character
12:23class rules I just create a rule that's
12:26indented for this specific character
12:28forever so a hero character will always
12:31give his shoulder a little backwards
12:33female for example has elbows bent
12:35inwards and a little more wiggling hips
12:37zombie case in here extremely been spine
12:39I could add the tremble if I want it and
12:41it's gonna apply on whatever this
12:43character plays and I can mix and match
12:47them I can have a zombie hero or I don't
12:50know
12:50rooty male if I want to prop operations
12:54doing it with weight we you factor in
12:56the weight of the prop for a different
12:58rule
12:59it looks like the artists prefer to have
13:01true boundary cases to create two
13:03boundary rules supervised super heavy
13:05and then work between them or possibly
13:07even have a middle case so depending on
13:12the weight of the prop change we would
13:14be somewhere in between at these states
13:17we can also as a set factor in how tired
13:20the character is we
13:21factor in his strengths so big guys will
13:24have easier time carrying rocket
13:25launchers and your grandma will not we
13:29can use the same approach for
13:30interaction with static props like
13:31doorknobs for example characters of
13:34different structure in size will still
13:36reach for the same place in the world to
13:38turn the doorknob mentioned we have a
13:40student game with multiple weapons
13:42conventional solution we trade several
13:45animation sets for different weapons so
13:48we have 50 weapons but we know every
13:50weapon we need to run crouch reload
13:52shoot too many things to mocap so we say
13:56our ek and our m16 will have the same
13:59placement of parts so we can use the
14:01same animation to play them now adding a
14:03weapon into the game would become
14:05complicated thing because now we need to
14:07create carrying this weapon with the
14:10Crouch without Crouch aiming shooting
14:12whatever in conventional setup we would
14:16also have to propagate every change that
14:18we want to do to every rig that we're
14:20using because maybe we have male and
14:22female rig and now we change the way the
14:25a.k behaves we need to modify it on both
14:27rigs and IKEA setting though every
14:30weapon preparation I carry grew
14:34simple as that so I can just go in and
14:36modify the rule to create a new type of
14:38behaviour with this weapon and aiming is
14:42separate from the motion itself and this
14:47will propagate to all the commotions and
14:49basically to all animations of the
14:50character if I want to at the same time
14:53grandmother and me will carry the rocket
14:56launcher in a different way because
14:57their prop weight is factored in should
15:01we want to add a new weapon we will make
15:03a new rule for it propagate is set to
15:05all characters if we want to
15:06should we have several aiming modes
15:09create a new aiming mode for hero for
15:11exam before female it will work the same
15:13thing one more new thing here is
15:16accurate rule for collision avoidance I
15:18like to have the hard collisions as we
15:20do but I also like to have the soft
15:22origins because we humans are smart
15:24things we apply the contradictory force
15:26before we hit ourselves with something
15:29so that creates a more more realistic
15:32and believable motion for proper
15:35operations like this so speaking about
15:37proper interaction now that the
15:40characters touch the props using the I K
15:42which we have full control over we can
15:45start scaling the characters and scaling
15:47animations and now you're thinking Alex
15:50we know how to scale the characters and
15:52we know how to scale the animations
15:53that's not the thing yes we do we do and
15:56every time we uniform scale the rig we
15:59also scale the animations we end up with
16:01lots of issues speed of character
16:04changes for once and this is not what we
16:06always want to illustrate this better I
16:09made some fan art for Star Wars just to
16:12show you so it's the same animation for
16:15both of these guys but rustic speed
16:19difference happens because of drastic
16:20scale difference in extreme case like
16:23this that's okay but more often than not
16:25speed must be dictated by the design not
16:29by the conversion result usually to fix
16:32that we'd start scaling the animation
16:33timing up and down but if we do it more
16:37than 10 15 % speeding up or slowing down
16:39the animation we start losing a data
16:41about velocity about the mass of the
16:43object and it's just not too good so
16:45here's the fun trick in Ikea I scale
16:48characters independently from the
16:50animations so animation is the same for
16:54these guys but the rebbe goes faster why
16:57well what's done here is the scaling
16:59just of the footsteps notice that all
17:02the animations are still in complete
17:04sync however I control the amount of
17:07footstep scaling and now that I have
17:10this complete control of this speed I
17:12don't need to adjust animation speed or
17:15timing I don't need to introduce the
17:17food sliding or do any other wicked
17:19thing and degree to which I can speed up
17:22or slow down is immense
17:23there is 400% difference between the
17:26slowest and the fastest guy and to
17:29illustrate this magnitude better it's
17:32pretty much that were we to scale the
17:35character up or down we wouldn't
17:36necessarily want to scale the speed of
17:38his motion we can just keep it we can
17:41set it as much as we want so the change
17:43there happens reverse we write the
17:45characters were scaled animation
17:46remained the same footsteps still
17:48have scale of one and I want to
17:51emphasize them the importance of using
17:54different skills yep
17:58and yep and why we know why we thing is
18:05that our games usually just have one
18:07character really I mean we can do
18:09different texture variation thing is we
18:11can push the geometry and the models but
18:13it's the same rig and yeah that's that's
18:17gonna look this way even like it's
