Neutral diagrams isolate one decision at a time. Compare a cut with a move, then inspect staggering, overlap, continuity, and attention. These are new demonstrations—not copied footage or working product interfaces.
Experiment 01 / Original diagram
A cut is not a fast zoom.
Same start. Same destination. Different continuity.
01 / Cut
Workspace
02 / Move
Workspace
Watch for. The left view changes in one discrete step. The right traverses intermediate scales and slows before it stops. Similar before/after stills would hide the difference.
Chosen settings. Both wait 600 ms. The cut switches with a step-end transition; the move lasts 900 ms with cubic-bezier(.16, 1, .3, 1). Both then hold. These are original demonstration values, not measured source-film presets.
What differs. The schematic geometry and curve are invented. The Atlas frame check establishes a cut; the Codex check supports a qualitative deceleration, not this exact curve.
Can a new surface enter without losing the request?
RequestCompare these options.
Result
Watch for. The request moves first. The result begins entering before that move finishes, so the transition reads as a connected re-layout rather than two unrelated events.
Chosen settings. Request re-layout: 450–1,150 ms. Result entrance: 700–1,350 ms. Divider: 600–1,100 ms. Moves use cubic-bezier(.22, 1, .36, 1); result opacity is linear. These are original demonstration values, not measured source-film presets.
What differs. This is an invented split layout. The source reference supports overlapping growth and annotation, not the placement, timings, or capabilities shown here.
A phrase can build without every word competing at once.
One phrase / three arrivals
Buildwithcontext.
Watch for. Words start 120 ms apart and settle onto one unchanged baseline. The completed phrase holds longer than any one entrance.
Chosen settings. First entrance at 400 ms; 120 ms stagger; 600 ms per word; 24 px vertical travel; cubic-bezier(.22, 1, .36, 1). Final hold begins at 1,240 ms. These are original demonstration values, not measured source-film presets.
What differs. The words, horizontal arrangement, travel distance, and timings are original. The references use different layouts and include a later re-centering phase; do not collapse those into one generic text effect.
Does the audience recognize the same card inside a device?
Field note / 04
A place to begin.
Watch for. The same DOM object moves and scales; it is not replaced with a second look-alike. The device outline arrives during the move, then both settle together.
Chosen settings. Card transforms from 500–1,500 ms using cubic-bezier(.22, 1, .36, 1). Device fades in from 800–1,500 ms. The final composition holds 1,500 ms. These are original demonstration values, not measured source-film presets.
What differs. A neutral card and device outline replace all source UI, art, and branding. This experiment tests continuity; it does not claim how the source compositor constructed its layers.
A selection can explain a reframe without another camera move.
Document
One detail worth noticing.
Keep this context visible.
Reframed detailThe change has a reason.
Watch for. A highlight appears, a callout follows, and only then does the diagram cut to the detail. The final surface stays still long enough to read.
Chosen settings. Highlight: 350–700 ms, linear. Callout: 850–1,300 ms, cubic-bezier(.22, 1, .36, 1). Discrete detail change at 1,750 ms, followed by a 1,250 ms hold. These are original demonstration values, not measured source-film presets.
What differs. This teaching sequence combines separate observed ideas. It is not an actual product workflow, a latency promise, or a reconstruction of one source film.
Overlap, scale, and a short stagger can establish depth without a bounce.
03 / Background
02 / Context
01 / FocusOne clear foreground.
Watch for. Back, middle, and foreground layers enter with overlapping moves. Different final scales and positions establish the hierarchy; every move settles without an added overshoot.
Chosen settings. Entrances begin at 400, 560, and 720 ms; each lasts 850 ms using cubic-bezier(.16, 1, .3, 1). The foreground is last. Final hold: 1,230 ms. These are original demonstration values, not measured source-film presets.
What differs. These invented cards test a layer hierarchy, not the source's imagery or exact path. No spring or overshoot parameter was recovered from the source study.