Automagical Head Turns
Moderators: Víctor Paredes, Belgarath, slowtiger
Automagical Head Turns
I see that Mike is back (somewhat) so I'm hoping he will at least read some of the threads here and consider some things for the next release. While bug fixes (are there really any major ones?) and logical enhancements have been mentioned, I thought I would blue sky something (hey, you can't get what you don't ask for, right?).
Doesn't everyone agree here that head turns (followed closely by body turns) are the holy grail of AS? I mean, I can count no less than a half a dozen different methods that have been posted, and many people are still trying to come up with new ways to accomplish this.
While some of the methods are ingenuous and even work rather well, they all share one thing in common -- they are a PITA to setup and all out of proportion to the way most of the tools in AS work (which are fairly easy and straightforward). I mean, truly, can ANY of you describe in a few sentences how to do them? You need diagrams and flow charts, and heaven forbid if you miss a step. Some of the bone setups are the most complicated things I've ever seen -- and trying to do those on top of an already complicated rig is just insane.
So I'm asking for an automagical process -- hey, I don't even have a clear idea how it would work. That's up to Mike, the magical programmer. I just want a spell I can cast and -- presto! a head turn. Okay, to be more specific, I want to be able to draw a front and 3/4 view and have the program handle all the rest of the work between those two.
You can *kind* of do this by drawing the same vector points in a switch layer, but it's a huge PITA (moreso than perhaps even the bone rig). Since the various bone rigs can do the math properly, perhaps all I want is this "facial rig" to be done with one button press or some such. Or maybe I want a switch layer that interpolates regardless of the number of points (it could average points on a shape -- I can see the math even if can't express it well, but it should be able to work that way at "guessing" what points should fall correspondingly to others). But whatever magic is done it needs to be simple enough that anyone can sit down and work with the two head views and get the turn working within a few minutes.
If AS had quick and easy head turns I suspect it would rule the world.
Doesn't everyone agree here that head turns (followed closely by body turns) are the holy grail of AS? I mean, I can count no less than a half a dozen different methods that have been posted, and many people are still trying to come up with new ways to accomplish this.
While some of the methods are ingenuous and even work rather well, they all share one thing in common -- they are a PITA to setup and all out of proportion to the way most of the tools in AS work (which are fairly easy and straightforward). I mean, truly, can ANY of you describe in a few sentences how to do them? You need diagrams and flow charts, and heaven forbid if you miss a step. Some of the bone setups are the most complicated things I've ever seen -- and trying to do those on top of an already complicated rig is just insane.
So I'm asking for an automagical process -- hey, I don't even have a clear idea how it would work. That's up to Mike, the magical programmer. I just want a spell I can cast and -- presto! a head turn. Okay, to be more specific, I want to be able to draw a front and 3/4 view and have the program handle all the rest of the work between those two.
You can *kind* of do this by drawing the same vector points in a switch layer, but it's a huge PITA (moreso than perhaps even the bone rig). Since the various bone rigs can do the math properly, perhaps all I want is this "facial rig" to be done with one button press or some such. Or maybe I want a switch layer that interpolates regardless of the number of points (it could average points on a shape -- I can see the math even if can't express it well, but it should be able to work that way at "guessing" what points should fall correspondingly to others). But whatever magic is done it needs to be simple enough that anyone can sit down and work with the two head views and get the turn working within a few minutes.
If AS had quick and easy head turns I suspect it would rule the world.
Re: Automagical Head Turns
This is not a shortcoming of the software.mkelley wrote:I want to be able to draw a front and 3/4 view and have the program handle all the rest of the work between those two.
This is a shortcoming of information theory.
The front view and the 3/4 view are projections of a 3D object onto a 2D surface.
In the instant the projection is cast, however, you are discarding the complicated 3D surfaces of cheekbones, eyelids, nostrils, and so on.
These are unique to the individual.*
That is, in fact, precisely why these subtle details matter so much to us as social animals.
The result is that there is no universal algorithm which will interpolate the rotating curves of both a stocky man's jaw and a slender girl's jaw.
* Actually, they are not unique, because so many people strongly resemble others. It would be fascinating to learn just how many human facial "types" there are. But it's bound to be in the hundreds, at least.
Yes, I understand all that (I've been rendering in 3D for a couple of decades now) but a computer CAN make interpolations between a 3/4 view and a front view even if those are not 100% accurate. After all, that's all the bone rigs are describing -- just a *possible* deformation of the face that does not take into account all the different possibilities. But that one possible deformation is good enough for a whole lot of people.