18:21funniest when we share the rig between
18:23male and female like barbarian and
18:25Amazon but okay let's say we we added
18:27the female rig now barbarian shares the
18:29rig with a necromancer Amazon shares the
18:32rig with her grandmother it's not good
18:34for any of those characters and adding
18:37more exists of course expensive
18:38thousands of mocap takes for several
18:40actors feeding all of this into a
18:43monstrous blend three controlling is
18:46somehow discovery now that adding an
18:48umbrella to the game is a huge overkill
18:51because you need to more careful of this
18:52and it's it's not something we want it's
18:59something that we just endure and it's
19:01easier for the guys in the movie
19:03industry because they can have any scale
19:04they want for their Shrek or for the
19:06earth
19:07how can Spider Man and Captain America
19:09for us it's hard so what we try to do is
19:12we try to frame the guys so they look
19:14different even though we can't scale and
19:16we can try to fake it we basically try
19:18to make them look something like this
19:21we're not building star war games in
19:23Ubisoft okay it's I just made it for you
19:26guys just disclaimer and I scaled them
19:29all to their actual whoops PDR height
19:30and I've compensated the scale in
19:33animation so they still move in perfect
19:35sync and we can speed them up and we can
19:39slow them down we can give them props
19:40and they will look fun and the moment
19:44you set the scale is different like this
19:46something magical happens something that
19:49you can't really unsee
19:51because going back now is very very
19:55noticeable and okay so we know how to do
19:59the scaling now we know what to do with
20:01it next
20:02and we can keep still keep the speed of
20:04the changed speed of the scale or change
20:07scale if we want but we just don't have
20:08to and just by the way these guys as
20:11every character that I showed so far are
20:14still playing the same incoming mail
20:16walking animation uniform scaling is
20:20actually boring it's not fun I come on
20:22every every human has his own sample
20:24proportion differences here we have the
20:27base dude he's feeling just fine but I
20:29swap his right his left hand and right
20:32leg too much longer bigger versions now
20:36I I call this guy Quasimodo for easy
20:38reference this is the base layer of the
20:40IKEA notice the omission is exactly the
20:42same right it's not enough we need to
20:45explain them how to move so I make a
20:48list of Commandments for Quasimodo I
20:50tell them you know if your hand is heavy
20:53you should win if your center of mass
20:55has shifted compensate for it with your
20:58spine sorry
20:59and if the camp gets heavy don't wiggle
21:03it around drag it and the fun part is
21:06that the scaling sorry just really warm
21:09in here this scaling can happen at
21:11runtime and that's super good news for
21:13all the monsters and mutants and robots
21:16because now that the quasi model knows
21:18how to work with normal hand and super
21:20big hand he can do it with any size of
21:23the hand and just because we told them
21:26how to do it
21:27detour a little so I often say I tell
21:31the characters what to do and they do it
21:33and that's a little misleading because
21:34people visualize the process like this I
21:36told them to do this and that and they
21:38do in fact it's more of this and this
21:42because I control the action by creating
21:45those rules by plugging them together
21:47but it's more of playing with the regul
21:51kit without actual schematics so it's
21:53fun job still job and of Deeter so yes
21:58everybody is different and since we
22:00already captured emotions we inherently
22:03capture the motion styles and I'm
22:06talking just about the mocap right now
22:07so this is the prototype feature but
22:10it's worth mentioning nonetheless I call
22:12it motion GNA and the idea is that we
22:15just
22:15dress a guy up in a suit we let them
22:18move around the mocap studio for a while
22:20and we're not capturing the exact loops
22:23of what he does know what we capture is
22:25instead how does he carry his spine what
22:28is the width of his footsteps
22:32how much hand swinging is there and
22:34these things we can parameterize and
22:37there's about 20 parameters that so far
22:40I came up with and if I take those
22:43things I can store them as accurate rule
22:46and I can apply it to the animation of
22:49completely different guy
22:50it's not going to make them exactly
22:52believable but it's gonna be way closer
22:54to this after who just moved around for
22:56a couple of minutes in my mocap studio
22:59so just oversell it a little maybe some
23:02time in a few sure you walk between
23:04before your Kinect and we pick up how
23:06you do this and then apply to Cotulla
23:09dancing on the screen even if you just
23:11have the controller it looks plausible
23:15so let's recap for a minute we can take
23:18one animation we can change its style we
23:20can scale our characters we can scale or
23:22animations we can change the proportions
23:24on per character basis while we still
23:26interact with props consistently and we
23:29can even control the speed long story
23:31short we know how to turn our walking
23:33mail into crouching aim in female what's
23:39missing Oh clicks so what is missing
23:45from all of this
23:47then even terrain navigation of course
23:48but before we go into unreal Tori
23:51navigation there's this thing that's
23:53something we dance for so long we don't
23:56really notice it anymore and I love to
23:59kick this thing and I will kick it again
24:00now what I mean is the term locomotion
24:04modes so we do so many things weapons
24:06aiming reloading shooting opening doors
24:09that will limit the type of work
24:10emotions will let the characters walk or
24:13run sometimes jog and there are
24:15brilliant people trying to blend those
24:16states together we let the characters
24:18walk or crouch sometimes crawl and there
