Old Man Willies Walks

[fae_alba]

I've started tackling a basic walk cycle for Old Man Willies. I'd like the forum's more critical view of the first (actually it's about the dozenth!) pass. I can already see some areas that need tweeking, but I want your crits first.

 

walk_sample.mov

[robcat2075]

briefly...

 

-the job of the rear foot is to push the guy onto the foot that just got planted. If that rear foot is already sliding forward it can't be pushing much.

 

-that gives the appearance that his center of gravity is way too far back. There's no way he could have enough weight off his rear foot to let him slide it forward when he does.

 

-You can get more extension of the rear leg so it can push longer by pushing off with the toe.

 

-after you finish pushing off with the toe and the foot is in the air, the toe doesn't remain curled up. In fact in animation we tend to let it lag behind as the foot is pulled forward thru the air.

 

-after the heel contacts the ground the foot should slap down in another frame or maybe two at most in a normal walk. In animation we often lag the toe one frame after the foot

 

-a real body is falling onto the leading foot and continues down for an instant after the heel contact, then the planted leg starts to push the body back up until it is ready to fall onto the next foot plant.

 

- the front leg should be pretty straight when the heel contacts the ground and you got that right! Don't lose that when you are fixing the other stuff.

 

- I made some videos on posing walks. look in the link in my sig.

 

Richard Williams also covers fundamentals of walks in great detail in his book.

[fae_alba]

-the job of the rear foot is to push the guy onto the foot that just got planted. If that rear foot is already sliding forward it can't be pushing much.

 

-that gives the appearance that his center of gravity is way too far back. There's no way he could have enough weight off his rear foot to let him slide it forward when he does.

 

Robert, thanks for your tips, they will be incorporated today...that rear foot sliding was the bane of my existence last night. For the life of me I couldn't get it to stop, even though I was keyframing the living hell out of the foot null to get it to stay put....

[fae_alba]

take two.... as easy it this stuff may seem...it aint!

 

 

walk_sample.mov

[robcat2075]

What's the character rigged with? Raising the heel seems to be giving you trouble.

 

The biggy I see is that he's reaching his lowest point before his heel hits the ground. It comes after.

 

Have you watched my vid? And/or studied Richard Williams section on basic walk?

[fae_alba]

What's the character rigged with? Raising the heel seems to be giving you trouble.

 

I am using the lite rig

 

The biggy I see is that he's reaching his lowest point before his heel hits the ground. It comes after.

 

That "should" be easy enough to fix...

 

Have you watched my vid? And/or studied Richard Williams section on basic walk?

 

I did watch your vid(s) ( a 3 part er no less!). I haven't reviewed the R Williams section yet and will hunt it down today.

 

Thanks for your input!

[robcat2075]

I think walks are THE hardest common activity to get right. It's really a bunch of different body mechanics problems all rolled into one.

 

I like that you are doing the walk moving forward rather than treadmill style. That's a much better tactic.

[fae_alba]

I like that you are doing the walk moving forward rather than treadmill style. That's a much better tactic.

 

Funny you should mention that: I was contemplating changing the cycle to match the tut in the TAO:AM. So allow me to ask you this...if I continue with the walk cycle as I have started, do I need to define stride length, and does it alter how I approach path constraints etc?

[robcat2075]

Funny you should mention that: I was contemplating changing the cycle to match the tut in the TAO:AM. So allow me to ask you this...if I continue with the walk cycle as I have started, do I need to define stride length, and does it alter how I approach path constraints etc?

 

If you're wanting to eventually use a path constraint you'll want to do it treadmill style since that is the convention path constraints anticipate you will use.

 

But animating with real motion forward is a more traditional method and makes the body mechanics problems more realistic. In most feature animation walk "cycles" aren't used because they look like robotic repetition after the first repeat. If a character needs to walk somewhere they animate every step and show the character thinking about why they are going where they are going and appropriately starting and stopping along the way.

 

A Hanna-Barbera cartoon uses walk cycles disguise what would otherwise be a static dialog scene. They do lots of walking and talking in "Scooby Doo" for example. But feature animation tends to avoid such contrivances.

 

Walk cycles got used in TWO for economy's sake, but they are pretty obvious.

 

A good use of cycles in TWO was the parade scene. Ken Heslip animated a cycle that marched forward two steps and made the beginning match the end. Then at the end of the cycle he moved the character's model bone forward the exact distance of two steps to pick up where the last cycle ended. This has the advantage of not altering the length of the cycle as it was animated; a path constraint stretches the timing to fit the distance needing to be covered.

