Inbetween automation solutions and ideas

[OwenWelsh]

One of the greatest production bottlenecks is still inbetweening full traditional animation (frame by frame). This can be especially true for subtle animation where the character doesn't change very much, like moving holds. Has TVPaint done any research on automating the inbetweening process? Even if the inbetween isn't completely accurate, we could use it as reference and trace over it keeping the good information and throwing away the bad. Simple interpolation might even help.

The other idea I had was for more action oriented animation where the character would change perspective in a major way infront of the camera. This would require a 3D reference model of your character that could be matched to the extreme drawings. Then, perhaps with some user control, we could control the spacing of the inbetween.

To create the 3d reference model, perhaps TVPaint could look at your model sheet and build a simple 3d model based on the silhouete of your drawings and line art (perhaps we would have to shade our characters in grayscale with shadows for extra information). But then TVPaint would need a pretty good toonshader to accurately render the automated inbetweens.

There is a company trying to do this called Cacani, but it relies on too much vector information needed by the user, and for people not working in vector is kinda useless.

Just like to get the discussion rolling if it hasn't been already.

[slowtiger]

Oh, is it that time of the year already?

I'm afraid this discussion already has taken place. Ever once in a while someone pops up with the brilliant new idea to automate inbetweens, even presents some software hack, then, after some unsatisfying tests, vanishes into oblivion again.

The solutions are out there, they're called vector animation and 3D. TVP is not the place and framework to include either of them.

(Short answer because I'm really tired.)

[habashtvpaint]

I second slowtiger on his thoughts (but with less anger )...
The beauty of TVpaint is the honesty to traditional animation approach and simplicity of use.
The application has hundreds of features that suits various animators in their workflow, thus such debates occur from time to time. Personally I would strongly prefer that TVPaint stays honest and simple... cause once it integrates 3D, vector, tweening, automated animation, etc, it will become like the long list of applications out there trying to create animation with one click. (nothing wrong with that of course from a fast-paced digital industry perspective) but ... think of TVpaint as the -missed but founded-link between technology and this atmosphere inside an animation studio from the 20's - 60's era.
In addition, there is nothing wrong with using 3D references for TVPaint, Professional animators always been acting and using reference footage for animation through rotoscope or simple guideline! I'm not metioning this to debate wither this is cheating or not. Like I said earlier, artists have different workflows to achieve their results! if you check this recently:

you'll see how Juanjo Guarnido and the team of awesome artists, used 3D as reference for TVPaint. just import the footage and start drawing on a new layer. I believe adding a 3D platform inside TVPaint isn't a solution but merely unimportant feature to consider.
still a one humble opinion, maybe a better statistic study and poll can find what the majority of TVpainters want.

[Fabrice]

dear forum veterans, please keep this topic a nice place

One of the greatest production bottlenecks is still inbetweening full traditional animation (frame by frame). This can be especially true for subtle animation where the character doesn't change very much, like moving holds. Has TVPaint done any research on automating the inbetweening process? Even if the inbetween isn't completely accurate, we could use it as reference and trace over it keeping the good information and throwing away the bad. Simple interpolation might even help.

Something you could check is the FX Rendering > Rototracking :
http://www.tvpaint.com/doc/tvp11/index. ... ng&lang=en" onclick="window.open(this.href);return false;
Or the FX Distorsion > Wrapping grid :
http://www.tvpaint.com/doc/tvp11/index. ... n-wrapping" onclick="window.open(this.href);return false;
Even the new warp tool might help :
http://www.tvpaint.com/doc/tvp11/index. ... sform-warp" onclick="window.open(this.href);return false;

But it's far from being an automatic inbetweening process.
We have some ideas in mind for inbetweening simple movements in the future, but nothing replace the human hand (and brain) behind this process.

EDIT : we will add the .tvpp project files in the manual soon.

[Peter Wassink]

Here Owen, maybe this is something for you:

"Just because its easy doesn't mean that its simple"

[schwarzgrau]

I work a lot with reference videos and images, if the shot got a really complex perspective I sometimes create even simple 3D reference stuff in Blender, but I don't want this to be a part of my 2D animation software. The software you use to draw and animate (in this case TVPaint) should be lightweight and responsitive and not a bloated battleship like Photoshop or even worser.

My problem with the automated inbetweens is always the placement: If you create your whole animation in a tween-based environment like Maya, After Effects etc. you can edit the speed-graphs, but if you draw on twos and want the ones to be interpolated you will get the inbetweens exactly in the center of your two keys, but that's usually the wrong position.

[Paul Fierlinger]

I don't mind drawing, even when I have to draw more.

[OwenWelsh]

Thank you to everyone who responded with helpful tools, suggestions and ideas. Technology is now increasing at a rapid rate and I don't see any reason why it can't be used to help us in our traditional workflows.

