Video reference is hard to read at full speed. A jump is a blur with a landing at the end, and the three frames where the weight arrives go by in an eighth of a second.
So slow the clip down. Find the poses that carry the action, count the frames between them, and compare those frames with each other. In ArtDeck, all of it happens on the board, next to the rest of the reference for the shot.
Timing and spacing are two different questions
They’re easy to mix up. Get the timing right and the spacing wrong, and the movement still looks dead.
Timing is how many frames an action takes. The same head turn reads as a flinch in six frames and as a decision in eighteen.
Spacing is where the positions fall inside those frames. Two head turns can both last twelve frames and look nothing alike. Spaced evenly, the head moves like a machine part. Bunched at both ends, the same twelve frames ease out of one look and settle into the next.
Even spacing
Bunched at both ends
A clip answers both questions, one at a time: you get the timing by counting frames, and the spacing by comparing them.
The frames worth counting between are the key poses: the extreme an arm reaches before it reverses, the contact where a foot takes the weight, and the breakdown that decides how the body travels between the two.
Slow the clip until you can see the poses
Pick a clip you can actually read: one continuous shot, a camera that holds still, and the whole body in frame. A cut in the middle of the action, or a camera that pans with it, will cost you more time than the clip saves.
If nothing you saved fits, film yourself doing the action and bring the clip in from Photos. That way you choose the framing.
Put the video or GIF on the board and use the controls on the item itself. Video speed cycles through 0.5×, 1×, 1.5×, and 2×, and GIFs go down to 0.25×. Scrub to the right part, then step one frame at a time. The counter next to the transport shows the current frame and the total.
Let a video play through once before you start stepping. A paused video shows a coarser preview than it does in playback, and you’ll misread a pose on a soft frame.
For a GIF, open the filmstrip under the item to see the sequence as thumbnails. Extremes are usually easy to spot even that small, because the silhouette at the top of a jump looks different from the frames on either side. Tap the one you want to jump straight to it.
Video has frame stepping and onion skinning too, but no filmstrip.
Count the frames to get the timing
Step to the first key pose and note the frame number. Step to the next one and subtract. Write the counts in a note on the board while you have them. A note that says “crouch to launch: 3, launch to peak: 7, peak to contact: 6” is more use next week than the clip is.
The counter shows frames, not seconds, and a count only carries over if the clip and your scene run at the same rate. Phone footage is usually 30 or 60 fps and most animation is done at 24, so convert through seconds: fifteen frames at 30 fps is half a second, which is twelve frames at 24.
Twelve frames at 24 is worth remembering. It’s roughly one step of an ordinary walk, and half of Williams’s twenty-four-frame cycle.
Treat the numbers as a starting point, not a target. Animation that matches live action frame for frame looks slack. Disney found this on Snow White: the human characters were traced from footage and still came out as caricatures of the models.
Turn on onion skinning to see the spacing
Onion skinning draws the frames around the current one on top of it, so you see the spacing directly: it’s the gap between the positions of one landmark from frame to frame.
The settings, in panel order:
- Line Art traces the neighbors as contours instead of solid shapes, which helps when the figure is dark or the background is busy.
- Before and After set how many frames get drawn on each side. One or two on each side keeps a contact readable, and more than that turns a whole arc into a single image.
- Step sets how far apart those frames are sampled, in tenths of a second, so you can compare a fast action frame by frame and a slower one at wider intervals.
- Opacity and Falloff (Fade or Trail) set how the outer frames drop off: Fade to read one neighbor clearly, Trail to read the shape of the whole path.
- Colors are set per side, so you don’t confuse the frames behind with the frames ahead.
Then pick one landmark and follow it, like the heel, the top of the head, or the tip of a hand. Even gaps mean constant speed. Gaps that tighten toward a frame mean the action slows into it, and gaps that open up mean it speeds away. Animators call those slow in and slow out.
Sketch those gaps as a ladder next to a key drawing and you have a timing chart, the note traditional animators write on a drawing to tell the inbetweener where the frames fall. Some software calls it a spacing chart, which is a better name for what it holds.
