How to Make Paper Animation: 3 Working Methods

Pick the route before you buy anything, because the three routes share almost no equipment and produce visibly different results.

The physical route is the only one where the shadows are real, and shadow is what the eye checks first. The rigged route is the only one that lets you retime a movement after the fact. The generated route is the only one that finishes before you have made a coffee. Nothing here is a substitute for anything else.

Searches for how to make paper animation edit belong to the middle route. That phrase almost always means a social edit where the card never physically existed and everything was assembled on a timeline.

Cut-out animation is stop motion done with flat materials. The reference definition is worth keeping in mind while you build: as Wikipedia's stop motion article puts it, stop motion of flat materials such as paper, fabrics or photographs is usually called cutout animation. The tradition is long and cheap: John Halas made the Snip and Snap shorts in 1960 and 1961 around a dog cut from a sheet of paper, and Trey Parker and Matt Stone built The Spirit of Christmas shorts of 1992 and 1995 out of construction paper. If you want to see how far the craft goes, Aardman works in the same frame-by-frame discipline at a studio scale, and the BFI holds a deep collection covering the flat-material tradition on film.

1. 1

   Cut the parts, not the character

   Head, torso, upper arm, forearm, thigh, shin, foot. Cut each piece separately with a blade rather than scissors, because a blade edge stays crisp under a light and a scissor edge curls.

2. 2

   Pin the joints

   Split-pin brads through the overlap, loose enough to move with a fingertip. Put the pin at the anatomical joint, not at the corner of the shape, or the limb will swing from the wrong point.

3. 3

   Build the stage flat

   Background on the table, characters on top, and small foam pads under anything that should float above the base. Those pads are what give you a shadow gap, and the gap is what makes the layers separate on camera.

4. 4

   Lock the camera

   A phone on a copy stand or a clamp, pointing straight down. Tape the stand to the table. Any drift between frames shows up as a wobble that no stabiliser will fix, because the wobble is in the artwork, not the lens.

5. 5

   Kill the ambient light

   Two lamps at forty-five degrees, curtains shut, autoexposure and autofocus locked. A cloud passing outside your window will change the exposure of one frame in the middle of a shot, and you will see it.

6. 6

   Move a little, shoot, repeat

   Two to three millimetres per pose for a walk, more for a throw. Shoot each pose, do not chimp between frames, and keep your hands out of shot rather than out of frame.

The rigged route trades realism for control. You scan the parts once, then move them forever.

• Scan each cut part on a flat bed at 600 DPI, or photograph them flat in even light, then remove the background so every part has its own transparency.
• Import each part as its own layer and set the anchor point at the joint, not at the layer centre. This single step is where most rigs go wrong.
• Parent forearm to upper arm, upper arm to torso, and so on, so rotating the torso carries the whole arm with it.
• Puppet pins are for bending a single sheet, not for jointing. Use them on a flag or a curling page; use parenting on a limb.
• Add the texture pass last: a subtle paper grain over the whole comp, plus a soft drop shadow on each part offset a few pixels in the direction your imaginary lamp is.

That last item is what separates a rigged card animation from flat vector art. Without grain and an offset shadow, layered shapes read as a slide deck. The same rig works in any layer-based editor; if you are on DaVinci Resolve 21, the Fusion page is where the node version of this lives, and the free edition includes it, per Blackmagic's product page read on 7 September 2026.

The generated route skips the cutting entirely. You supply a photograph, the tool isolates the subject, and the motion comes from a fixed repertoire rather than from your hands.

1. 1

   Start with a clean subject

   One clear subject, no motion blur, and plenty of empty space around it. The isolation step decides everything downstream, and it cannot recover an edge that was already soft.

2. 2

   Choose the move before the tool

   A fold, an unfold, a tear, a page turn or a headline reveal. Tools that ship four honest modes will do one of those well; a general prompt box will do all of them inconsistently.

3. 3

   Render at the largest size offered

   Card texture is fine detail, and fine detail is the first thing a social encoder discards. Whatever the tool exports, keep the master and let the platform do the compressing.

4. 4

   Composite if you need a background

   Ask for a green screen version, then key it over your own footage. An overlay is reusable; a baked background is a single shot.

LazyKiwi's paper animation tool is the version of this route I use for reference, because its stated behaviour is unusually specific: four fixed modes instead of a prompt box, every frame drawn locally instead of queued on a server, a stated 8 frames per second, and free 720p output with no stamp on the frame, all as described on the tool page on 7 September 2026. ImageToPaper is the other specialist worth knowing; its composer exposes a green-screen toggle and six sliders, including Shadow Intensity and Edge Intensity, which are the two that matter for making card read as card.