 

Cycles work for the parade because that's a situation where the characters really are trying to repeat their motion.

 

The longest stretch of walk cycle animation I've seen in a Disney movie was the opening of "Robin Hood" and that was... a parade.

[fae_alba]

Funny you should mention that: I was contemplating changing the cycle to match the tut in the TAO:AM. So allow me to ask you this...if I continue with the walk cycle as I have started, do I need to define stride length, and does it alter how I approach path constraints etc?

 

If you're wanting to eventually use a path constraint you'll want to do it treadmill style since that is the convention path constraints anticipate you will use.

 

But animating with real motion forward is a more traditional method and makes the body mechanics problems more realistic. In most feature animation walk "cycles" aren't used because they look like robotic repetition after the first repeat. If a character needs to walk somewhere they animate every step and show the character thinking about why they are going where they are going and appropriately starting and stopping along the way.

 

A Hanna-Barbera cartoon uses walk cycles disguise what would otherwise be a static dialog scene. They do lots of walking and talking in "Scooby Doo" for example. But feature animation tends to avoid such contrivances.

 

Walk cycles got used in TWO for economy's sake, but they are pretty obvious.

 

A good use of cycles in TWO was the parade scene. Ken Heslip animated a cycle that marched forward two steps and made the beginning match the end. Then at the end of the cycle he moved the character's model bone forward the exact distance of two steps to pick up where the last cycle ended. This has the advantage of not altering the length of the cycle as it was animated; a path constraint stretches the timing to fit the distance needing to be covered.

 

Cycles work for the parade because that's a situation where the characters really are trying to repeat their motion.

 

The longest stretch of walk cycle animation I've seen in a Disney movie was the opening of "Robin Hood" and that was... a parade.

 

Robert,

 

that goes a long long way in explaining/confirming what I already suspected. So treadmill walkcycles and path constraints are not the way I will be going, since my goal is the more realistic/natural feel.

[fae_alba]

So, here we go with a basic walk take 2. Quickly browsed thru Richard William's Animator's Took Kit, then started over. This is what I wound up with...

 

walk_sample.mov

 

Robert, one question for you; I've built the action in the non-treadmill way discussed earlier, but how do you approach it repeating it in the chor so that it doesn't keep popping back to the start position?

 

Crits all around please. I want this character to behave naturally, so the goal is to avoid the "floaty" feeling

[robcat2075]

It looks too fast partly because your animating 12 frame steps but rendering at 30 fps

 

animate at 24 fps. Almost all animation uses that convention.

 

He's also stepping too far for a casual walk.

 

 

 

I know i've said this:

-a real body is falling onto the leading foot and continues down for an instant after the heel contact, then the planted leg starts to push the body back up until it is ready to fall onto the next foot plant.

 

And Richard Williams shows it in numerous diagrams. The body should be heading DOWN thru the heel contact. If it's higher on the next frame that is not going down, that is going up.

 

Look at the diagram on pg 102 (my edition), the one labeled "useless fact" at the top. Does the body go down or up after the heel contact?

 

he's showing you exactly what to do... but you gotta do it.

 

 

 

Robert, one question for you; I've built the action in the non-treadmill way discussed earlier, but how do you approach it repeating it in the chor so that it doesn't keep popping back to the start position?

 

you jump the model bone forward if you intend to use this as a cycle. Or... just animate more steps forward.

[mouseman]

(Sorry to Paul for getting a little off-topic on his post.)

 

If you're wanting to eventually use a path constraint you'll want to do it treadmill style since that is the convention path constraints anticipate you will use.

 

But animating with real motion forward is a more traditional method and makes the body mechanics problems more realistic.

I wish there were an option for walk cycles to be made with motion forward instead of treadmill style. With treadmill style, you have to magically pick a cycle length that matches the speed and getting feet to move at a constant rate on the treadmill is extremely difficult! The trick of using stride length never really works for me.

 

If there were a motion-forward variant, with the last frame of the walk cycle enforced to be the same as the first, and you could move it to wherever you wanted, that would be FABULOUS! Such a time-saver. Plus it would make beginners have a far easier time to get really good results. Animation:Master can compute the movement much better than any human being could.

 

In most feature animation walk "cycles" aren't used because they look like robotic repetition after the first repeat.

And any flaw (such as a popping knee common with IK) in the animation becomes even more noticeable. I think I noticed a few walk cycles in the first Shrek movie.

[robcat2075]

I wish there were an option for walk cycles to be made with motion forward instead of treadmill style.

 

We have that option... animate the forward moving cycle and jump the model bone forward at the start of every new cycle.