One of the nice things about 2D as opposed to my far more comfortable 3D is that people will accept things that are not physically correct. A front view face can have a nose laying on its side and still look reasonable. A person can walk in ways that no human could actually do so and we accept it. We probably (most of us) don't WANT a physically accurate model anyway -- these are freaking cartoons for goodness sake!
The more I think about the more easily I think Mike could add this feature (or perhaps a set of features -- he could add the bone rig as well as the interpolation between views).
One of the nice things about 2D as opposed to my far more comfortable 3D is that people will accept things that are not physically correct. A front view face can have a nose laying on its side and still look reasonable. A person can walk in ways that no human could actually do so and we accept it. We probably (most of us) don't WANT a physically accurate model anyway -- these are freaking cartoons for goodness sake!
The more I think about the more easily I think Mike could add this feature (or perhaps a set of features -- he could add the bone rig as well as the interpolation between views).
Not exactly the same as your feature request but it would lead to solutions in that area...
Rather than a "specific" application feature to do head turns, I have been pushing for ADDING to existing features (bone constraints, scripting) so these solutions can be built more easily by pro users like myself.
Instead of poor overworked Mike creating a "head turn tool", if he added in extra bone constraint features and/or more access to the application through scripting then I or someone else could create a head turn tool.
Just a few bone constraint modifications would kick some arse baby and it would eliminate the need for scripting to achieve the same results.
Just three new constraint features to start would be great:
Translation limits
Translation constraint includes rotation of target bone
Multiple constraint targets (more than 1 for each... at least 3)
These by themselves would be fantastic for doing head turns. It would be like an explosion of incredible possibilities.
I have a very very very complex bone rig that doesn't even need any view but the front view. The resulting turns are very realistic. With just those 3 new bone constraint features it would make it so much easier to do... like 1000% easier.
I would like to figure out a way to put the bone placement in an automated menu script the same way Genete did with his 3D vector scripts.
So you see, by giving us some "simple" new features we can create new features that no one would have thought of.
-vern
Rather than a "specific" application feature to do head turns, I have been pushing for ADDING to existing features (bone constraints, scripting) so these solutions can be built more easily by pro users like myself.
Instead of poor overworked Mike creating a "head turn tool", if he added in extra bone constraint features and/or more access to the application through scripting then I or someone else could create a head turn tool.
Just a few bone constraint modifications would kick some arse baby and it would eliminate the need for scripting to achieve the same results.
Just three new constraint features to start would be great:
Translation limits
Translation constraint includes rotation of target bone
Multiple constraint targets (more than 1 for each... at least 3)
These by themselves would be fantastic for doing head turns. It would be like an explosion of incredible possibilities.
I have a very very very complex bone rig that doesn't even need any view but the front view. The resulting turns are very realistic. With just those 3 new bone constraint features it would make it so much easier to do... like 1000% easier.
I would like to figure out a way to put the bone placement in an automated menu script the same way Genete did with his 3D vector scripts.
So you see, by giving us some "simple" new features we can create new features that no one would have thought of.
-vern
But you see, that's *exactly* what worries me.heyvern wrote:I have a very very very complex bone rig
I agree with you that Mike should add on to existing features, particularly if they seem relatively easy (as at least a couple of your ideas seem to me). But if all this buys me is the abiliy to make a very very very complex bone rig... thanks but no thanks.
Even if you could make a script to do it for me, what are the odds that such an insanely complicated bone rig would work properly all the time? And something that complicated is *bound* to have issues with other features.
No, I still think that something automagically is called for in the case of head turns -- once again, the holy grail of 2D animation. It needs to be simple and workable even by an idiot as myself. Then you couldn't stop people from buying AS if you wanted to.
I think asking for stuff like, automatic head turns, is kind of silly. You already have the tools to do it. Just take the time to learn how. Such as "How To..." books or video tutorials on the net. This isn't that hard to do at all, just takes more time for the impatient person.
If we keep asking for request like this, then we'll end up with a tool that does everything for us. Where's the talent in that?
If we keep asking for request like this, then we'll end up with a tool that does everything for us. Where's the talent in that?
You missed my point. It's complicated NOW but if I had those new features it wouldn't be.mkelley wrote:But if all this buys me is the abiliy to make a very very very complex bone rig... thanks but no thanks.heyvern wrote:I have a very very very complex bone rig
For instance, one bone with one translation constraint. If I need more I need 2 or even 3 bones in the same spot sharing influence with constraints to 3 other bones... it starts to get "very complex". If one bone could have multiple targets... it would be a very simple set up and not hard to understand or change. It would be a dream... I would dance the happy dance.