If you want the motion directed rather than picked from a list, that is a general video model's job instead. The photo animation template animates a still, and models such as Kling v3 Omni hold a described camera move well. They will not hold a card surface across eight shots, so use them for the hero and a specialist for the series.

This is the single most searched move in the format, and it is done three ways depending on the route you chose.

• Physically: crumple a real sheet, then flatten it one crease at a time in reverse, shooting each state. Play the sequence backwards and it crumples; play it forwards and it opens. Ten to fifteen states is enough.
• Rigged: build the sheet from four or six triangles that share edges, then rotate each triangle around its shared edge in sequence. Add a gradient that darkens each facet as it turns away from the light.
• Generated: name the move in the prompt and let the tool pick the geometry. You get consistency and lose the ability to time an individual crease.

a folded sheet of matte card opening outward one crease at a time, creases catching light as they turn, subject printed on the inner face revealed as it opens, static camera, no zoom, no morphing, no gloss

The trick that makes any of the three believable is the physical reference. Crumple an actual sheet, photograph it under a hard lamp, and look at where the light lands: highlights sit along the fold ridges and shadow pools in the valleys, and the ridges are never perfectly straight. A digital version that misses that reads as folded plastic.

Animation has run on a two-frame hold for as long as film has had a standard. Wikipedia's frame rate article records that 24 frames per second became the standard for 35 mm sound film between 1927 and 1930, and that showing one drawing for every two frames results in a display of only 12 drawings per second, which is enough for most subjects; fast movement is the case where animators switch to a new drawing every frame.

• Shoot on twos as the default. Halving the number of poses halves the work and looks correct for card, which is a slightly stiff medium anyway.
• Switch to ones only for a fast action beat, then go back. Mixed cadence within a shot is normal and invisible.
• Generated tools make this decision for you. The LazyKiwi tool renders at 8 frames per second, which is deliberately below film cadence because a slight stutter is part of the medium's signature.
• Lock exposure, white balance and focus before frame one, and shoot to a memory card rather than a preview.
• Tape everything. The stand, the cable, the background sheet, your own footing near the table.

One extra habit: shoot two identical frames at the head and tail of every shot. Those give you a handle to trim against in the edit, and they cost four seconds.

Any comparison has to be at a fixed output, so plan against one ten-second clip delivered vertical for a feed. These are planning estimates for someone competent but not fast, not stopwatch measurements.

• Physical: about eight hours in total. Roughly two on artwork, one on rig and lighting, four shooting around 120 poses on twos, one on the edit. The variance is in the artwork, not the shooting.
• Rigged: about four hours the first time and forty minutes on the second clip, because the rig is reusable. Scanning and background removal is the boring hour.
• Generated: under five minutes for a first result and about twenty if you iterate on the source photo and the prompt. Most of that twenty is choosing a better photograph.

Treat those as planning figures rather than a benchmark, and note that production write-ups on sites like No Film School budget far longer for anything with a story in it, and they scale badly in one direction: doubling clip length roughly doubles the physical route and barely moves the other two. That is the real argument for a rig or a tool when a series is coming.

Effort is not visible in the result. Nobody watching a two-second reveal can tell whether it took eight hours or five minutes; they can tell whether the shadow is real.

— Daniel Okafor

Four faults account for nearly every disappointing result, and all four are fixable before you render anything.

1. 1

   Flat lighting

   One light straight on kills the shadow that separates the layers, and without separation you have a collage, not an animation. Move the lamp off axis.

2. 2

   No gap between layers

   If every element sits on the same plane, nothing casts on anything. Physically that is a foam pad; digitally it is an offset shadow per layer.

3. 3

   Plastic texture

   Gloss, bevel and a uniform surface are renderer defaults. Card is matte, slightly uneven, and has a furred edge where it was cut or torn.

4. 4

   Motion that is too smooth

   Perfectly eased movement belongs to vector graphics. Card is heavier and stiffer than that, and a small hold at the end of each move sells the weight.

A last sanity check before you publish. Export, upload as an unlisted post, and watch it on a phone at arm's length. Texture that survived your monitor often does not survive the platform's encoder, and if the fibre has gone you are better off rendering larger or simplifying the artwork than adding more grain.

If you want the shortlist rather than the build, the companion comparison covers what each tool ships and what its free tier exports.