 

 

With treadmill style, you have to magically pick a cycle length that matches the speed and getting feet to move at a constant rate on the treadmill is extremely difficult!

This is easy. Set the interpolation of the channel from foot contact to foot lift to be "linear"

If there were a motion-forward variant, with the last frame of the walk cycle enforced to be the same as the first, and you could move it to wherever you wanted, that would be FABULOUS!

 

What do you mean by "wherever you wanted"?

 

 

Animation:Master can compute the movement much better than any human being could.

 

I'll disagree with that. There's about 10 basic ways ( and a zillion variations ) for a computer to interpolate motion between two keyframes; in a walk (and in all character animation) the "way" that needs to be used is different for each pair of keyframes. There's no way for A:M ( or Maya or Softimage or 3DS or any other app) to know what choice to make.

 

A:M only knows it's moving bones around. YOU are the one who knows those bones represent human body parts and you need to make those interpolation choices for A:M.

 

In hand drawn animation the "Animator" would draw the key frames and the "inbetweener" would fill in the missing frames. But the Animator had to explain the interpolation in every case to the Inbetweener. Richard Williams spend a lot of pages discussing what would go wrong when an Animator just assumed the Inbetweener would get it right. (He doesn't call it Interpolation, but that's what the Inbetweener is doing)

 

He's not wasting your time with those inbetweener stories, they are absolutely relevant to CG, because...

 

In CG, A:M is your Inbetweener. If you manage it right you get great results. I f you leave it to do the default each time, you won't.

[dblhelix]

fast progress, fae_alba.

learn it, then re-invent it!

 

robcat,

why is there a model bone? when do i need it? are there common rules to use it, like in a walk cycle "just keep it with the left foot and don't worry" or something?

[fae_alba]

Richard Williams spend a lot of pages discussing what would go wrong when an Animator just assumed the Inbetweener would get it right. (He doesn't call it Interpolation, but that's what the Inbetweener is doing)

 

He's not wasting your time with those inbetweener stories, they are absolutely relevant to CG, because...

 

I ordered The Animator's Toolkit on Thursday from Amazon. Got it on Friday. And those inbetweener stories were some of the first I picked up on in the book. They are relevant, because you need to understand the process in order to make it work..

 

In CG, A:M is your Inbetweener. If you manage it right you get great results. I f you leave it to do the default each time, you won't.

 

This I have learned. At first I thought it was better to not put keyframes in willie-nillie. But as I got more into a walk that is exactly what I had to do.

[fae_alba]

(Sorry to Paul for getting a little off-topic on his post.)

 

Hijack away..I've done it enough to other threads! When I want it back,I'll take it!

[robcat2075]

This I have learned. At first I thought it was better to not put keyframes in willie-nillie. But as I got more into a walk that is exactly what I had to do.

 

Our goal is to use as few keyframes a possible and manage the interpolation inbetween, but a lot of times you just gotta go in and nail something down where it needs to be nailed down.

[robcat2075]

robcat,

why is there a model bone? when do i need it? are there common rules to use it, like in a walk cycle "just keep it with the left foot and don't worry" or something?

 

For characters, I just place the model bone once at the beginning of the scene to get the character in the right area and leave it. After that everything is posing the character parts

 

There has to be some structure in the model that all else is a child of. In A:M it's the model bone.

[mouseman]

I wish there were an option for walk cycles to be made with motion forward instead of treadmill style.

We have that option... animate the forward moving cycle and jump the model bone forward at the start of every new cycle.

Oh, I did not realize. The process is still not clear. Would I have to manually move the bone forward? If so, how could I be guaranteed to get it in the exact right location? Where is this jumping of the model bone done? In a Chor action right below or above the link to the walk cycle? Is there an example or tutorial somewhere that includes both the action and the use of the action in a Choreography? (I have TWO and SO on my HD, if you know of an example there.)

 

With treadmill style, you have to magically pick a cycle length that matches the speed and getting feet to move at a constant rate on the treadmill is extremely difficult!

This is easy. Set the interpolation of the channel from foot contact to foot lift to be "linear"

If there were a motion-forward variant, with the last frame of the walk cycle enforced to be the same as the first, and you could move it to wherever you wanted, that would be FABULOUS!

What do you mean by "wherever you wanted"?

There were two aspects to the idea. You might have gotten the first, but I'll just repeat/restate that.

 

I've thought it would be nice instead of copying the first frame of the animation if the system would automatically do a virtual copy-paste. It would be like Conway's game of life, where the borders wrap around. From the user's point of view, it just wraps. This would make incoming interpolations at the beginning match outgoing interpolations at the end. (Of course I have no idea how much work that would be to program.)