Drop the colors when the footage is real
Red for the frames behind and blue for the frames ahead works on drawn animation, like the GIF at the top of this page. The image is flat there, and the silhouette is most of what you read.
On live-action footage, the color fights the picture. Set both sides to no color and the neighbors stay photographic, as ghosts at the opacity you chose.
With three frames on each side, sampled 0.3 seconds apart at 70 percent opacity, a seven-second clip of someone running becomes one image of the same stride repeated across the pavement. You don’t have to estimate the spacing between foot plants anymore: it’s laid out end to end, at a known interval, in the same frame.
Save the overlay back to the board as a still and draw on it: the arc the head travels, where each foot plants, and the count between them. An arc is close to invisible in playback and obvious once it’s a drawn line.
What you worked out stays on the board as its own item, next to the source and its credit.
Study a paused frame like a still
Stop on a pose and open the ToolBox panel. The five lenses that work on a still image work on a paused frame too, and a key pose has its own questions.
A two-value Notan in Values tells you whether the pose reads as a silhouette. If an arm disappears into the torso in black and white, it wasn’t clear in the reference either, and copying it won’t fix that. Read groups the frame into its big shapes, so you can squint without the blur.
In Measure, a plumb line through the supporting foot and the head shows where the weight really sits. Angle Finder reads the angle between the shoulder line and the hip line, so you don’t have to guess at the twist. Composition puts grids and a viewfinder over the frame when the question is staging rather than the body.
None of this changes the source clip. Save the result on the same board as the frame it came from, or export it. To see how these tools compare with the desktop reference apps, read the reference board app comparison.
Keep the clip with the rest of the shot
Reference for a shot is never just one video. There’s the clip, the character you have to stay on model to, a note about what was asked for, the link back to where the clip came from, and the two or three frames you pulled out of it. On one board, they stay together.
If the board is missing a model sheet or a second clip, the animation reference guide lists production archives, and the gesture drawing guide has motion libraries.
On iPad, you might animate in Procreate Dreams, RoughAnimator, Callipeg, or something else. ArtDeck runs in Split View, Slide Over, and Stage Manager, so if your animation app supports the same, both stay on screen and you can check the count on frame 54 without switching apps.
Boards work offline, and a board set to sync through iCloud is on all your devices when you need it.
What ArtDeck will not do for you
ArtDeck doesn’t key, in-between, or composite, and it has no timeline to put drawings on. You can annotate, though: pull a frame or an onion-skin overlay out to the board and draw over that copy, which is where the arcs and counts above end up. The source clip stays as it was.
What you take away is a set of numbers and comparisons for the app where the animation actually gets made.
The motion study page lists the playback controls, and the study tools page covers the five lenses.
Common questions
How do I step through a video frame by frame on an iPad?
Put the clip on an ArtDeck board and use the controls on the item itself. Video speed cycles through 0.5×, 1×, 1.5×, and 2×, and GIFs go down to 0.25×. The counter next to the transport shows the current frame and the total, so you always know where you are.
Let a video play through once before stepping, because a paused video shows a coarser preview than it does in playback.
What is onion skinning for, if I am not animating?
It shows you the spacing in someone else’s footage. With the frames around the current one drawn on top of it, you can see the gaps a landmark leaves from frame to frame: even gaps mean constant speed, and gaps that tighten toward one frame mean a slow in. You’re reading the motion out of the reference, not making a drawing.
Do the frame counts transfer to my own animation?
Only if the clip and your scene run at the same frame rate. Phone footage is usually 30 frames per second, which doesn’t line up with a 24fps scene unless you convert. Either way, treat the counts as a starting point, because animation that matches live action frame for frame looks slack.
Can I animate in ArtDeck?
No. There’s no timeline, and ArtDeck doesn’t key, in-between, or composite. It’s for reading reference, and it gives you numbers and comparisons for the app where the animation actually gets made. You can still draw over a frame or an onion-skin overlay pulled out to the board, which is where arcs and timing charts end up.