24:21are other brilliant people trying to
24:23blend these things together but the
24:25sophisticated blends are just masking
24:27the problem but there
24:28not solving the problem the problem is
24:29our navigation States in games are
24:32integers and in life they are not
24:34integers and this is inheritance from
24:37the idea that we will want to do
24:39different things as we navigate
24:41different ways so we can't afford to
24:43mocap everything in all the interim
24:45States so what's wrong with state a or
24:47stage B switch but honestly there is
24:51nothing wrong with a switch it works for
24:53past 30 years and we keep doing
24:56this Nets
24:58what's worse is we build our levels with
25:00this in mind your doors must be fixed
25:03with or the character won't walk through
25:04them plausibly your obstacles must be
25:07fixed height or he won't jump over them
25:09plausibly the ceiling must be fixed
25:11height because under otherwise you want
25:14to crawl under them plausibly and every
25:17now and then there is those little just
25:19missing inches that the animation
25:20suggests a different height but we just
25:23have to stay so that's it what if this
25:26collision is only halfway down what if
25:28it's moving what if it's a little lower
25:30or higher that's character remember what
25:32you just did does he have a history is
25:35he limited to these states if we start
25:38retracing forward not in time but just
25:40in space we can start predicting what
25:43happens and we can change just the right
25:46amount we don't have to go all the way
25:48down if we just need to know their head
25:50but if we have to we will do and we will
25:53our motion will suggest the physical
25:56appearance of the world around us and
25:58it's going to make more sense to us
26:00humans we don't collide by the by our
26:03own choice so we don't clip like this we
26:06avoid were smart beings and it's not
26:09hard to mimic if we strap the idea of
26:12integer States and it's not just
26:15integers that are obsolete the whole
26:16concept of metrics is insulting and it's
26:19ridiculous and they know the questions
26:22that we get at this moment metrics are
26:24at the core of our pipelines without
26:26metrics how do we even design or model
26:28how do we build levels how do we build
26:31the game with arbitrary scales like like
26:33the real one like the real world we it's
26:37scary we can't pull it off we're not
26:39going to risk it
26:40and first of all let me call these old
26:43ways what they are limitations and you
26:46know a guy breaks a leg spends couple
26:49weeks in the hospital then he has this
26:52casket for several months then the
26:53casket is gone his wreck is healed he
26:55still keeps limping because he got used
26:58to that and we got used to that maybe we
27:02don't have to do that
27:04you see I thrown in some uneven training
27:06here just for kicks so without further
27:08ado let's take a look at different ways
27:11to navigate and even common solution
27:13rates raised down and be happy well the
27:15sad part is there is no history in this
27:18walking down walking up looks the same
27:20and of course we can't have stairs in
27:23here because if we just ship this height
27:27between two frames too drastically we're
27:29gonna have the food popping we can't use
27:31the stairs ok ok and again if we if we
27:35can retrace forward right cast forward
27:38we can start anticipating say hi to
27:40anticipation say hi to follow through
27:42and we can propagate this change to the
27:45whole body not just the legs because
27:46we're not splitting between upper body
27:48lower body anymore we can act and it's
27:51fun and our slope can be arbitrary
27:54we were not picking animations from the
27:56pool and blending between them we
27:58generate the proper motion for this
28:00specific case for the specific guy with
28:02a specific scale no loads railing no
28:05feet flipping no metrics what do we do
28:08here after we fire the level designer
28:11normally we we just say ok it's going to
28:14be a flat plane and we're gonna clip and
28:15that's fun because it's it's a video
28:17game okay cut us some slack that's fine
28:19yes no like a solution can handle stuff
28:22like this usually usually we fake it and
28:27be happy but again you know what we can
28:30do instead is we can switch into
28:31nostradamus mode or what I call
28:33nostradamus mode basically what I do
28:34with we throw in the animation into the
28:37engine and then we analyze our feet and
28:39we just store couple floats per frame
28:41for each foot we remember how long ago
28:44did I get lifted up
28:46how soon am I am going to be planted
28:48down where will I be doing this so now
28:53per frame I know that the food for
28:55example initiated the lift up thing and
28:59since I know where it's gonna land and
29:02when I can modify the whole curve and
29:05arrive to the proper location yes some
29:09weight of the body is lost in here that
29:11will be fixed on the Polish but at least
29:13I planned correctly in fact I'd slow
29:17down the extra here by twice because
29:18human wouldn't really risk that fast
29:20moving that fast but that's fine
29:23in famous case of discontinuous terrain
29:25stairs I love it
29:26so okay if we have stairs in our game
29:28all the stairs gonna be the same because
29:30you guys know will you be more capped
29:32for these specific stairs going up going
29:34down we can't have stairs of different
29:36right what happens if he walks
29:38diagonally because any mission is not
29:40going to play anymore what happens if he
29:44crouches the Wemo kept crouched up
29:47crouched down and maybe in diagonals
29:49what happens if you wanted to limp what
29:51happens if I have a female rigged now I
29:53need to propagate all of this so again
29:56conventionally I would guess root
29:58sliding and some foot clipping and it's
30:01fine
30:01but remember we have this Nostradamus
30:03mode right so we're tracing the
30:07footsteps and when the food gets lifted