 

Now the second part of the idea is that you just animate forward based on a stationary ground, which is more natural. We still need to specify where the ending location is. So for example, if you started off with the left foot forward and the toe of the right foot taking off at 0, the end of the walk cycle should have the left foot forward and the right foot taking off at some position. This position is where the new cycle will have to start, and thus where the model would be moved to. I was basically thinking that this final frame could be "wherever you wanted" - in both space and time. This is important because I never quite know how long the walk cycle will be ... 1:15? 2:02? And I don't know where it will be ... Z = 95cm? 172cm?

 

I hope that is more clear.

 

Animation:Master can compute the movement much better than any human being could.

 

I'll disagree with that. There's about 10 basic ways ( and a zillion variations ) for a computer to interpolate motion between two keyframes; in a walk (and in all character animation) the "way" that needs to be used is different for each pair of keyframes. There's no way for A:M ( or Maya or Softimage or 3DS or any other app) to know what choice to make.

 

A:M only knows it's moving bones around. YOU are the one who knows those bones represent human body parts and you need to make those interpolation choices for A:M.

My statement was too broad -- sorry about that. I was not trying to make a general statement about all interpolation and inbetweens. The interpolation I was talking about was the rate that the ground moves under the character within a treadmill-style walk cycle. Being human, I have a very hard time getting my character's foot to go at a constant rate on the treadmill so the foot appears to be stationary in the choreography. This is the specific movement computation that I was trying to indicate that the computer could compute much better than humans. The problem is that the theoretically simple operation of keeping a foot planted on the ground is tremendously difficult for a human in the treadmill style of walk cycle. It is much easier for a human being to deal with a fixed point of reference than try to get bones to match up with some rate of floor moving beneath the character. What is even harder is that I am always changing the length (thus the rate) when making a walk cycle, so all of the rate-based assumptions I originally made have to be re-computed.

 

I agree with all of the other stuff you mentioned about interpolation.

[robcat2075]

Would I have to manually move the bone forward? If so, how could I be guaranteed to get it in the exact right location?

Yes. In your action you can measure how far forward the character moved. Use that as a basis for eye balling it. Easy to do since youcan scrub the chor in real time to look for jumps.

 

 

 

Where is this jumping of the model bone done? In a Chor action right below or above the link to the walk cycle?

In the main chor action.

Is there an example or tutorial somewhere that includes both the action and the use of the action in a Choreography? (I have TWO and SO on my HD, if you know of an example there.)

 

The marching band scene in TWO did this for all the band members. 2_01_80? (be warned that it takes about an hour to load on my PC. YOu'll need to hide all but a few characters and change them to bounding box mode to scrub anything) You might try dropping one marching band winkie in a chor

 

 

There were two aspects to the idea. You might have gotten the first, but I'll just repeat/restate that.

 

I've thought it would be nice instead of copying the first frame of the animation if the system would automatically do a virtual copy-paste. It would be like Conway's game of life, where the borders wrap around. From the user's point of view, it just wraps. This would make incoming interpolations at the beginning match outgoing interpolations at the end. (Of course I have no idea how much work that would be to program.)

It could be done but the current copy/paste + thoughtful brain power can get the same result so it's unlikely thei feature will be implemented.

Now the second part of the idea is that you just animate forward based on a stationary ground, which is more natural. We still need to specify where the ending location is. So for example, if you started off with the left foot forward and the toe of the right foot taking off at 0, the end of the walk cycle should have the left foot forward and the right foot taking off at some position. This position is where the new cycle will have to start, and thus where the model would be moved to. I was basically thinking that this final frame could be "wherever you wanted" - in both space and time. This is important because I never quite know how long the walk cycle will be ... 1:15? 2:02? And I don't know where it will be ... Z = 95cm? 172cm?

As Martin woudl say it's possible like going to Mars is possible.

 

Ultimately the complications of canned walk cycles are good reasons to animate walks directly.

 

At AnimationMentor the lecture on walks began with a clip of every animator they knew saying how much they hated doing walks. So it's not a trivial thing.

 

But the walk cycle is more trouble than it's worth if you need to do things besides walk.

[fae_alba]

The saga continues....I think it's getting better, at least the process is getting easier. There are still issues, there is a popping at the transition, but I'm thinking that's where I'm moving the body bone forward at the end of each stride. Plus it seems that he is rising up a bit too far for an elderly gent, and I need to relax the hands quite a bit. But give me your thoughts...

 

walk_sample.mov

[TheSpleen]

looks much better hpwever it kinda looks like he is tip toeing. sort of.

