Can a drawing feel like an object?

What to look for

Move slowly and erratically

Speed up, slow down, reverse direction, stop dead in the middle of a drag. The deformation should track all of it continuously. The failure mode this is testing for is the opposite: a canned animation that plays on touch and looks roughly right but isn't actually following your finger. That distinction is invisible in a screenshot and obvious in the hand.

Two fingers work too. So does dragging past the edge of the shape.

The comparison

Flat is not the same as soft

The flat one is the drawing itself being pushed around: bold outline, flat fill, the white die-cut edge. It deforms convincingly, but the thing you're deforming still reads as a picture of something. A four-year-old put it better than any of my notes had — she looked at it and said it was a sticker, not a squish toy. She was right, and no amount of shading was going to argue her out of it.

So the solid one drops the outline and the die-cut edge entirely, because those are what make a sticker a sticker, and gets actual volume instead. Its outline isn't redrawn by hand — it's the flat version's own silhouette, which is why it still reads as the same character. Press it and the whole body responds rather than just the spot under your thumb.

Worth trying both with a slow press and holding still, then again with a fast flick. The two differ more under a stationary press than under a drag, which is the opposite of what I expected going in.

The buttons under the solid one swap its finish — gold, galaxy, holo and so on. Same character, same silhouette, same mesh underneath; only the surface changes. Gold and chrome are worth a slow tilt of the phone, and the holo one only really makes sense in motion, because what it's doing is shifting with the angle you look at it from rather than wearing a printed rainbow.

Why it's the hard part

Dragging is not tapping

There's a well-known result in the HCI literature that makes this specific interaction awkward to build. Jota et al. measured perceptible latency separately for tapping and dragging, and the thresholds are nowhere near each other: tap latency goes unnoticed below roughly 24 ms, while dragging latency stays detectable down to about 2.4 ms — an order of magnitude stricter.

Most fidget and squish apps are built around taps, which is a forgiving place to live. A stretch is a drag, which isn't. Commercial touchscreens already spend somewhere in the range of 50–200 ms end to end, most of it on the display side and none of it mine to fix, so the only sane strategy is to not add to it.

So the contact point isn't simulated at all. The mesh vertices under your finger are driven straight from the raw pointer position with no spring and no smoothing of any kind — any filter on finger position is latency. The springs only run where latency can't be seen: the dent that blooms under a press, and the settle after you let go. The contact point is pinned, not simulated. Everything else is a spring.

Honest caveats

What this isn't

It's a web build, so it's capped at 60 fps in Safari and can't use iOS's predicted-touch API — both of which a native version would get. It's one deformation model, so everything here squishes like the same material. The artwork is placeholder test shapes chosen to be hard to deform well, not finished characters. There is no game attached to it.

If it feels good anyway, that's the useful signal. If it doesn't, that's a more useful one.

Parameters

Tuned by eye, because there's nothing to look up

I went looking for a number — how fast does gelatin actually wobble, what frequency reads as "soft" — across animation, game-development and rheology writing. There isn't one. No published constant tells you how a drawing of a dumpling ought to move, and anything quoting one is making it up.

So the settle is underdamped on purpose, which is why it rings a little instead of stopping dead, and it holds at full stretch for a beat before snapping back, because releasing instantly reads as rubber rather than something soft. Both of those were dialled in by poking it several hundred times, which is the least sophisticated and most reliable instrument available.