30:09up we can figure out where it's going to
30:11land and we get we can figure out if we
30:13want to land there or maybe a little
30:14different place so now we can have
30:16arbitrary steps staircases I mean and we
30:21can have arbitrary scale of the
30:22character and he will still walk I mean
30:24it's just depending on his height he
30:26will be missing stairs the steps or
30:29maybe he will step on each of them just
30:30like we do because we have different
30:33scales and that's how we behave and
30:37since we generate this animation on the
30:40fly I mean we can have any type of steps
30:42that we want for any character scale
30:45that we want yes we need to polish this
30:47but this is already better than what we
30:49normally have with just flipping through
30:52the stuff final test of this continuous
30:54and even Turing conventionally we would
30:57have a minigame for this because you
31:00can't really move cap unless you mock up
31:02for just this specific set of obstacles
31:04and just way this one animation
31:06but again I mean if if in real life a
31:09five-year-old would have fun doing this
31:10then our hero should be able to walk
31:13through this as well just by pressing
31:15the button without no minigames so I
31:18know it's a lot to take him let me show
31:20you the combined result of these rules
31:22please welcome the marching band this
31:25operates directly from the guy with the
31:27controller all right then we added Ikea
31:32we have different proportions in scale
31:34remember that was another guy right even
31:36though they're all skilled they work in
31:38perfect sync remember we have the
31:40footstep scaling remember our IKEA rule
31:43for crouching notice that the tall guys
31:45and short guys crouch differently
31:47because they need to go to different
31:48height as a result I mean some of the
31:51male some of them female because it was
31:53easy and a brilliant single one of them
31:56carries his or her own musical
31:58instrument and plays it to some tune
32:01that I will expose weights wrong and one
32:07more fun part there is this waving brown
32:10demeanor you see that we're gonna go
32:12there because they're the water you like
32:15I call it the cheese's test
32:17it's the ultimate thing I mean if I can
32:19work on this I can work on anything plus
32:22an asteroid I must move for the steps so
32:25this is one source animation right
32:28everything else generated per frame
32:30right here controlled separately this
32:36nice visualization for the curves that I
32:38threw in for them to play so each of
32:40those characters uses one of these
32:42curves to play his or her musical
32:44instrument so there is the drum the
32:46dishes the tambourine the violin and you
32:51know what fräulein oh I could just throw
32:52in actual musical piece and take the
32:55curves from different musical
32:56instruments and freedom here and see the
32:58result not sure it would be super cool
33:00from the start but I can try and yeah i
33:04scaled them all differently because well
33:06because I could human height usually
33:10like 1.7 meters male 1.6 female plus
33:13minus 10 15 % wrote different and I
33:16choose to respect that because it's very
33:18small and cheap wait
33:19add realism to this so marching band
33:25marching band is one single actor turned
33:28into nine characters some males and
33:31female nine different scales and
33:33proportions nine animated props synced
33:36to the tune controlled separately they
33:39can Crouch they can climb time to make
33:42all of this I had the rules ready
33:44remember one day took me several months
33:48to realize one more thing so crouching
33:50is actually just a subset of a bigger
33:52thing of avoiding and if we can avoid we
33:56can attract as well so we start ray
33:59casting in two sides we can have a lot
34:02of fun with character just walking
34:03around simply by the level and coolest
34:08part is I've created these rules for
34:10avoidance and for traction and I can
34:13apply them so my Crouch in zombie a mean
34:15female carrying whatever and I can tell
34:19the different characters will do
34:21different things maybe somebody will
34:22reach further maybe somebody will be
34:24recasting into other characters and
34:26touching other characters maybe somebody
34:28will be touching him as he goes on I
34:30don't know so let's have some more fun
34:36the big section the quadrupeds first
34:39thing of course I have the 4c I turn a
34:42horse into IKEA rig for C because I
34:44can't work with bones so once I turn the
34:4814 to Ikea rig for C I just have those
34:50nodes now I can play with it but the
34:52thing is there's less to do with
34:53quadrupeds conceivably they don't
34:55operate with props but what they do
35:00operate with is other characters so I
35:03put an Ikea rig and an Ikea and we can
35:06tell this rider what to do we can give
35:09them props and so on you know as much as
35:11we want as for the horsey we don't
35:13usually distinguish between with or
35:15without ride your animation we never
35:18really mocap for with rider without
35:21rider it doesn't seem like we need to
35:24question of course like would there be a
35:26need to modify the horse animation in
35:29here well does the player need visual
35:31feedback that the horsey
35:33is that he's overloading the right right
35:36maybe there is an extreme case where we
35:38want to tell them that this horse is not
35:40a happy horse and it's nice to be able
35:44to do this fun stuff on the fly of
35:45course I exaggerated but you know what I
35:47mean guys and if you don't have the
35:50horsey with a heavy rider mocap you're
35:54still fine in fact I didn't have horsey
35:57mocap at all and I'm still fine because
35:59I had Jerry and Jerry has been my
36:02champion for the past year so you know
36:07if you're still feeling bad for the
36:08green horse II don't because none of
36:11them exist but Jerry does and through my
36:14test Jerry has been several breeds of
36:17dogs capped mountain lion a bear horse