[robcat2075]

One easy check in a basic walk... is the body moving forward at a consistent speed? Frame after frame after frame after frame... It almost always should. Is yours?

[fae_alba]

looks much better hpwever it kinda looks like he is tip toeing. sort of.

 

Sort of?? For an old guy he's practically tip toeing thru the tulips!

[fae_alba]

so...here we are yet again. I've started doing the walk cycle entirely in the chor, and figured to work on just the legs for now. Once I've go the legs down, I'll move on to the rest.

 

walk_sample.mov

[robcat2075]

That's looking better. The foot is being lifted higher than is really should.

 

 

however, lets try an experiment

 

Start a new chor and pose out just the heel contact poses for about 4-6 steps. pose the whole body, arms and torso and everything, but for now just key frames at the moment of the heel contact poses.

 

Like Richard Williams has drawn in his book.

 

Use hold interpolation so the character snap's from one key to the next with no in between.

 

Then post that and we'll look at it.

[fae_alba]

Start a new chor and pose out just the heel contact poses for about 4-6 steps. pose the whole body, arms and torso and everything, but for now just key frames at the moment of the heel contact poses.

 

Like Richard Williams has drawn in his book.

 

Use hold interpolation so the character snap's from one key to the next with no in between.

 

Then post that and we'll look at it.

 

walk_sample.mov

[robcat2075]

walk_sample.mov

 

It's promising but that's only 3 steps. Can you do at least 4 (6 would be better) and then we have something to work with. With more steps it's easier to see if something wrong is just a fluke or if you are posing it that way repeatedly.

 

Leave out the "passing pose"

 

And to make it easier to see, pull the light back so his feet are casting shadows for every step. When his feet don't cast shadows it's near impossible to tell if they are contacting the ground.

 

You got the Hold interpolation right! Most people are baffled by that.

[dblhelix]

the ground grid could prove helpful when using walk cycles. scale it so the meridians match the length of the stride and then rotate it so it's perpendicular to the models orientation - so for as long as the model walks at the same pace you can match/anticipate where s/he'll end up.

[fae_alba]

Thanks guys! I'll get right on it asap (tonight I have to do paying work though)

[fae_alba]

So I started back into the walk in a chor last night. I thought that I could simply copy the key frames from each foot contact, paste them in each consecutive frame, and move the body bone forward when needed to keep the poor old man moving forward. Sadly that didn't seem to work out. Each time I tried the poor old guy looked like he was taking a flying leap. I think it has something to do with the Lite Rig, and the body controls null. It would seem, that I will have to manually key frame each contact point, which in the long run I guess is optimum since it provides a more organic look (potentially), but also opens me up to not being entirely consistent as well.

 

[robcat2075]

So I started back into the walk in a chor last night. I thought that I could simply copy the key frames from each foot contact, paste them in each consecutive frame, and move the body bone forward when needed to keep the poor old man moving forward. Sadly that didn't seem to work out. Each time I tried the poor old guy looked like he was taking a flying leap. I think it has something to do with the Lite Rig, and the body controls null. It would seem, that I will have to manually key frame each contact point, which in the long run I guess is optimum since it provides a more organic look (potentially), but also opens me up to not being entirely consistent as well.

 

 

One way...

 

-use Markers to plot where every heel contact is going to be

-move the body (the bone that controls the hips and everything above) between the first two markers

-pose the front foot (right or left, make a choice) to make the heel contact pose at the front marker

-pose the rear foot to make the push off pose at the back marker. Remember the back marker marks where the rear foot's heel came up from, not where its toe is

-adjust the body up/down/left/right so that the legs are not too bent or pulled too tight.

(-at this point you may decide you need to fine tune the foot poses, then fine tune the body position some more, then the feet again...)

 

-pose the arms and spine and head and whatever, then make sure it's all keyed on this frame

 

advance to the frame of the next heel contact

 

-move the body to between the 2nd and 3rd markers (the legs will look like they are left behind)

-bring the rear foot forward and pose it to make the heel contact at the 3rd marker (now it's the front foot)

-repose the old front foot (now the rear foot) to make the push off pose

-adjust the body again so the legs are right

-pose everything else and key it

 

advance to the frame of the next heel contact

-move the body to between the 3rd and 4th markers...

 

wash rinse repeat... see the pattern?

 

 

On another note, are you sure the Lite rig doesn't have a heel raiser? that would simplify posing.

[HomeSlice]

On another note, are you sure the Lite rig doesn't have a heel raiser? that would simplify posing.

The LiteRig does have a heel raiser