Crow flock test V1

Following on from techniques explored over the past couple of weeks, I put together a test of the flock of crows from the opening shot. It should serve to give us an idea of whether the timing is working, the angle of the birds and the pattern of their flight. It also gives us an idea of how we might go about constructing the rest of the puppets for the animation and any potential issues we might encounter!
 
I was considering how best to approach the construction of the puppet — creating separate wing shapes as experimented with previously would look really nice if pulled off correctly but it might just be too complicated and fiddly to animate convincingly. I came across this video on Youtube of a bird somebody animated in After Effects, using one solid shape for each wing and rotated on a 3D axis to simulate the beating of the wings, very similar to what I did with my first bird test:

It's not terribly realistic, but it got me thinking and I decided to return to a similar technique. Instead of using one solid mass, however, I thought I might try breaking the wing into two separate components — much like we did in Digital Skills, with separate layers for each joint. This would give a degree of flexibility when animating the wings,  allowing me to break and freely rotate the joint to give some delay and overlap to each section.

I drew out a simple wing shape in Photoshop and split it in half, each section on a separate layer, with the connecting sections rounded out to allow for neat rotation at each joint. Then imported it into After Effects as a layered composition and created a hierarchy of parent/child relationships — the wing tip a child of the wing base, the head a child of the body, etc. This meant that if the body or base of a wing was moved, the head or tip would follow, but I would still be able to rotate the tip and head independently.

Once everything was parented up correctly I moved the anchor point of each limb so that it would rotate from the joint rather than the centre. I then just keyframed a basic up and down motion of the base of each wing, giving me a very static but distinct 'flapping' motion.

You can already start to see problems with the puppet at this point. The base of the wing sticks out when the tip is rotated, and the base has a really sharp edge that sticks out of the body when the wing rotates downwards. All totally fixable, though!

To loosen things up a bit I then added a little delay/overlap to the tips of the wings, so that they would drag behind as the base of the wing travelled down, and continued forward when the rest of the wing was being pulled up. I also added a little rotation to the body and head, as well as a slight jerk to the body caused by the beginning of the flap, the force of the movement pushing the body upwards.

I also turned on the motion blur switch at this point, just to smooth things out a little.

I did also have a little poke around in After Effects' curve/graph editor, applying what I'd learned about function curves in Digital Skills 2, just to give a slight overshoot to the wings as they reach the top of the movement. It was far trickier than I thought. The curves don't behave entirely as you'd expect (sometimes, when you drag slope handles, absolutely nothing happens!)

 I don't know if I'm just not used to it or if it's genuinely less intuitive. It has a separate graph for speed and value of keyframes which was massively confusing and just seems needlessly complicated. The top graph is the Value curve, and the bottom is the speed graph. If you look closely towards the last keyframe on the top graph, you can see that the curve extends very slightly above the keyframe. It's this that gives it the overshoot — the wing travels a little higher, then slowly comes back into the pose defined on the keyframe.

The speed graph is a little trickier to read but it's the same principle. I'm not entirely comfortable with how it works — I need to read into it a little more — but basically, the vertical axis represents an increase or decrease in speed (represented by high and low points on the curve, respectively) and the horizontal axis represents the 'influence,' controlled by bezier handles. The 'influence' basically controls  how quickly the object reaches the speed indicated on the graph — much like Softimage, the wider the curve, the slower the acceleration.

To reach the graph editor (if you dare!) simply click on the parameter whose speed you want to alter and click on the weird graph-y icon thing next to the giant stopwatch/keyframe button. The graph editor will open up in the timeline panel, displaying either the speed or value graphs.

To change which graph you're looking at, just right click and choose 'Edit Speed' or 'Edit Value,' depending on which you want to alter.

Anywho! As a result of all that tinkering, this is what I eventually ended up with:

 

It looks just as bad as my first attempt did at this point, but experience has taught me that once it's actually composited and put together it (hopefully!) will look much better. Fingers crossed.

It looks too quick to me — I think that the wings should probably delay slightly at the lowest point before beginning the upstroke and vice versa. The body movement is off as well. It looks like it comes up too soon, before the wings have even started beating, so that needs to be shifted a bit — probably just as the wings come down, or after they hit the mid-point? The motion blur does seem to disguise the dodgy rotation sticky-out bits, though. To some extent.

Being just a test, I didn't go all-out trying to make it perfect and gorgeous, so I just dropped it straight into a new composition.  I used the same techniques to put it all together as I did before, assembling multiple birds into one composition and adjusting their start positions so that they didn't all flap in perfect sync. Using the animatic timing as reference, I simply keyframed the scale and position parameters to fly them across the screen and towards the camera. I duplicated the flock layer a couple of times and slightly shifted the start/end points to create a larger flock.

One of the flocks originates from the wrong point and I think they need to spend longer on the horizon at the beginning of their approach. It's quite a distance they're covering so, realistically, they need to spend a bit more time as mere dots, gradually getting bigger and then increasing in speed as they draw nearer.

For the purposes of a test, though, I think it's alright. It's given me some ideas, highlighted some issues and should hopefully aid us in the composition of the rest of the shot!