36:21elephant and even more bizarre things
36:23like this so in all honesty he deserves
36:26an Oscar and yes I was not here I scaled
36:29this animation made it slower made sense
36:32in this case so at least as far as the
36:35locomotions go yes I can share but then
36:38certain extent of course the elephant
36:42would still piss with his leg lifted we
36:44wouldn't do that but happily we have
36:47talented animators and of course all of
36:50this is only possible because we can
36:51scale the footsteps quadrupeds are super
36:54different in our proportions as we go
36:56from breed to breed what else is good
36:59about IKEA rigged for quadrupeds I guess
37:01yes maybe this maybe didn't even terrain
37:05for horse for the rider slope speed any
37:08direction as we please not only that we
37:12remember we did this Nostradamus thing
37:14right for foot placement you probably
37:18know it's coming
37:19another thing that you will not and see
37:22at least I didn't see it anywhere and as
37:26I said Jerry is amazing let me talk
37:31about main reasons to introduce the
37:33systemic animation solution to the
37:36pipeline which I carried or any other
37:38one first of all it's about character
37:41design character design that is driven
37:44by the character artist and concept
37:47not by the rig that we share and we want
37:50to add versatility to sizes we want to
37:54add versatility to proportions of our
37:56characters we want to add different
37:59behaviors as they move around to show
38:01that they move differently and we want
38:04to provide visual clues whenever working
38:05with props to show it's heavy that his
38:07limping is hurt we can show it in the
38:10animation content on demand that means
38:15generating animations right as we play
38:18right as we need them so for this
38:21specific guy in this specific case in a
38:23specific level right now I need this and
38:26we make it so hence the term systemic
38:29animation that I like so much we don't
38:31pick from the pool of stuff we pre
38:33created we make it on the fine first
38:38iterations and more important fast
38:42prototypes anytime you want to add
38:45umbrellas to your game you just throw in
38:47a prop and you create a cure group now
38:49I'm not saying this is going to be the
38:51perfect umbrella carrying animation I'm
38:53saying you can do it within one day and
38:55see the result and play with it and
38:57figure out what else do you want you
39:00don't need to mocap shoot this off line
39:02a loop of retargeting for your female
39:04rigs carrying umbrellas and stuff like
39:06that import expert now right there right
39:09now fast and furious final reason I
39:13first called this saving money but I
39:17know you guys we get ten million dollar
39:20envelope we spent when you ask for more
39:22so whatever we save here we would still
39:26waste in some other place so I'm calling
39:28it cost allocation let's save some time
39:31we let's save some money let's invest
39:33this time and money into maybe
39:34generating more content for the game
39:36maybe investing it into the rendering
39:38maybe investing it into the sound and
39:40whatnot we will figure that out I know
39:45what are the pipeline changes so
39:49characters yes they're created based on
39:51concept art and common sense of course
39:53not on based on the rig
39:55basically we're cutting away a part of
39:57the pipeline where your curve
39:58just made this wonderful model and now
40:01technical animator comes and says now
40:04you have to adjust it to this rig and
40:05he's crying rigs can be unique but for
40:10the most human cases it's like you know
40:13with any bugs in your Mme image they
40:16have this screen at the start where they
40:17say that height weight which changes
40:21those are within centimeters they're
40:23super important for the fight and for
40:24the fighters but they are not too far
40:26away so normally I would say character
40:29artist can just modify the rig on his
40:31own if you trust him enough just to
40:32scale things up and down as he pleases
40:35and in engine riggers and tech guys map
40:38this incoming rig to the unified AI
40:40clerics structure and they define which
40:43parts of the bones we want to each to
40:46which chain in full body i key and yeah
40:50they also set up the constraints for
40:52secondary things like muscles and
40:54whatnot
40:56animators they work with the IKEA rig
40:58rules there are two very important
41:01things here first of all to work with
41:04this stuff you need the real-time
41:05preview so if you build something like
41:08this make sure your animator sees the
41:10changes and can experiment with them
41:12second of all there is in need and we're
41:15actually discussing it right now I need
41:17to connect the animation engine of your
41:20game engine with motion builder or any
41:23other software of choice for the
41:24animator so he can go back and forth and
41:27set the pose for this particular trumpet
41:30if he wants to that's the simplest thing
41:32and because he would be doing this a lot
41:35with trumpets there is a need to to have
41:38this process simple and fast but the
41:42largest largest difference of all is
41:44actually for creative directors for game
41:47designers because the holy moment
41:50happens when you realize you can spam
41:53these marching bands on the fly as much
41:55as you want that you can write different
41:58animals that you can Telegraph the
42:00characters state and health through the
42:02animation without this extra cost of no
42:05capital that you can have cover system
42:08destructible covers and characters of
42:10different scale will use those cover
42:12different ways they get destroyed that
42:15you can have some very interesting fun
42:17with AI and I will talk about this maybe
42:21next time
42:22and yeah that you can have characters of
42:26different sizes and have this
42:28versatility in your virtual city or
42:30whatnot and that all of these things can
42:35be connected to your game design because
42:38now you're showing way more maybe you
