Roadrunner

[Simon Edmondson]

Have been trying to achieve a running style like that of the roadrunner in the WB shorts with very mixed results. Thought of another way to try in the meantime and wondered if anyone could kindly offer help on a couple of points ?

 

Initially I set up an action for each foot then tried to put that on a path in a chor but that didn't seem to work, so tried to set it up using a path, but paths don't seem to be an option in actions ?

Then tried to set it up in the chor but two questions,

How do you get it to loop ( post extrapolation ? ) and can you then move it along using the track by moving the path ?

regards

simon

Road_Runner.prj

[robcat2075]

I'm not sure I understand teh PRJ. The path is SO huge compared to the feet. Is it really the motion they are supposed to be making under the body?

 

If you are wanting feet to do a circular motion under the body in an Action it will be vastly simpler to keyframe that circular motion. Four keys, maybe 8 at most?

 

I'll note that making them blur like the road runner's feet will probably take another technique to create that appearance rather than actual moving, modeled feet.

[Simon Edmondson]

I'm not sure I understand teh PRJ. The path is SO huge compared to the feet. Is it really the motion they are supposed to be making under the body?

 

If you are wanting feet to do a circular motion under the body in an Action it will be vastly simpler to keyframe that circular motion. Four keys, maybe 8 at most?

 

I'll note that making them blur like the road runner's feet will probably take another technique to create that appearance rather than actual moving, modeled feet.

 

 

Robert

Thank you for you reply. I may have done it, albeit with a fair bit of refinement needed ?

 

This is it with 75% Blur on the render,

 

Runner.mov

 

and this is it with 15% blur

 

Runner15.mov

 

I'll do some more tests tomorrow and see what happens.

My present thinking is to animate the path along the track as the figure moves, so the feet stay under him as he travels the distance. Will try to refine this tomorrow before attempting the distance.

regards

simon

[robcat2075]

You'll need more blur.

 

Here's a quick attempt that uses 9 keyframes to circuit the loop twice. That is so when it is going around the second time it is not exactly duplicating the placement of the first circuit. This helps reduce the appearance of the object showing up inteh same place too soon.

 

i think Richard Williams touches on his in his book.

 

Footblur.mov

 

This effect could be improved with refinements.

 

RoadRunnerBlur.prj

[Simon Edmondson]

You'll need more blur.

 

Here's a quick attempt that uses 9 keyframes to circuit the loop twice. That is so when it is going around the second time it is not exactly duplicating the placement of the first circuit. This helps reduce the appearance of the object showing up inteh same place too soon.

 

i think Richard Williams touches on his in his book.

 

Footblur.mov

 

This effect could be improved with refinements.

 

RoadRunnerBlur.prj

 

 

Robert

Thank you very much for your help.

I will look at them tomorrow as sleep beckons.

regards

simon

[Wildsided]

Just throwing this out there. Could you make the legs a separate model that could be swapped out for a loop shape with a blurred texture and maybe a transparency map to make it look like the legs were moving fast. then when he stops swap the other legs back in?

[NancyGormezano]

Had to give this a try - I have always wanted to do a roadrunner type leg. But I wanted the legs/feet to look like spokes of a wheel.

 

1st animation is a 2 pass render with 20% blur

2nd is 2 pass with 75 % blur

3rd is Multi-pass OFF - 75% blur

 

There are 3 sets of legs, and 1 complete model. 1 set of legs (leader) is constrained to a path, all other models are constrained to the leader (with translate and orient with offsets). All models have a looping action with repeat for the wheel action of the legs.

 

If you need more explanation or want the project - let me know

7roadrunner2pass20Blursoren3LOOP.mov

7roadrunner75blur2passSoren3LOOP.mov

7roadrunner75blurNoMPsoren3LOOP.mov

Edited January 23, 2014 by NancyGormezano

[robcat2075]

the motion blur will definitely work better in multipass

 

those are funny, Nancy!

[Simon Edmondson]

Had to give this a try - I have always wanted to do a roadrunner type leg. But I wanted the legs/feet to look like spokes of a wheel.

 

1st animation is a 2 pass render with 20% blur

2nd is 2 pass with 75 % blur

3rd is Multi-pass OFF - 75% blur

 

There are 3 sets of legs, and 1 complete model. 1 set of legs (leader) is constrained to a path, all other models are constrained to the leader (with translate and orient with offsets). All models have a looping action with repeat for the wheel action of the legs.

 

If you need more explanation or want the project - let me know

 

 

Nancy

Thank you very much for your helpful reply.

I tried multiple legs, but on separate renders composited together. Back to the drawing board.

regards

simon

[Simon Edmondson]

Just throwing this out there. Could you make the legs a separate model that could be swapped out for a loop shape with a blurred texture and maybe a transparency map to make it look like the legs were moving fast. then when he stops swap the other legs back in?

