Ghosting And Motion Trails

[newtonvfx]

Hi,

I'm trying to do something I think should be really simple to do, but finding it a hassle. What I'm trying to do is create multiple duplicates of an animated object which follow the original x number of frames behind it. For example. "curve" is copied to "curve1" which has the same animation 2 frames later and "curve2" has the same animation again 4 frames later and so on.

Now, an animated snapshot won't work, because the resulting objects have no animation on them. Duplicating the object's input graph and offsetting is messy because if I make a change to the animation of the original I have to redo it all again. What I want is something that will automatically update when I set and edit keyframes on my original object.

Ghosting is what I want, but it doesn't actually produce geometry I can get at and I want the curves to loft a trail.

The best solutions I have come up with is to either write a MEL script to make the "duplicate input graph" less painful or write a script job that updates the position of the duplicates with the appropriate offset whenever the transform of the original is changes. The second what is not so good because I have to step through the animation to manually keyframe it (or keyframe on every time the script job is run).

So, am I missing something really obvious!? Or do I really need to script a solution. As I see it it would be easy to do this if Expressions could read values from different frames, Is that possible?

Thanks

PS, Im using Maya 2008

[gmask]

If it's less than a dozen or so objects I'd just export the object and path and then reimport it x number of times and offset the animation on each or write a mel script

[newtonvfx]

Thanks gmask,

Yes, that was something else I was considering, but from that reply, it seems I haven't missed anything. I really do need to do something special to get this effect.

Damn! I don't understand why Maya doesn't have the ability to query the value of a attribute from any frame in an expression. Even Nuke, a composting package can do that and animation is less important in comp.

Thanks again for your input

[gmask]

QUOTE(newtonvfx @ 03/09/09, 04:30 PM) [snapback]302677[/snapback]
Damn! I don't understand why Maya doesn't have the ability to query the value of a attribute from any frame in an expression.

of course it can.. you just need to use the -t (time) flag of getAttr

QUOTE(newtonvfx)
Even Nuke, a composting package can do that and animation is less important in comp.

you must not be familiar with motion graphics

[joojaa]

Well theres actually a node that does this trailing effect. So what you do is make a animated snapshot and then animate the snapshot range.

Its just that for maya asing for values in previous frames is a bit unptimal for the system as it must evaluate the context 2 times. So maya actually tries to guide you to other options. So while maya could do a lot more complicated things with the time of evaluating a node 2 times, seriously. You can have 100 independent cubes for the cost of one off time lookup.

Theres also a second node that can do this its called cache.

[newtonvfx]

Thanks for your posts gmask and Joojaa,

gmask, I'm aware of getAttr's -time flag, but you can't use MEL in expressions (can you?). I use "getAttr -t " in the script job approach I tried out. An expression would have been neater for me as it would automatically update, but I do take Joojaa's point on the evaluation at 2 or more times.

I guess if Maya has that restriction for performance reasons, it important to supply the user with enough tools and control to get around it. My feeling is that if an architecture is flexible, scripting and coding is only needed for complex work. I'm starting to come around to Houdini's way of doing things, but maybe that's just my turn of mind.

Thanks for the comment on animated snapshot, that's a great idea Joojaa. I'll have a play with that. So it seems Maya has the tools in this case.

BTW, I don't think many people use Nuke for motion graphics, it's mostly set up for straight up compositing, but what do I know, as you might have noticed from my ID, I work in visual effects, so you're right I'm not that familiar with motion graphics.

Cheers!

[gmask]

QUOTE(newtonvfx)
gmask, I'm aware of getAttr's -time flag, but you can't use MEL in expressions (can you?). I use "getAttr -t " in the script job approach I tried out. An expression would have been neater for me as it would automatically update, but I do take Joojaa's point on the evaluation at 2 or more times.

of course you can but there are limitations.. for example piping the result of a command such as getAttr using syntax such as $x=getAttr object.tx will break when you reference that expression. Anything in between will not get namespaced.

QUOTE(newtonvfx)
I'm starting to come around to Houdini's way of doing things, but maybe that's just my turn of mind.

Sure and that's what it was designed for but for the typical user it goes over way over their head and budget.