42:40want the player to be noticed in this
42:41for a particular reason when to
42:47introduce systemic animation solution to
42:49the pipeline as early as possible I
42:52would say if you had your first mocap
42:54shoot probably it is too late because
42:57there is a lot of things to consider
42:58first of all you need to decide what is
43:00it specifically that you want to do do
43:03you just want to have different
43:04proportions of humans or do you want to
43:07be writing quadrupeds or maybe you have
43:09a game with focus on proper interaction
43:12and you want this interaction to be
43:13realistic so you need to decide how much
43:16of it you need to do because there is no
43:18need to build the full shebang from the
43:20start we need to build something simple
43:22to see how it works also there are
43:24things in pipeline which will be cut out
43:26and a lot of this is about iTunes and a
43:30lot of this is about motion capture for
43:34every possible Crouch with every
43:35possible weapon because probably you
43:37don't really need this anymore you focus
43:39your animation on things that matter on
43:42cutscenes on the character of main
43:44characters but not on making endless
43:46variations for male and female rig for
43:49several different props yeah also you
43:53need to think about your levels because
43:55if you start discarding the metrics of
43:58the levels that changes things a little
44:01you can start thing and you start
44:03thinking about dynamics in their proper
44:05collisions with dynamics it's
44:07interesting but whenever you do this I
44:12suggest you guys do it soon because
44:15couple years from now systemic animation
44:18stuff will start shipping and it's going
44:21to be hard to keep up with this
44:24it's so good but I also want to be clear
44:28when not to use anything like this and
44:30first example is the
44:32obsessive-compulsive game mechanics like
44:33Mortal Kombat for example they have a
44:36number of characters yes but every
44:38single animation every single keyframe
44:40of every single animation has been
44:42visited in revisited million times
44:43because it's super important the reach
44:45of the sub zero's punch the speed of
44:48this punch this is the game mechanic you
44:50can't allow this to be controlled by
44:52some arbitrary system you want this to
44:54be specific so in this case you don't
44:56need any solution like that especially
44:58because in Mortal Kombat you're not
44:59really working with props or navigation
45:01and even terrain as much a massive
45:05strategy games conceivably I mean the
45:07games where you have 10,000 characters
45:08on the screen even if you had the
45:11fastest fullbodyik a solution ever if
45:15the character size is 10 pixels in the
45:17screen just there is no reason to go too
45:19deep with fine-tuning the foot placement
45:23of him so of course we want to keep it
45:26smart and solo games meaning games where
45:28you only have one character or the
45:30amount of character animation is low in
45:34such cases maybe you want to invest in
45:36some sort of systemic animation solution
45:38but only if only if that's the key part
45:42of your game and there are games where
45:43there's one character but dynamically
45:45interacting with environment it's very
45:47smart
45:47unless that's the case of course there
45:51is no need to invest in a humongous
45:53system if the total time of animated
45:56character on screen is 30 minutes so let
45:59me answer the three main questions you
46:02guys probably have what stage are we at
46:05with IKEA Alex nice videos how usable is
46:09it so I came up with this idea about a
46:11year ago as I said I spend some time
46:14bouncing it around in my head and with
46:17my friends at Ubisoft and back in the
46:20day we were only interested in about 10%
46:22of what I'm showing today because
46:25frankly we didn't know all of this as
46:27possible I did I built the first
46:30prototype in maxscript
46:32and that was a proof of concept stage we
46:36were playing with it we were seeing what
46:38can happen dogs two cats
46:40I remember the challenge from Christian
46:43and that was fun but then my friend
46:46Michael Butner took interest in it and
46:49he built the runtime part of it and he
46:51keeps building it as we speak
46:53like right now in Toronto and yeah we we
46:57keep developing it further to be honest
46:59the list of complete features is now 10
47:01percent of the new 10 times increase of
47:05the dream and I don't know how much do
47:07we need right now but we'll see and the
47:13key parts are there right now namely
47:16they're optimized for body and mechanics
47:19for creating and sharing Nike rules
47:21between different characters and seeing
47:23it in the engine main road yeah resource
47:28heavy expensive main road the expense
47:31goes on your full body a key solution
47:34all the IKEA rules they have very
47:37noticeable results right but in fact
47:40they're very short in simple lines of
47:42math this is what I liked a lot I mean
47:45it's simple mathematics
47:46they're independent of the game engine
47:48or the programming language it's very
47:51simple move rotate constrain you may
47:56think of the IKEA rules as motion
47:57shaders but with way less instructions
47:59in way were more output so after this
48:03math is done then a full body a key
48:05starts working and you can use any full
48:10body a cake you want and that is your
48:12cost the good part is that full body a
48:16cake can be optimized very far in case
48:19you wanted to and depends of course on a
48:22rig complexity so it can be L Odede
48:24things like great casts you understand
48:27how to control those so up to you
48:32finally working I downloaded the
48:37reaction I got a lot from the moment
48:40that I did the original IKEA reveal
48:42speech was this and
48:47let me ask you what do you download the
48:50normal Maps you don't you can introduce
48:55the feature to your engine yes or you
48:57can purchase an engine that has supports
