AuthorTopic: Walk (failure)  (Read 5341 times)

Offline Tourist

  • 0010
  • *
  • Posts: 376
  • Karma: +1/-0
    • View Profile

Walk (failure)

on: December 30, 2014, 08:44:22 pm


Lesson learned:  Mocap data is rather boring.  Also, if you don't include the toes, the walk looks clumsy.  I guess I have to actually learn this animation stuff instead of taking shortcuts.

Image is mocap on a Poser skeleton figure, colored lines and head added by me.  The bobbing head is just a copy-paste thing that I didn't bother to edit out.

Tourist

Offline tim

  • 0010
  • *
  • Posts: 240
  • Karma: +2/-0
  • Founder of Odd Tales - Art Director
    • View Profile
    • Tim Soret - Showreel

Re: Walk (failure)

Reply #1 on: December 30, 2014, 09:31:45 pm
How is it a failure ?
It's better than 95% of walk cycles seen in video games these days.
Consider it a base animation you're going to iterate upon.

It's just a bit "bland" as is, but you can enhance it, specifically the toes. The arms are a bit boring too. But overall the animation is really smooth, the character is grounded correctly in the floor, we can sense the weight, the speed, the balance, so I guess it's a success, not a fail at all. Just add a bit more spice & flavor to make it pop and a bit less realistic and it should be good.
Founder of Odd Tales
Art Director - Game Director - Game designer - Motion designer

Offline Tourist

  • 0010
  • *
  • Posts: 376
  • Karma: +1/-0
    • View Profile

Re: Walk (failure)

Reply #2 on: December 31, 2014, 01:44:18 am
It looks realistic because it is motion capture (mocap) data.  Probably from someone on a treadmill.   

The problems are:

The 3d model has no toe joint, the data does, so the ankles and feet look wonky.  This is fixable by changing the model and re-rendering.

The real human motion doesn't correspond nicely to the grid.  There are a lot of small movements and angles that don't work well at this scale.  The head should swivel or tilt a bit, but at this scale that requires sub-pixel work.  Likewise the angles of the limbs either produce jaggy edges or shape distortions; watch the shape of the bright red shin/calf.

This process needs some more work.

Tourist

Offline Cyangmou

  • 0011
  • **
  • Posts: 929
  • Karma: +3/-0
    • cyangmou
    • http://pixeljoint.com/p/32234.htm
    • cyangmou
    • View Profile
    • Pixwerk Homepage

Re: Walk (failure)

Reply #3 on: December 31, 2014, 01:59:23 am
the main problem with this is that it has 15 frames.
One step takes 7 frames, the other one 8 frames, the speeds of the legs are different which will make it really hard to get it working for an exact underlying 2D code, to prevent animation-clipping between different animations.
To keep the left and right step consistent you need to work with an even amount of frames.
THis also needs to more possibilities to balance out the movement arcs and play with the timing.

Little details like the toe movement can be easily edited in.


But with animation-capturing (aka rotoscoping) it's like with realism. It's not possible to reach a visual more exact result then with just filming something.
However the question is "how much realism do you need and where can you freely exaggerate"
It usually takes the same time to construct it with pixelart from scratch, since this guarantees that there also won't be any inconsistencies you have to iron out otherwise (which also might take quite a while)

Mocap is a great reference.
« Last Edit: December 31, 2014, 02:02:28 am by Cyangmou »
"Because the beauty of the human body is that it hasn't a single muscle which doesn't serve its purpose; that there's not a line wasted; that every detail of it fits one idea, the idea of a man and the life of a man."

Dev-Art
Twitter

Offline Gil

  • 0100
  • ***
  • Posts: 1543
  • Karma: +1/-0
  • Too square to be hip
    • http://pixeljoint.com/p/475.htm
    • View Profile
    • My Portfolio

Re: Walk (failure)

Reply #4 on: December 31, 2014, 10:40:25 am
I agree with Tourist, it looks bland and mocap is not a good way to go, UNLESS you know how to act well. Exaggerate your mocap movements. Theater actors know that to have the audience see something interesting from that far away, they have to really push movement. A regular walk, mocapped and then transformed to animation is going to look bland, but if you do a completely over the top walk, it'll look just right.

