Automated inbetweening

[sykosan]

...

[slowtiger]

Any 3D software could do this, any 2D vector animation software could do this. If I'd need that feature, I'd use one of those. I can't see how this could be incorporated into a bitmap software like TVP in an elegant and working way.

In order to do this, the software must be able to know the identity of strokes troughout the movement (that's what RETAS does, each stroke is a vector line), or the user must be able to define matching ares which will be dissolved like in a morph program. I haven't heard of any other way in the last 20 years. So which of those is your approach?

you need to make the drawings in a particular way

This is the part where you lose me. What I like about TVP is that it doesn't force me to draw in a certain way. If I want to be restricted, I choose another program.

[sykosan]

...

[Sewie]

Interesting.

[Paul Fierlinger]

Not very interesting (for the record )

[idragosani]

Any kind of automated tweening or morphing still looks 'computer generated' to my eyes and doesn't have the same spark of life and organic feel as fully hand-drawn tweens do. This goes for the examples above, too. They still look CG

[Paul Fierlinger]

Some might think anything that makes 2D work should be part of TVPaint, and some, me included, will believe that TVPaint should concentrate only on those features and techniques that are considered hand drawn and hand animated.

My fear is that if this software would start moving into the gray area of 2 1/2 D, the developers will be spending less time on bringing to us all the features that fall strictly under 2D animation and that we yet don't have.

[idragosani]

Some might think anything that makes 2D work should be part of TVPaint, and some, me included, will believe that TVPaint should concentrate only on those features and techniques that are considered hand drawn and hand animated.

Me, too. To quote Dennis Ritchie "If you want X, you know where to find it", where X = automated tweening.

[Paul Fierlinger]

Or to quote Terry Gilliam, "if you want smooth, go to the ballet, and who wants to do that?"

[neonnoodle]

Obviously I don't want to see any abandonment of 2D features in TVP, and it's the best and fullest 2D animation program in existence right now, no contest.

But I'm also intrigued by the prospect of automating some parts of the process, and it would be nice to have the choice to do so in appropriate situations. I'm not one to reject a potential feature on the assumption that somehow it means losing another feature.

[sykosan]

...

[Sewie]

Any kind of automated tweening or morphing still looks 'computer generated' to my eyes and doesn't have the same spark of life and organic feel as fully hand-drawn tweens do. This goes for the examples above, too. They still look CG

For that matter I have seen a lot of hand drawn animation that lack any spark of life and and organic feel. So there goes your argument.

I don't want to reject any new input before I can see (or preferably experience)for myself what can be done with it. And I am very curious about where this can go.

[Elodie]

Hello Sykosan,

Thank you for taking the time to subscribe the forum and submit your idea.
Sykosan contacted us in first in order to propose us its technique. I encouraged him to post on the forum and collect different opinions.

Sykosan : how have you been using your technique ?
Did you use it for bigger projects, like short movies ?
Do you think your technique could be used in a studio, with hundred of person ? I mean, is there a way to be sure people will interpolate in the same way and obtain an homogeneous result ?

I would like to understand in what your technique is different from this one :

http://youtu.be/Gj2AUsPx1OE?t=15s" onclick="window.open(this.href);return false;
(NB : Ankama mixes traditional animation on paper and inbetweens on Flash)

[slowtiger]

My fear is that if this software would start moving into the gray area of 2 1/2 D, the developers will be spending less time on bringing to us all the features that fall strictly under 2D animation and that we yet don't have.

I'm with Paul here, especially since I see that Anime Studio went down this way quite a bit. (To be correct: they included lots of stuff to make animation look easier to newbies, while neglecting to improve the user interface or even worsen some aspects.)

Still I'm not so convinced about the "new and revolutionary" technique. I've seen lots of solutions over the years, and recently much stuff from researchers, so I'm quite sure that it's nothing new to me.

The whole problem boils down to this set of tasks:
1. you need to separate elements in a way the program understands, so it will get the identity of a line (or whatever) from beginning to end of movement,
2. you need to apply any kind of mesh over each element, so the program knows how to transform it during motion,
3. you need to implement any kind of 3D ordering in arranging your elments, so the program will be able to do X and Y turns in a more convincing way.
Add to this a bunch of controls over the inbetweening process.

There is no solution without these. Please explain and show your idea, because I will not believe any promises without proof.

I could, however, imagine a software which combines a 3D framework, a skeleton, discreet elements (bitmaps/strokes), "symbols" (like in Flash), a very complex X-sheet system, and all the drawing abilities of TVP in one. But that would be a totally different kind of program.

I see that there's a kind of general movement in the field towards this "universal" functionality. There's CGI software where you can paint directly onto the objects. There's vector software which does a great job to convincingly mimick natural media. And there's TVP which concentrates on a strict bitmap workflow, thus avoiding that "object lifetime" nuisance which bothers me and others a lot in the aforementioned programs.

[Fabrice]

The whole problem boils down to this set of tasks:
1. you need to separate elements in a way the program understands, so it will get the identity of a line (or whatever) from beginning to end of movement,
2. you need to apply any kind of mesh over each element, so the program knows how to transform it during motion,
3. you need to implement any kind of 3D ordering in arranging your elments, so the program will be able to do X and Y turns in a more convincing way.
Add to this a bunch of controls over the inbetweening process.

There is no solution without these. Please explain and show your idea, because I will not believe any promises without proof.

Well, I agree with Slowtiger.
I can't imagine any other solution than the process he described.

Maybe I'm wrong, but if so : please show the steps of your technique, or at least the drawings without inbetween, so we can have a better idea.