Rotoscoping running - How?

Rasheed · Post by **Rasheed** » Thu Sep 21, 2006 5:03 pm

I'm wondering how one could do rotoscoping with AS, just as a study to improve my animation skills. I think I should try to copy nature first and only then try to make it more "cartoony" and entertaining.

I have downloaded (iPod video) this stock footage video from Google and distilled the following runcycle using several techniques.

I already have the 36 frames of the above runcycle prepared in a separate folder.

Now how can I do rotoscope this footage? Should I study the movements and transfer the observations to an animation, or should I use the footage as a template and copy it to a bone animation, or perhaps both? My first observation is that the center of the pelvis is absolutely stable in one position.

heyvern · Post by **heyvern** » Thu Sep 21, 2006 6:10 pm

I would actually use a movie file... only need one layer.

If you step through the timeline a movie file layer (image layer) will step through the frames and you can "trace" each frame or just use it as a reference.

Bones or vector motion? Uh... up to you. I would use bones... but that's just me.

EDIT:

The "frozen" hips look very odd to me. I don't think... it looks quite natural.

-vern

Rasheed · Post by **Rasheed** » Thu Sep 21, 2006 7:44 pm

heyvern wrote:The "frozen" hips look very odd to me. I don't think... it looks quite natural.

It looks like the data was taken from a non-world class runner on a tredmill, probably a 18-20 years old student in some CGI media lab with markers stuck on his joints. Seasoned runners tend to have a more stable head, although there are exceptions (e.g. Paula Radcliffe).

I will try to learn from copying nature and see how this differs from walking. Then I will have a looksy at Richard Williams' book. I think I'll take a stick figure as a model. Perhaps I also should make some footage of my own at my local track and field club, because there are many different styles of running and the example in the animation is just one of many.

I'm looking forward to this study, because I like running

heyvern · Post by **heyvern** » Thu Sep 21, 2006 9:35 pm

Those hips are PERFECTLY motionless... not even a wobble. There is some kind of... forced constraint.

Even someone on a treadmill would have hip motion. That sucker is locked solid.

-vern

Rasheed · Post by **Rasheed** » Fri Sep 22, 2006 2:53 pm

I guess when I start rotoscoping footage of actual runners, it will become clear what actually happens with the hips. For now, I want to learn how to rotoscope footage, so I keep using the running CGI skeleton.

After trying to rotoscope using bones, I saw several things happening:

the limbs shorten and lengthen, so I need to use bone scaling
the arms are joined by the shoulders, which rotate during the action
the legs are joined by the hips, which rotate during the action
there are a lot of keyframes needed to keep up with the complex movements of the arms and legs, especially the shortening and lengthening of the bones

I hadn't connected arms nor legs, so I had to move the shoulder and hip joints with bone translation. That wasn't a problem with the legs, because all is visible, but the left shoulder keeps apearng and disapearing behind the chest, leading to a muddy animation of the left arm.

The head and chest are problems area as well, because both turn during the animation. I guess I have to solve that using different views of the head within switched layers and do the same with the chest.

7feet · Post by **7feet** » Fri Sep 22, 2006 6:55 pm

Instead of translation for the shoulders, try including a parent bone for each arm that starts at the spine and goes to the shoulders, the way your scapula moves with your sholders. For the hips, since they don't have that individual movement, you could make a similar setup, but have one of the hipbones angle constrained (by minus 1 to mirror) the other, as well as scale constrained by 1. That'll allow you to adjust the angle and perspective of the hips using only one bone.

I've been meaning to make one of my "everything and the kitchen sink" tools for bones, I've done stuff like this and not having one is dumb, going back and forth between all the bone tools while animating is a huge pain in the bum. I'll probably do that momentarily.

Rasheed · Post by **Rasheed** » Fri Sep 22, 2006 8:17 pm

7feet wrote:For the hips, since they don't have that individual movement, you could make a similar setup, but have one of the hipbones angle constrained (by minus 1 to mirror) the other, as well as scale constrained by 1. That'll allow you to adjust the angle and perspective of the hips using only one bone.

That was actually a very good tip (only the angle constraint needed to be 1, instead of minus 1), because this allowed me to use the visual part of the shoulder to put the invisible part in the correct position.

Here's the animation on YouTube:

http://www.youtube.com/watch?v=5CcOFA1LDsM

And there were some interesting additional observations using this method:

unlike what you might think, the hips DO go up and down
the back is lengthened and shortened through perspective, so the mid between the shoulders also goes up and down and back and forth

So the trick seems to be to do the following:

Track the position of the center of the hips, using bone translation of the root bone
Track the position of the mid between the shoulders, using bone rotation and scaling of the back bone
Track the positions of the hips and the shoulders, using bone rotation and scaling
Align the upper legs and arms, using bone rotation and scaling
Align the lower arms and legs, using bone rotation and scaling
Align the feet, using bone rotation and scaling

This is only the primary animation. There is of course secundary animation, such as the turning of the the head, flapping of skin folds, muscles and such. Furthermore, not everything has been considered (e.g. the turning of the upper body, changing perspective of the hands and feet). I guess these things are only important if you use a more realistic model.

heyvern · Post by **heyvern** » Fri Sep 22, 2006 8:31 pm

Hey that looks great!