 

 

Dan

Thank you for your suggestion

I tried a similar idea but, compositing layers. A separate model, allied with Nancy's suggestion might be a avenue to explore...

regards

simon

[Rodney]

A separate model, allied with Nancy's suggestion might be a avenue to explore...

 

I had a chance to speak with Keith Osborn a few years ago. He was involved in making the new CG version of Warner Bros 'Roadrunner' shorts with Wile E. Coyote.

These shorts featured several segments with the stylistic mutliple arms and legs that you are working toward here.

 

He used the multiple model method to create the fast moving appendage effect.

 

In A:M You could basically do the same thing by saving several copies of your model and either deleting or hiding the unnecessary parts of it in the various models.

Then place them all in the same location in the Chor and go to work.

 

One thing I recall Keith saying is that the multiple models really slowed down his system which I think wouldn't be quite as much of a problem with A:M due to the lightweight spline tech that we enjoy.

Once in the Chor we could even export out a new character that had the multiple appendages... although you'd want to make sure you rename the various duplicate named bones!

 

 

You can see the multiple model method used on heads, feet, arms, etc. on Keith's website.

The shots are at the latter part of his reel and primarily consists of one shot with roadrunner (multiple heads) and coyote (multiple hands).

[Simon Edmondson]

I post this for amusement.

Still testing the ideas others put forward and some hanging over from yesterday. This was the result 10 mins ago.

 

RunnerB75.mov

 

When played it looks as though it is going backwards ( clockwise ) ?

Actually, its animated anti clockwise...

simon

[Simon Edmondson]

...

 

I had a chance to speak with Keith Osborn a few years ago. He was involved in making the new CG version of Warner Bros 'Roadrunner' shorts with Wile E. Coyote.

These shorts featured several segments with the stylistic mutliple arms and legs that you are working toward here.

..

 

Rodney

Pardon me, I missed your post, when posting.

Thank you very much for the info, I shall head there straight away.

regards

simon

[NancyGormezano]

A separate model, allied with Nancy's suggestion might be a avenue to explore...

 

I had a chance to speak with Keith Osborn a few years ago.

 

He used the multiple model method to create the fast moving appendage effect.

 

In A:M You could basically do the same thing by saving several copies of your model and either deleting or hiding the unnecessary parts of it in the various models.

Then place them all in the same location in the Chor and go to work.

 

Maybe I wasn't clear in my above post - but that is exactly what I was doing. I was using a wheel like action for the legs in the above post.

 

Here is another example where I change the leg movement to be just back and forth (no wheel like motion). I did not use blur so that you could see more clearly what the legs were doing. And I did not zero the translate and rotate offsets at the end of the path, like I did in the previous post

 

EDIT: To be more efficient, I could have just constrained the "whole model" to the path, and then constrained the two "legs only" model instances to the "whole model" (with translate, orient and offsets). I did not need the extra "legs only" instance.

7roadrunner0blurbackforth2passKeepoffsetsoren3LOOP.mov

Edited January 23, 2014 by NancyGormezano

[Simon Edmondson]

Nancy and Rodney

Thank you both for your help.

Have been working away at it today with varying success.

Still working on the feet at this point. This is the most recent ( of about 8 )variations.

Going a bit stir crazy so will be keen to try Sports Day at the weekend.

regards

simon

 

Runner_001.mov

[johnl3d]

I think Nancy has the answer

[Bobby]

My idea.

RoadRunner_ak_0000.mov

Road_Runner_Test_a.prj

[Simon Edmondson]

My idea.

RoadRunner_ak_0000.mov

Road_Runner_Test_a.prj

 

 

Bobby

Thank you for your help. Thats an effective idea.

regards

simon

[Simon Edmondson]

Todays test.

Same chor, rendered with blur, and without blur, with alpha channels and those applied as cookie cutter decals.

The blur is too soft I think ?

 

Going to try the other suggested methods tomorrow.

|Thanks to everyone for their help.

simon

 

Running.mov

[Simon Edmondson]

Two pairs of legs, 12 feet on one path, path moved along track. 64 passes, motion blur set to 60%

 

Think at least one extra pair of legs needed.

 

RunningC.mov

[Rodney]

Yeah, that last one isn't reading that the feet/legs are what is being blurred.

It's like something has been superimposed on top of the running.

 

For the effect as I am seeing it play out in my head I don't think you want to approach this so much from a moving sequence perspective but rather one of still imagery.

In each frame the multiple images of fee/legs should be clear and readable. It's the inbetweening then that gets the blur effect.

 

Suggestion: Render out a sequence with static imagery showing exactly where the (multiple) feet and legs are to be placed (note that this will likely NOT play as well as a moving image at this stage).

Now setup a Chor with that sequence as a Rotoscope or Layer and render that sequence again with Multipass on (higher levels should be better) and blur effects on as well.

 

Note that what you are doing here (or can be doing) is basically retiming the animation so you may want to render a longer sequence in the first step than what you will be targetting for length in the second.