An "automatic" head turn feature will require a ton of thought and work from a developer's point of view. Many of us have done most of that work with our own custom solutions. If we had the tools in the program... simple features like better scripting and extra bone constraints... then we can make the instructions to set them up much easier to understand.
Mike has already told me that I'm doing stuff with AS scripting that he never planned on. He developed it but he couldn't forsee how it will be used. That is the beauty of simple powerful tools like bone constraints and scripting. It is like an erector set or legos. Do you think the original inventor of legos expected someone to build an Eiffel tower replica with them?
Any solution for head turns will have its own complexities to deal with. Even a solution created by the developer will have to be "learned". Plus if it is created by the developer only he can fix it or add to it when he gets around to it or when the distributor lets him. PLUS they won't be putting out updates as often as we'd like.
If others could create better tools and features based on expanding existing features, these could be updated much quicker. We would be in control. "Scratch your own itch."
-vern
ideas on how a script could work
I am no programmer, but I've thought about this one. It kinda goes in line with shape recognition (ai-ish), but here's how it "could" work...
you have a front view of the face.
the script would ask you to highlight either one facial feature or multiple features at a time.
It would then calculate the X scaling and X position of each feature and presto! automagical head turn.
That's basically it (I know, I pretend that would be easy, but it's not impossible.)
If the computer couldn't handle that you could dumb it down more-
divide your front view face into 3 sections- left, middle, right (this is how I do my head turns for lighting/shading purposes). You'd have to divide the nose and mouth into at least 2 slices (or 3).
THIS REQUIRES MUCH SHAPE ORDERING!
This way, at a 3/4 turn the middle shapes cover the left shapes, and then when completely turned the far right shapes cover the middle and left shapes.
Maybe the script would know to do this- divide your face into 3 groups of shapes, background, middle, foreground, then do the x scaling and positioning of lines/shapes.
It is sort of a mathematical process correct? Scripts love math!
you have a front view of the face.
the script would ask you to highlight either one facial feature or multiple features at a time.
It would then calculate the X scaling and X position of each feature and presto! automagical head turn.
That's basically it (I know, I pretend that would be easy, but it's not impossible.)
If the computer couldn't handle that you could dumb it down more-
divide your front view face into 3 sections- left, middle, right (this is how I do my head turns for lighting/shading purposes). You'd have to divide the nose and mouth into at least 2 slices (or 3).
THIS REQUIRES MUCH SHAPE ORDERING!
This way, at a 3/4 turn the middle shapes cover the left shapes, and then when completely turned the far right shapes cover the middle and left shapes.
Maybe the script would know to do this- divide your face into 3 groups of shapes, background, middle, foreground, then do the x scaling and positioning of lines/shapes.
It is sort of a mathematical process correct? Scripts love math!
One of the reasons I have been doing most of my work so far in 3D is the existence of such a tool in the iClone software I use. It lets you take a front view and a side view picture of a head, and then has a tool to map those onto a head by letting you drag the eyes, nose, and mouth to the appropriate points of the picture you've loaded on the front view, and then the Nose, mouth, ears and chin on the side view. From that it gives you a pretty decent 3D head. Lets you use any decently large JPG or PNG file...
Unfortunately it doesn't have a render to 2D feature, Of course if Anime Studio which does 2D let you do something like this...
Just my .02 worth,
Ewan
Unfortunately it doesn't have a render to 2D feature, Of course if Anime Studio which does 2D let you do something like this...
Just my .02 worth,
Ewan
I think it will only be useful in very specific (photo-realistic) circumstances. Many 2D animation isn't just a flat version of 3D animation, but has its own cartoon physics. Often faces are deformed for sake of clarity. You want to see the mouth shapes when a character is talking in profile view.
Of course, many things can be solved with proper staging. You don't always have to animate a full head turn. With proper staging, you can get away with a "flip". Many TV cartoons do it that way, and I don't hear anyone complaining about bad head turns in TV cartoons.
The darn thing has to be done in time and on budget, so IMO there is no need to overdo something for artistic reasons.
My 1/50-th of a euro.
Of course, many things can be solved with proper staging. You don't always have to animate a full head turn. With proper staging, you can get away with a "flip". Many TV cartoons do it that way, and I don't hear anyone complaining about bad head turns in TV cartoons.
The darn thing has to be done in time and on budget, so IMO there is no need to overdo something for artistic reasons.
My 1/50-th of a euro.