If you were to see people recording a music video in real life, it'd look silly as all hell, because they are asked by the director to exaggerate in such a silly way it's hard not to burst out laughing. When filmed, it looks "just right".

Or, you know, just learn animation ;)

Offline Cyangmou

  • 0011
  • **
  • Posts: 929
  • Karma: +3/-0
    • cyangmou
    • http://pixeljoint.com/p/32234.htm
    • cyangmou
    • View Profile
    • Pixwerk Homepage

Re: Walk (failure)

Reply #5 on: December 31, 2014, 03:13:03 pm
feel like I need to add a few sentences here.
Practically the problem we have with the discussion here is, that we neither know the exact tone tourist wants to have for the animation, nor what he is trying to achieve.


A good animation is always just good in the context of the overall production. If something stands out too much it simply feels weird.


I agree with Tourist, it looks bland and mocap is not a good way to go, UNLESS you know how to act well. Exaggerate your mocap movements. Theater actors know that to have the audience see something interesting from that far away, they have to really push movement.

Yeah one needs to be careful where he gets his mocap references from.
But what's always right in a mocap reference is the relative movement of the arcs - you can easily find out when each part has which position which makes it a great learning source, even if the model don't acted that great.

In real sports which also has a big a big audience, none of the athletes is exaggerating any movement and nonetheless if you know what you look for, you can easily tell what they are exactly doing. Those are mostly efficient and very elegant movements, because they are on spot and also a great learning source of how the human body moves.

The factor of exaggeration however always depends what one wants to make - the tone of the project
For an exact simulation of a movement, exaggeration won't work that good.
While for most games and films it's a great possibility, but it always has to be in line with the style.
A lighthearted Disney movie can get away with broken lings and over the top animations (just look at Goofie's walks) while a more serious production has overall less exaggeration.
We could go and compare goofy with the prince of Snow White here - both are great, but both have a very different tone to them.

I would say as rule of thumb: the more earnest the tone you want to have, the more realistic you have to be and therefore you need less exaggeration.

A regular walk, mocapped and then transformed to animation is going to look bland, but if you do a completely over the top walk, it'll look just right.

I think for pixel art as a medium mocap isn't really that effective to go from, since a big part of any pixel art is stylization, even for big resolutions you have to stylize a lot.
We would need to work in an insane big resolution (which is unfitting for pixel art) to be able to capture all the small details in a readable way, so it goes anyways against what pixelart is effective for.
Nonetheless we can get all hose small details working and good looking if we stylize and exaggerate them in the right way (right way depends on the tone).

If we look at all the stylized NES and SNES games I would say that strong exaggeration is the way to go for any pixel art game. The lower the res, the more exaggeration you will have anyways, because the distance a single pixel travels relative to the char size increases.

some side-notes:
For 3D animation it's also worthwhile to look up examples like Polar Express or Beowulf, which were made by using motion capturing, but both films touch the uncanny valley. Nonetheless
Both films are great examples of what you want to avoid.

What was also really interesting for me was to look at the graphics of the 3D zelda series. They went from "realistic low poly" (N64 Zeldas) to hyper stylized and exaggerated (Wind Waker) back to a more realistic style (Twilight princess) to a mixture of both previous artstyles (Skyward Sword)
In terms of aestethics or by the question "what artstyle might be the most fitting one for my game", Zelda is definitely one of the best subjects to study out there, because they had a lot of very diverse, but interesting artstyles.
The interesting experience I had with those games is that I really disliked WindWaker's artstyle as it came out, but by now it feels to me like it has the most "timeless" graphics of all games released so far, while others in comparison have aged badly.
"Because the beauty of the human body is that it hasn't a single muscle which doesn't serve its purpose; that there's not a line wasted; that every detail of it fits one idea, the idea of a man and the life of a man."