Yes, very good tip on the bone constraints... I use that with eyes and it never ocurred to me to use it anywhere else.

-vern

Rasheed · Post by **Rasheed** » Fri Sep 22, 2006 10:42 pm

I redid the rotoscoping with points attached to the joints and a keyframe every three frames. Now the animation becomes even more realistic. However, I need to remember not to animate mechanically, but rather try to see and understand what is happening in four dimensions (space and time) and check and re-check if my animation reflects that. It still is a very mental process, trying to bring that stick figure to life. Lots of trial and error and a slow process.

http://www.youtube.com/watch?v=4UzF51GnZ-8

I think the next step is creating a 2D model, with which I can recreate the illusion of a running individual. I should test the model in animation and see if any refinements are necessary.

7feet · Post by **7feet** » Fri Sep 22, 2006 11:53 pm

Oops, I was thinking of another technique, was just trying to get the idea out before I had to run out to a shoot (I like Comedy Central, they pay on time).

You lay out the problem well. I've thought, for quite a while now, about setting something up to interpret mocap files into bone motion. I just really hate the matrix math, or at least setting it up. I try and write it and I have to fish through dumb errors for hours.

(not entirely related, but somewhat) That's one of the things that I find people forgetting. Bone scaling is a very useful, even necessary tool to create the imrepssion of perspective. If you do a walk toward camera, everything is about perspective.

Nice job on that roto, it shouldn't be hard to wrap a character around it. Hey, there's a market! Premade actions/bone structures that folks can draw their own stuff around.

jahnocli · Post by **jahnocli** » Sat Sep 23, 2006 10:51 am

Interesting. The main problem for me with rotoscoping animation is the weight -- or lack of it. It's a funny thing, but something completely credible in real life (whatever that is) looks all "floaty" when you make an animation of it. Be interested to see where this leads...

Patmals · Post by **Patmals** » Sat Sep 23, 2006 11:14 am

I am wondering if this skeleton is a poser animation.. this is easily done within the software..there is a skeleton model and a walk designer which has attributes that can be changed to run

Rasheed · Post by **Rasheed** » Sat Sep 23, 2006 11:45 am

I refined the model a little, to include toes and a head.

http://www.youtube.com/watch?v=_269n1BHXtA

I seems the feet and toes are essential for creating an illusion of weight, as is the head and the movement of the arms. When force is issued by the muscles, the head and arms suddenly shift position. And the feet... they seem to stick to the floor and seem to like it there. It could be that this actually gives our brains the clue how heavy this runner is. The feet of heavier runner stick longer to the floor than those of lighter runners.

Boy, you learn a lot from rotoscoping footage. I guess, you would never learn this from observation alone, even if you would watch the footage frame-by-frame. You need to do the animation (and not merely watch it) to see how it all works and a stick figure is probably the most useful tool of all.

jahnocli wrote:Interesting. The main problem for me with rotoscoping animation is the weight -- or lack of it. It's a funny thing, but something completely credible in real life (whatever that is) looks all "floaty" when you make an animation of it. Be interested to see where this leads...

I have done some rotoscoping before and my problem then was that rotoscoping can't be translated directly to a cartoon character. Some translation has to take place. Perhaps I should now look what Richard Williams has to say about running to check my findings.

Rasheed · Post by **Rasheed** » Sat Sep 23, 2006 2:22 pm

Before I check the Animator's Survival Kit, I noticed two strikes things:

each foot is in contact with the floor for 12 frames over a running cycles of 36 frames (so, one third of the time); this is obviously a heel striker
because this motion seems to be captured from a treadmill, the feet are making strange movements, as compared to running on a non-moving surface; it is known by experienced runners that running on a treadmill is quite different from running on a road

So while this may be a fine exercise in interpreting running video, it is probably not very typical for normal runcycles.

I need to be looking for different footage of a runner. At least now I know what to look for and how to rotoscope the footage.

heyvern · Post by **heyvern** » Sat Sep 23, 2006 8:30 pm

Thanks for this tip Rasheed!

I needed a "sexy walk" and... I was tired... and... lazy...

So I used a couple of screen grabs from a poser walk cycle as reference.

Saved a ton of time and looks great.

I did "exagerate" the heck out of the motion. Basically... if the hips swing in my reference... I tripled the amount of swing in Moho... I just pushed the motion a lot further.

-vern