What could be more simple than making a character blink?
Animating blinks are a basic part of any living character. Blinks occur in most scenes, yet it’s a place where beginning animators, and occasionally vets, screw up. For today we’ll be looking at vanilla blinks, the spontaneous blinks that occur spontaneously and regularly. Later we’ll look at blinks triggered by external stimuli and special cases. First, how often do we blink?
The short answer, for adults, is 4-16 times per minute (different people tend to have different baseline rates), with about 10 blinks/minute being average. Babies and young children blink much less, about 1-2 times a minute. How constant is that frequency? To answer that, we need to understand what affects blink frequency.
The primary function of blinking is to keep the cornea lubricated with tear fluid, and also to clear dust from the eye. It might stand to reason that normal blinks are triggered by dryness or dustiness, thereby affecting the baseline blink rate. That doesn’t happen. There’s an internal oscillator that triggers blinks, usually before the eye has a chance to dry out under normal conditions. This oscillator doesn’t significantly adjust for environment. We blink at a fairly constant rate, regardless of whether we’re in a humid, dust-free Turkish bath (where we physiologically need fewer blinks) or walking around the desert (assuming there’s no driving wind blowing sand into the eyes).
However, when we do something that requires concentration, our blink rate drops dramatically, to 3-4 times per minute. The more difficult the task, the less we blink. It doesn’t have to be a visual task — carefully attending to sounds or music likewise lowers the blink rate. Typically, there is a short burst of several rapid blinks just before anticipated difficulty, then a greatly reduced blink-rate during the task. The reduced blinking during sustained attention is why long bouts of animating at the computer can dry out the cornea and eventually damage the eye — so take frequent short breaks!
Now let’s look at the timing of individual blinks. Let’s start with some practical nuts and bolts — how does one animate a ‘standard’ blink? This is pretty much the one bit about eye movements that some animation books actually address, however briefly. Unfortunately, I think these tips don’t quite cut it.
The system I was taught (in hand-drawn animation) was to close the eye in two frames, keep it closed for two frames, and open it over two frames, with a key element being that the inbetween in the closing phase should be a one-third favor of the open position, and the inbetween on the opening a one-third favor of the closed position. That yields a 6-frame, or 250 millisecond, blink, and tends to look fine in 2-d animation.
Varying the two inbetweens serves two functions — it looks more naturalistic, and it avoids the strobing that can occur when the inbetweens going into and out of an extreme pose are identical (it’s the same reason, when animating a bouncing ball, one avoids having the ball at the same height off the ground in the frames before and after the contact frame).
Other’s have taught variations of the above template, as mentioned in Richard Williams’s book. Should we stick with these simple templates when doing CG?
No, I don’t think we should — we have to go beyond what worked in hand-drawn animation. CG animation gives us the opportunity to do facial animation with far more subtlety than was possible in hand-drawn animation. Having spent plenty of hours honing a 0.3 mm B-lead mechanical pencil to surgical sharpness for literally each pencil stroke during the clean-up drawing of a face, I can tell you we were fairly limited in the nuance we could put into things like eye lids and blinks.
But more than having the ability in CG to animate finer detail, I think we’re required to do just that. The detail and texture that goes with CG animation gives us a much lower threshold for dead-looking facial animation. We HAVE to get more nuance into CG faces than hand-drawn faces, just to get the same pleasing result achieved with a few pencil strokes and some flat coloring, otherwise we start to fall into the uncanny valley.
Let’s do a little research. Here’s a set of clips of several typical looking blinks, from the BBC Gallery website:
These are all the European-standard video frame rate — 25 fps, which is close enough to the feature film standard of 24 fps for comparison. Note how many frames it takes for upper lid to close, how long it stays closed, and how long it takes to open.
According to one physiological study, a typical spontaneous blink last 334 ± 67 ms, or right about 8 frames, give or take a couple of frames. The closing phase gets 92 ± 17 ms and the opening phase 242 ± 55 ms. Note that the opening phase lasts almost three times as long. The average is therefore 2 frames to close, and 6 frames to open, with only a single frame at the closed position. (Voluntary eye blinks tend to be slightly faster, around 275 ms, if we want to get picky.)
Now let’s look at the clip above showing a variety of blinks. I’ve averaged these and others I found on the BBS site, and while the video data is similar to what I found in published physiology studios, the blinks tend to be a little longer. The average is about 2-3 frames to close, full closure for a single frame, and 6-14 more frames to full open. Total blink duration, at 25 fps, turned out to range between 8-18 frames. Note how in the closing there is a definite slow-out from open (with the first frame of movement only being the slightest contraction of the lid complex), but little to no slow-in to closure. The opening, however, tends to have both a definite slow-out and a slow-in, so that the eye is mostly open within about 6-7 frames but continues to open for another quarter of a second.
I think the difference between these videos and the published data is that the close-up videos show a very subtle but definite slow-in to the final lid position after the lid is mostly open, and that this final tiny movement is likely disregarded as insignificant by physiologists (in these studies, blinks are usually measured by the electrical field generated by muscle activity, not my measuring actual displacement, so watching high speed video is probably slightly more accurate).
Note also that there is no ‘compression’ on the closure (the lid closing for one frame, then closing even tighter for a second frame). Also, we see that there is no ‘overshoot’ at the end of a blink — it’s a straight slow in. I’ve seen animators advocate both ‘compression’ and ‘overshoot’ in a standard animated blink, but it’s always looked unnatural to me, and in these clips we see they are unnatural. (Now, if you’re animating someone scrunching up their face, or blinking into a surprised expression, we’re talking about something else).
Note also that virtually all the blink comes from the upper lid. The lower lid moves, as we see by the blurring in the close-up videos, but they don’t really move upwards to meet the upper lid. When the lids close, four muscles are involved, and the generate both contraction and rotation. It is this slight rotation that ‘tweaks’ the lower lid, pulling it slightly towards the nose, making it blur in the videos. But the lower lids are mostly staying put.
That said, since we don’t usually have controls to animate that slight pulling inwards of the lower lid, it’s useful to animate a slight upward motion of the lower lid, so that part of the eye-lid complex isn’t inert and frozen looking.
Physiological studies also tell us that blinks are faster when our gaze is directed upwards, and slower when our gaze is downwards. This is not intuitive, since the upper lid has much farther to travel before it’s closed when the eye is in an upward position, and a much shorter distance when the eye is gazing downwards. Look closely at the following clip, which is consistent with this data — this is about the fastest blink I found on the BBC site:
Physiology studies, and the above clips, also show that human blinks are extremely conjugate. That is, blinks tend to happen very much at the same time. Disconjugate blinks are not usual. That said, it’s useful to add slight variation between the two lids in the closure and opening — not so much that it’s obvious, but just enough to feel natural.