 

Where the magic kicks in is where you adjust your keyframes and timing so that on the frames you wish to be blurred more than one keyframe is set on a frame that will be rendered (this will guarantee more blur during those frames.

 

Consideration should also go for manual blur effects such as elongating or deforming the mesh itself so that the renderer has more to use in the blurring effect.

 

Working from a known set of prerendered images can yield better (and more controllable) blur effects than if we render using the actual geometry.

 

The beauty of blurring rendered images is that if you use copies of your image sequence rather than the original files you can easily set the project up again and render with a different effect.

A quick way to duplicate the sequence being to copy/paste the folder that contains the images to another folder and then rename the folder.

I like to use a folder named 'RenderFolder' as my target folder as I can then simply duplicate that folder and then render right over the top of the old imagery.

This can have an added benefit if I only need to rerender specific frames of a given animation.

It's like filling thousands of glasses with colored water and keeping them full and precisely the color desired... some may not need to be filled repeatedly or their color changed while others may.

In this way if there is a 'perfect frame' that you know you want to read clearly in the sequence you can keep that outside the render que and replace any other image that tries to take its place.

 

Hope that isn't too esoteric.

[Simon Edmondson]

Rodney

Thank you for your helpful reply. I will give it a go with the rendering as stills then use as layers, after the current test goes through, that was the technique I used first, or a variation thereof.

The one going through now has three sets of legs with feet. Each leg set is doing the same actions but with the feet targets aligned slightly differently. I tried offsetting the action cycle but the extremes and speed of the cycle hindered the 'read'. So I constrained the additional targets to the originals . Will post the result later, its going to take some time to go through.

regards

simon

[Rodney]

I went spelunking for my copy of 'Animation Crash Course' by Eric Goldberg because I recall some nice suggestions on blur in that book.

Sadly, I didn't find it so I cannot confirm.

 

I believe the key to good blur here is going to be persistence of shape even though the motion is blurred.

The idea is core and very basic to animation in that...

 

Let's say we have three frames of animation

- Starting Position

- Ending Position

- Breakdown or Inbetween

 

Depending on your stylistic goal you will mostly likely need to include one frame that captures all of those three frames in one to achieve the proper smear/smudge/blur.

 

The late Michael Sporn has a variety of examples online to include some stills from Chuck Jone's classic (and majorly blurred) classic 'Dover Boys'.

 

Article:

http://www.michaelspornanimation.com/splog/?p=2927

 

Note that in this article there is a recipe of sorts for CG blur used in Disney's 'Raphunzel'.

In the scene (where the character opens and pops quickly through a door) he says:

Technically, you start slow, you compress the middle to one or two images, and you slow out over a long beat.

He then follows this up with a nice Timing Chart that accounts for the 17 frame sequence of animation.

 

Disclaimer: It should be noted that Michael is suggesting that Disney over did the blurring in 'Raphunzel' and that made the film entirely too cartoony.

The point being that blur works best with cartoon/exaggerated motion.

 

Example image:

 

In the context of a single frame the image looks terribly wrong but when put in motion the effect is (literally and figuratively) transforming.

 

In order to achieve this effect in animation we usually have to either replace the model with a different deformed shape (a copy of the model changed to meet the requirement) or distort the model to account for the perceived motion. At 24 or 30 fps there simply aren't enough frames to capture that fast a motion.

[Simon Edmondson]

Rodney

Thank you very much for your continued help. Much appreciated. I will follow up those references tomorrow.

I watched a Chuck Jones ( always my favourite WB Director ) DVD were they explained about his use of blurred frames in general rather than specific details..

I tried some variations today. I think your earlier suggestions regarding layers will be tomorrow's task.

 

Regards

simon

 

RunningD.mov

 

 

RunningE.mov

[robcat2075]

I think those legs need to be offset more. It looks like he's got two pairs of legs rather than one pair of legs running fast.

[Rodney]

Here is the result of my layer test.

 

I dropped 4 instances of the same sequence into a Chor as a Layer and re-timed (and recolored) the layers.

 

It came out both better and worse than I thought it might.

Better in the sense that it came out mostly as I thought it would.

Worse in that there are movements when the blur proceeds the object (i.e. the blur gets in front of the moving leg)

 

This makes me wonder if A:M takes into account the actual geometry when blurring or whether it's more of a matter of applying a post effect to images rendered.

This test isn't a good one to determine that because all of the blur is created on images (i.e. no geometry is present).

As such there really is no way for A:M to know which direction the legs are moving.

It just blurs what it discovers there in the imagery.

 

If motion blur is a post effect that doesn't take into account direction then I'm not sure we can trust the results of the effect except in specific instances such as ideal movement of shapes and color via screen direction.

 

I suppose that should be easy enough to test.

 

 

Added: I should have also said that I rendered the layers with multipass set to 256 passes.

I did not attempt to adjust the blur except by indirect result of re-timing the ease of the frames of animation of each layer.

 

 

Added: Gif version (apparently the conversion to gif introduced some artifacting.

MidKnightRunning_000.mov