49:00normal Maps but the reason I keep
49:02calling this is just one of the possible
49:04systemic animation solutions is that
49:07it's not a software piece it's a concept
49:10and it's in development that Ubisoft yes
49:14but you don't have to wait and wonder if
49:17we revisit and when if I if I did a good
49:21job explaining idea you know how to
49:25build it with blackjack and whatever you
49:27want in the recent years we had very
49:31many advancements in rendering in in AI
49:34an amount of content on the screen and
49:37in the game we got amazing evolution of
49:40dynamics we evolved the streaming of
49:43data and a bunch of other things so it
49:46feels like now is the time that the
49:49animation is taking this step to the
49:50next gen as well and there are amazing
49:53developments out there there's cool
49:56stuff there is human NK by Autodesk
49:58there's euphoria by natural motion final
50:00okay bye finally K by root motion
50:03runtime by a Kinema I like to think that
50:06if we share the ideas and developments
50:09we can contribute to the industry as a
50:12whole and that would be cool traded you
50:15this smiling person is Michael partner
50:18decklid of animation in physics of
50:20Ubisoft Toronto he's the mind the father
50:22of motion fields tack motion matching
50:25tech and he's also the father of runtime
50:28implementation of the IKEA rig
50:29so whatever runs now and whatever we'll
50:33keep running
50:34I would not be possible without my phone
50:37and of course my gratitude to Ubisoft
50:40because this is the the craziest family
50:43I had for employer ever and we are
50:46shipping kick-ass games and you know
50:48it's coming yes to ship today's games we
50:51give money to smart people so if you
50:53want to work on epic stuff and if the
50:56dev force is strong with you
50:58there's a team that could welcome you
51:00there is Ubisoft booth on the premises
51:02second floor wait right outside here go
51:06talk to them figure out maybe it's
51:07interesting for you maybe it's not but
51:09don't be shy just give it a try
51:10it's fun or catch me up I can give you a
51:13shortcut to them why not two things I
51:19want you guys to I want to suggest you
51:21guys to do
51:22first thing tomorrow 11:30 there's a
51:26talk about motion measuring by saman
51:28trivia go there and listen to him listen
51:31to what he has to say if you don't have
51:33time make time because this is one of
51:37the coolest things in animation I've
51:39seen so far and I've seen a lot second
51:42thing we will have an extended Q&A
51:44session tomorrow I'm going to keep this
51:45fight for a while so you can take a
51:47picture so you don't forget West's whole
51:49second floor tomorrow 3:00 to 4:00 p.m.
51:52come over we can chat we can have some
51:55laughs somebody may try to recruit you
51:57I'm sorry it's going to be fun so thank
52:02you that's it
52:11who wants to ask something guys
52:13microphone don't be shy questions
52:16questions what middleware products are
52:24you using to achieve this you mentioned
52:26a few for body IT solutions are you
52:29using one in particular no throne
52:31blackjack and first thanks for sharing
52:38the idea I think we need more of this in
52:41game development why do you like
52:43defensive videos thanks videos but idea
52:47is much more important what I want to
52:49ask how much time and effort does it
52:52take to configure one character with all
52:56the parameters that you mentioned like
52:58how it moves and which depends from
53:00which part okay so how much time does it
53:04take to basically define a motion
53:07capture akin to the psyche with chains
53:09is that what you're asking yeah like how
53:13much time does it you create a new
53:14character using already ready they mock
53:21up I would say like way less than a day
53:26there is there a couple of crazy
53:29characters like you saw in one of the
53:31videos this guy with multiple legs and
53:33hands taken out of him that took me more
53:34time because well most of this time I
53:36was trying to figure out where just take
53:38an extra hand but if we're talking about
53:40bipeds yeah that's that's fast and
53:43simple actually because yes I have a
53:45different bones coming in but I know
53:47where my legs are I might even have a
53:50naming convention and streamline if I
53:51could just script it okay and all those
53:55definitions do you define them in the
53:57code or you made some kind of UI which
53:59is usable by artists so right now it's
54:03not as friendly but of course we need
54:06the UI we're thinking an old based
54:07system and yeah at the moment it's lines
54:13of code with some nodes in place but
54:16we're not complete yet we're not
54:18shipping it tomorrow
54:20but yeah whatever we do we have to make
54:22sure that like people working with this
54:24some of them are nerdy technical artists
54:27yes but a lot of them are animators and
54:29animators need to have nice interface
54:31ideally resembling something that they
54:33have experienced working with before
54:35Kansai mention the MotionBuilder ugly
54:38strips okay thanks
54:40thank you this is very cool you showed
54:44some examples where you said that like
54:47this is good initially but you would
54:48want to polish it can you talk about how
54:50you would polish these sorts of things
54:52like if an animator looks at something
54:54and says oh that's really good when the
54:56character crouches but it should look a
54:57little bit more like this is there some
54:59key framing like way to tweak it and
55:02polish it so yes it should that that's
55:06why I was talking about the
55:07MotionBuilder connection right directly
55:09because animators like to tweak things
55:11but to be honest this thing you're not
55:16you you're not supposed to modify this
55:19particular pose if it's broken you need
55:22to figure out why because at the end of
55:24the day we're generating we're
55:26generating the content as you play you
55:28never know what's going to be happening
55:29so if all of a sudden your leg that