Dev-Art
Twitter

Offline Tourist

  • 0010
  • *
  • Posts: 376
  • Karma: +1/-0
    • View Profile

Re: Walk (failure)

Reply #6 on: December 31, 2014, 06:05:37 pm
An even number of frames, of course!  I am enlightened.

The realism is nice, but more importantly the walk data is consistent across frames and different animations.  But consistent position (angles) and consistent shape (form) are mutually exclusive thanks to the grid.  Vectors and 3d tend to provide consistent shapes, mocap provides consistent angles.   

This animation wasn't made for any particular story to tell.  I have this lovely walk data, and wondered how to use that for pixel animation.  Exploring process and seeing how efficient and effective it was at producing animation.  Trying the oddball ways of doing things helps me understand why the traditional methods work.  Also, laziness.  :)

On to the next attempt,
Tourist

Offline Gil

  • 0100
  • ***
  • Posts: 1543
  • Karma: +1/-0
  • Too square to be hip
    • http://pixeljoint.com/p/475.htm
    • View Profile
    • My Portfolio

Re: Walk (failure)

Reply #7 on: January 01, 2015, 12:33:55 pm
Okay, yeah, Cyangmou, that's fair enough, good middle ground :)

Though I have to mention that I dislike the prince from Snow White. I consider him to be very stiff, compared to for example the prince from Sleeping Beauty :). Nitpicking now.

Offline Tourist

  • 0010
  • *
  • Posts: 376
  • Karma: +1/-0
    • View Profile

Re: Walk (failure)

Reply #8 on: February 02, 2015, 02:44:58 am
I got a little further on this and then stalled out.  In the interest of sharing, here's what else I learned.

Here is a basic run animation, 30 frame motion capture:



Software interpolated to 16 frames:



This works ok, but it's a little bit stiff and floaty because the interpolated frames don't capture the extremes of the joint positions.  I took a looking at the original frame by frame, and the timing is a mess.  The main problem is that the left and right sides hit the extreme joint positions a frame or two apart. 

The body is at its lowest on frame 4, with one foot on the ground and that knee at its most bent position.  The other leg, passing underneath, continues to bend until frame 5, after the body has started rising up.  The shoulder makes a large move between frames 5 and 6.

Similarly, the figure takes off from the ground on frame 7, but the other knee continues to push forward until frame 8.  And so on.

Spreading out the animation to 480 frames and then selecting frames that keep the extreme positions, but as spaced out as I could, gives this:



This looks pretty good.  It's not as smooth as the previous animation because it sacrifices timing for position.  It looks more bouncy.  Over the 16 frames it also spends one more frame on the ground so it looks more grounded rather than floaty.

Time taken for smooth interpolation: a couple of button presses, call it 2 minutes.
Time taken for manual frame selection: approx 20-30 minutes .

That's not a huge difference compared to the time required to slap down all the pixels, so the manual frame selection is probably worth it.


Then into the vector art application!  Oh wait, I don't have a vector art application.  So I wrote one.  It looks like this:



It's similar to Pivot stick animator, but it handles closed forms and curves.  This not only lets the user draw an outline on an image, but also add in other flowing lines like hair or clothing without worrying about cloth physics simulators.  It's also a clunky, clumsy and unfriendly tool. 

Setting up the first frame takes a 30-40 minutes, but simply moving the shapes to the next frame only takes 5 minutes per frame unless the figure changes shape significantly.   This doesn't save much time compared to just sketching lines like I did in the first post, but it is a net savings for longer animations (8 frames no, 16 frames yes). 

Then to pixels!

And here I stalled out, because I am just too slow at slapping down pixels.  I also noticed I had a tendency to simply fill in the vector shapes rather than use them as guides to actually draw.  That's no good.  If I just filled in the vectors I'd get animated vector art, and then I might as well use pure vector tools instead.   I don't really want to do that.

I need to go practice basic pixels and speed pixelling a bit.  I declare this effort to be on hold.  Hopefully this post is useful for others to learn from my blunders.

Tourist