QUOTE(newtonvfx)
BTW, I don't think many people use Nuke for motion graphics, it's mostly set up for straight up compositing, but what do I know, as you might have noticed from my ID, I work in visual effects, so you're right I'm not that familiar with motion graphics.

You know what they say about assumptions.. imo all node based compositing systems are terrible for motion graphics but I have seen people use nuke for it but after effects is the dominate in that area.

[newtonvfx]

Doh!!

Shame on me! I don't know where I got the idea that you can't evaluate MEL in an expression. I be having a Paris Hilton day! I didn't even try it!

Yup...

float $t = frame - 1;
pSphere2.translateX = getAttr -t $t pSphere1.tx;

...will do it.

But I went for Joojaa suggestion it's working well. No scripting needed only two simple expression.

I'll try to remember that for future though.

Thanks gmask

[joojaa]

QUOTE(newtonvfx @ 03/10/09, 06:46 PM) [snapback]302731[/snapback]
I'm starting to come around to Houdini's way of doing things, but maybe that's just my turn of mind.

Well i like houdini, but sometimes its clanky. Dont get me wrong houdini kindof is great but its also sometimes unlimittedly stupid about things. Maya is good at hiding the fact that your using a node based apporach. Houdini users allways talk about this as being somehow unique its not.

Houdini does however have 2 things going for itself, 1 houdini is being developped for the node based approach. Mayas not, it eats into maya a bit, as a result houdini has so much more nodes to use. Not that maya couldn't have but its just not seen as necessary for several incarnations of maya. 2 houdini can glob info in way maya can not, for exaample arrays are just that arrayable all the time and overridable in maya ist often conditional to the nodes you have in houdini its allways accessible. Now the globbing of many shapes into one is actually something maya should try to get working for it because its the only thing maya looses over big time.

But mayas gui is way better for so many reasons at the normal day to day things. Most of the time the amazing things you can do in houdini can be done in maya with a few lines of api code her and there.

By the way its not totally out of the question to evaluete the curve at a offset wiith a custom done without noticeable speed hits.

PS: due to the fact that not all maya users use it as a node based modeller you will find that finding INFO on all theese node tricks is not widely penetrated skillset. Sure riggers now it but most riggers have a very limitted rigging scope.

[newtonvfx]

Yes Joojaa,

I totally agree with you about Maya vs. Houdini. I seem to spend most of my time rummaging around in the undergrowth of the Hypergraph and graphing what I find in the Hypershade window to so I can connect on shader utility nodes. I agree the interface is great for the everyday, but is sure ain't built for node based TD work.

My frustration mostly comes from finding when I want to do something only slightly out of the ordinary, I have to script it.
Anyway, happily I was wrong in this case. But I have found that if I want something to be fast or I want something a little exotic, I have to learn the API. I'm not that much of a C++ programmer and often in production you don't have time to play around with coding.

Anyway, please can explain more about what you meant here...

QUOTE(Joojaa @ 03/10/09, 10:55 AM) [snapback]302744[/snapback]
By the way its not totally out of the question to evaluete the curve at a offset wiith a custom done without noticeable speed hits.

Cheers

BTW, the animated snapshot idea was great and did the job, but it wasn't ideal for my application, as the pivots of each snapshot are still static.

[joojaa]

QUOTE
My frustration mostly comes from finding when I want to do something only slightly out of the ordinary, I have to script it.

Well no often you dont need to, its just that i and most of the people who know what they are doing have evolved past a graphic user interface. Sure i do a lot of thing isn the hypergraph and connection editor, but when i develop stuff its often easier to go build it then delete it build it etc. So at the end of the day its often easier to wirete the recipie down once i get that one step done in the gui.

So what you end up seeing is either 1 of 2 things:

Somebody who doesent know what they are doing, usually they force their wy so they script it

Or you se somebidy who is so attuned to the way it works that doing it with script is as effective as doing it with the gui.

So 95% of anything scripted can be done without scripting, because maya scripting interface can NOT doe anything yo9u cant do in the gui for most parts. Its just that sometimes its a bit convoluted to do certain repetitions in gui And at the end of the day this is exactly the same workflow i do in houduini.