My interest is in creating full traditional animation like that of Disney's Bambi or Pinocchio. There are already solutions for vector animation like Flash, or puppet animation, so lets try and focus on the traditional paper/pencil/inked look.

I really like Freak of the Weeks approach of 3d reference. However sometimes we rely too much of this reference and need to caricature it further to make better use of the traditional medium. That being said I think its a great idea and their animation is terrific. One of my ideas was using photogrammetry. http://en.wikipedia.org/wiki/Photogrammetry" onclick="window.open(this.href);return false;

This would aid the traditional artist in creating 3d models of their characters or objects very quickly based on a model sheet with lighting information. Then using this very rough 3D model as reference in our TVP workspace as an underdrawing to draw over. The trick lies in how do we get the reference into the proper pose to match our extreme poses. If the 3d model matched the poses of your extremes and offered you a little slider so you could control the spacing for your inbetween, then you have a very nice reference in the correct pose and perspective as your underdrawing. This in no way automates the inbetween process, but it speeds up the work flow considerably as many of the perspective problems are solved / aided.

In terms of having TVP actually rendering a final line drawing for your inbetween I think that technology is very far off. But there could be some stepping stones to get us creating content faster with complete control over our drawings by hand.

I feel that the inbetweens have always been a major hurdle in the traditional animation process if we can find ways to speed up this tedious task with the same results, why not?

[Paul Fierlinger]

I feel that the inbetweens have always been a major hurdle in the traditional animation process if we can find ways to speed up this tedious task with the same results, why not?

You are over thinking the process. If you want to get away from the act of drawing find something else to do on the technical side of life but don't confuse art with the pursuit of a riddle in search of a problem. It irritates me to no end to have someone telling me that what I've been practicing my entire life has been a waste of time since there's an algorithm just around the corner which would liberate me from my drudgery if only I would put my pencil down and pick up a calculator. Your premise that inbetweening is the animator's bane is completely off base.

[Fabrice]

Have you tried the new option "Guide > Image" ?

It might help to keep the exact proportions of a character.

[Elodie]

Inbetweens are tedious, but they are necessary since they give the rhythm - and so, the life - to your animation.

Key images are just pretty drawings if they don't have good inbetweens

[slowtiger]

I feel that the inbetweens have always been a major hurdle in the traditional animation process if we can find ways to speed up this tedious task with the same results, why not?

On the contrary: the most difficult part is drawing meaningful poses. Inbetweening is easy. You can inbetween whatever you have, it will become smooth, but if the poses are bad so will be your animation.

[Paul Fierlinger]

Ever since the Project tab evolved within TVP I have found myself moving away from so called key poses. My first sketches are composition layouts within which I occupy myself more with the distribution of weight and negative space than the nuances of dramatic poses. I stay away from those messy, repetitious swirls of thick lines that can grind a hole into the screen and make anyone lose sight of the rest of the story.

I begin inbetweening as soon as I have my first clips with their preliminary sketches lined up to my satisfaction. I work on all of these at the same time, jumping back and forth between layouts, tightening spaces of action with light sketches that suggest motion and acting. I need to see my story in motion before I will care about what a single drawing looks like.

I will use the playback button long before I manually make a single flip between two drawings because I want to see how my clips interact with one another rather than how accurately my individual drawings look. They are not illustrations. Only after having enough of these clips going can I see an actual play emerge within their stages and I might have all my drawings sketched out in two's before I start tightening some details in a few images. Eventually I will redraw every frame as I add a new drawing between each of them.

Technology is now increasing at a rapid rate and I don't see any reason why it can't be used to help us in our traditional workflows.

I agree and this is how that thought has has already worked itself into TVPaint's GUI.

[schwarzgrau]

The good thing about something like the 3D guides you talking about is that they are absolutely accurate, but that's also their downside. For example I noticed if I use TVPaints Out-Of-Pegs function too much my animations get absolutely mechanical. I don't want to say you need to overthink every inbetween, but if you got a perfect guide for every frame you will propably stop thinking and loose the looseness of it.
It's like comparing pose-to-pose animation with straight-forward animation. Pose-to-pose is great for staging a shot and keeping the characters constant etc. but you're loosing the life of it really fast.
That's the reason why a film like paperman doesn't work for me (apart from the story). It misses the tiny bit of imperfection even the last hand-drawn Disney features had.

[momo]

It misses the tiny bit of imperfection even the last hand-drawn Disney features had.

+1 Totally that's the same issue I have with a lot of 3D productions, the "imperfection" human skills gives life to animation. It is a very similar feeling when comparing a hand man pot and a machine made pot. Most of time the machine made pot has no life at all.

Animation is to some degree going through a similar process as pottery. Machine are slowly taking over to make the production speed quicker.