55:32overstretched that means you need to
55:35introduce a rule to keep the leg within
55:37the limits you know how to do this so
55:39ideally I would say you try to fix by
55:42that now if my enemy juror comes to me
55:44and says you know what this female works
55:46looks like crap my first intention is
55:49okay I need to open a video female walk
55:51female walk that he likes and I want to
55:55try to parameterize what's happening
55:57because when I was keyframe animating
56:01stuff I was doing pretty much the same
56:02thing right I mean in the mind of
56:04enemies right you open the reference and
56:06you see what happens now I'm not just
56:08going frame by frame adjusting to the
56:10background image I'm trying to figure
56:11out what exactly is happening maybe
56:13there is a curve maybe every time the
56:14leg is lifted there's an extra X
56:16rotation in local of the hips I'm gonna
56:20do that and see this hope that it works
56:23thank you thank you when you're doing
56:28your Nostradamus footfall determination
56:32and you're under control of let's say
56:34the player who's you know constantly
56:36changing with the joystick which way
56:38they're going are you
56:40[Music]
56:42recalculating the footfalls like as you
56:44are moving your foot in the arc or do
56:47you lay in that foot and then just do
56:48your calculation off of the next
56:50footfall and account for that so if the
56:53foot is like if the food is in here my
56:57incoming stuff just from the blend
56:59machine spit out it already has the idea
57:02of the position and I also have my
57:04Nostradamus nodes and they're telling me
57:06I'm like 50% in there now if I start
57:09moving to a different direction I have
57:12to modify my target vending position but
57:15again if I if the direction I'm moving
57:19in is controlled by the blend machine
57:20not by some artificial leaning these new
57:23animations will have this target
57:26position relative to them so as my
57:29animation changes my target position
57:31also is grabbed from the new animations
57:34that I just walked into am I making
57:36sense yes okay okay thank you I heard
57:43another talk from Ubisoft about motion
57:48matching it's let relate to I can be so
57:53these two texts are different I started
57:57working on this taking the input from
58:00motion fields motion matching there was
58:02a rename but it's the same thing and
58:04that's why I advertised Simone's talk
58:07tomorrow so much I think everybody
58:09should do this but these two text they
58:13do different things so motion matching
58:15is all about a natural human motion with
58:18very effective use of mocap time so you
58:22create the motion I can rig is about
58:25modification I can rig is not about
58:27gesture this is the animators the mocap
58:30I can rig is about modification of walk
58:33into crouch simplest example for
58:35carrying something right so these things
58:37they sort of work one after another and
58:40they contemplate they add up
58:45Thanks thank you your encoding of the
58:49foot animation looked like it was
58:52distanced from hip to ankle with the
58:53angle of the knee or something like that
58:55I was curious if you ran into an
58:57animation where let's say it's like a
58:58Dick Van Dyke jumping up the air and
59:00clicking his heels together type
59:02animation where the distance between the
59:03heels was critical does that require you
59:06to modify your animation rig so you now
59:08have an encoding of the distance between
59:10the heels and have you run into cases
59:12where that happens a lot based on
59:14different content going into the game
59:15and does it is that important to your
59:18pipeline being able to kind of change
59:19the definition on the fly or is this
59:21very specific to like basic locomotion
59:23so any type I will try to turn this into
59:28a bigger question and you'll let me know
59:29if I got it right anytime our character
59:32has self collision of any sort or a
59:35collision with a prop of any sort how do
59:38we compensate for it in this system
59:41right yes so first of all of course we
59:44have the physical representation of the
59:45character I mean we have the full body
59:47okay but we still keep our red doll and
59:49stuff we need this what I really like
59:53and what I didn't have the time to do at
59:56the moment but what I really want to do
59:58is if we mow cap the guy going over a
1:00:01box for example an example of the proper
1:00:04interaction right
1:00:05what I do is in a small cap file I place
1:00:09the actual box and I detect where there
1:00:11was a correct collision
1:00:13now if the box is scaled my animation
1:00:16knows that here's the place where I
1:00:18collide but this place is inside
1:00:21something I can find the new placement
1:00:23for the leg basically if the guy was
1:00:26scaled but the box remained the same if
1:00:28I know where it's gonna happen
1:00:30I can try to get it out it's not going
1:00:32to work out of the box
1:00:33instantly when the character self
1:00:36collides if I scale them just uniformly
1:00:40it's going to be the same you understand
1:00:42but if I do the proportion scale I of
1:00:45course need to be smart about what's
1:00:48happening I need to make sure that I
1:00:50don't let him cross the new bounding
1:00:52volume so yes there's physics in it but
1:00:55as I was talking about the soft
1:00:57collisions and hurtful
1:00:58I also like to slow down before before I
1:01:00hit
1:01:01so I'm actually I need to keep this in
1:01:04mind I don't have this feature like
1:01:07fully developed so I'm talking from the
1:01:08top of my head but I hope this idea
1:01:12makes sense
1:01:13thanks thank you stat all alright thank
1:01:19you so much have a good day

About this transcript

This page contains the full transcript of IK Rig: Procedural Pose Animation by GDC Festival of Gaming, generated from the public captions YouTube serves with the video. The transcript has 10,329 words across 1,448 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.