IN FACT: I use maya 15% of the time form a text editor or console WITH NO GRAPHIC USER INTERFACE WHATSOEVER. I also regularily wirite eps files also by hand, sometimes objs and so on. But just because i do this doent mean its because i have to, it is because i CAN sometimes do stuff on lowvele easier because im doing something that nobody ever meant for me to do. Using a ccomputer is about doing what you want not what you can do easily.

Sometimes a written template beats all other things known to man. I mena graphical programming interfaces have been around for ages, but nobody uses them fro some slight reason.
QUOTE
BTW, the animated snapshot idea was great and did the job, but it wasn't ideal for my application, as the pivots of each snapshot are still static.

Well you can fix that easily with a node simple omne node connection to all the snapshots. Also maya can animate the pivot. So you can easily do this atthe snap stage.

[newtonvfx]

Putting the whole Maya, Houdini and scripting debate to one side for a while if I may, Joojaa.

Would you be kind enough to elaborate on what you said here...

QUOTE(Joojaa @ 03/10/09, 12:31 PM) [snapback]302751[/snapback]
Well you can fix that easily with a node simple omne node connection to all the snapshots. Also maya can animate the pivot. So you can easily do this atthe snap stage.

I'm not sure where I can find the appropriate pivot point for each snapshot on any given frame to connect it to each snapshot transform.

Thanks once more

[joojaa]

QUOTE(newtonvfx @ 03/11/09, 10:15 PM) [snapback]302842[/snapback]
I'm not sure where I can find the appropriate pivot point for each snapshot on any given frame to connect it to each snapshot transform.

What exactly do you want to achieve. Its easer than to guid e you to some hypotethical one approach. Usually its easier for me to do the entire thing than just your next step and then the next step. Because you might not be on the optimal solution path. So sometimes the first ting i do would be different than what you did because i have bigger view of the landscape

For example:

QUOTE
What I want is something that will automatically update when I set and edit keyframes on my original object

It is possible to do this a bit differently too whet you do is you connect the second objects to a frameCache node and inject this to each consequtive history isntead of separate anim curve. This way you coudl possibly inject uniquess in the network. THis has sevaral advantages.

[newtonvfx]

So to be more specific then, I'm looking to have exact duplicates of a curve trailing behind it. Copying it's translation and rotation. However, I'd ideally like the duplicates' scales to be independent. I want curves so I can loft them to create a kind of tail trailing behind something in it's wake (It's meant to look a bit like a bright, moving light source photographed on a long exposure).

I would like the scale to be independent so I can animates it to put subtle ripples through it, but it's not essential because I can scale the original curve, but that wouldn't give me the full control I'm looking for.

So you see the animated snapshot worked but the pivot of each snapshot was static. I can of course centre the piviot but it's stills static and it doesn't pick up the rotation either.

I hope that's specific enough.

So that said, your comment...

QUOTE(Joojaa @ 03/13/09, 02:39 AM) [snapback]302937[/snapback]
For example:
It is possible to do this a bit differently too whet you do is you connect the second objects to a frameCache node and inject this to each consequtive history isntead of separate anim curve. This way you coudl possibly inject uniquess in the network. THis has sevaral advantages.

...seems to be an interesting way forward.

Thanks in advance.

[joojaa]

Oh yeah the sword swipe effect, ufo trail etc.. oldest trick in the book. Use animated sweep, with softbody dynamics.

Well thing is theres several way s of doing this, i would invariably use some combination of dynamics for this. Sure youll get some serious ssues scrubbing things but you just cant get everything!

do what your after, its a bit unoptimal. As is but dont scale the trasorm scale the components this will converge to the eselectin center with possibility to offset that (use move component tool)

you can do this fully in dynamics, the problem is that youd end up sacrafcing ease of scrubbing for ease of implementation

You could simply use framecahce on x individual nodes

YOu could alter thinking in terms of scale ad instead think in terms of normal, this has a tendency to do more neturally what you think scale does this oone is probably the most versatile of them all. But it kindof perverst the jmission you set. BUt the effect is better.

Use the snapshot as joints

Use path a wrap, made form the sweep, his gives you much finer controll over the actual shape of the effects decay

Use history.

Path Anim

You could make a node to do this