Good uses
- Chapter endings in mystery or exploration games that set the next goal
- Novels and interactive fiction where choices return into the story
- Learning material that links understanding from one answer to the next question
01
Classification and repeated action
A stimulus engine is the axis of what invites another action. The visual world is chosen separately as an expression theme.
- Stimulus engine
- Narrative pull
- Repeated action
- Open a clue, choose an option, or work a mechanism in the scene to confirm the answer to a question.
- Core loop
- Question
- Clue
- Partial answer
- New question
- Typical feedback
- Chapter, Foreshadowing, Figure in shadow, Choices, Cliffhanger
02
Build the sequence
Design the opening state, the input, the result, and the next actionable state as one sequence.
Opening state
A chapter number, a short question, and a scene containing clues are visible, and it is readable where to touch to get closer to the answer.
Input
An input such as choosing an option, opening a clue, or working a mechanism is received with the scene moving exactly as far as the finger moves.
Result and reward
The question is marked "Solved" with a one-sentence partial answer, and the clue used joins the clue list as its count rises.
Next actionable state
A larger question built on that answer appears, with the next clue and the place to touch shown, so the user can choose to continue on their own.
03
Feedback across channels
Do not rely on motion or sound alone: every channel should make the change of state readable.
Visual
Visual: keep the question, answer, and clue count in fixed places on screen, and tie the answer to the new question with the same mark or color.
Motion
Motion: mechanisms move at the speed of the input, the chain behind an answer plays one link at a time, and figures or doors appear slowly.
Text, sound, and haptics
Text, sound, and haptics: one sentence each for question and answer, a short confirming tone on solving, and a pause with sound pulled back before the new question.
Without motion or audio
Without motion or audio: the "Solved" mark, the one-sentence answer, the higher clue count, and the new question text alone show how the story moved forward.
04
Production controls
Eight production controls decide the strength and order of the response.
Input
Choose an option, open a clue, or pull a key or mechanism. Make one input that connects directly to the question the center of the scene.
Timing
Respond to the input at once, and leave 0.5–1 s between the answer and the new question so the answer can be read first.
Hit stop
Hold only for one beat just before the new question appears. Keep the answer text still for as long as it takes to read, and never freeze choice or back controls.
Amplification
Let one answer set the scene's mechanisms moving in sequence until a hidden door or figure appears. Link the answer and the next question by lighting the same mark.
Text
State the question in one sentence and the answer in one sentence. Never trail off with an ellipsis to drag things out, and show the number of clues paid off.
Camera
Start close on the clue, pull back slowly with the chain of the answer to show the whole scene, and settle quietly on the place of the new question.
Sound & haptics
Build three steps: the sound of the mechanism releasing, a confirming tone for the solve, and quiet before the new question. Keep haptics optional and only at the solve.
Next expectation
Draw a line from the next clue to the place of the new question to show the input that opens the continuation. Do not foreground wait timers or paid unlocks.
05
Where it works—and where it fails
Good uses, failure modes, risk, accessibility, and the boundary that keeps the engine from being used to manipulate.
Failure modes
- Adding only questions and moving to the next chapter without paying off a single answer
- Cutting a sentence mid-way to force reading on, creating a wait with no answer
- Staging the answer so long that it is unclear what was actually learned
Risk
Open questions pull hard, so stalling without payoff makes users feel they are paying time for answers and costs the story their trust. A hook that cuts a sentence mid-way makes users want more through the discomfort of interruption, not through the content.
Accessibility
Keep the question, answer, and clue count on screen as text so a screen reader can follow them in order. In reduced-motion mode, replace the mechanism chain with state changes, and let users extend how long text stays on screen.
Manipulation boundary
Do not hide the continuation or the answer behind paid unlocks, wait timers, or ad views, and do not make users pay to relieve the unease of a question. Do not plant questions that cannot be paid off; mark where the story ends and where to rest, and keep progress when the user stops mid-way.
06
Watch the interaction grammar
A 6-second vertical 1080×1920, 30 fps specimen with no audio, and a text alternative of what it shows.
A study at night. The grandfather clock has stopped at 3:59; neither the pendulum nor the gears move. A key with a crescent-moon head is pushed into the gear window, its tip caught where the gears mesh. At the top: "Episode 2: The stopped clock", "Clues 3", and the question "Why did it stop at 3:59?". A finger grips the key head and pulls; the key comes out exactly as far as the finger moves, and on the frame its tip clears the gear teeth, the gears turn, the pendulum starts to swing, and the second hand moves. The question is stamped "Solved" with the answer "A key was blocking the gears", and the key goes into the fourth slot of the clue tray, "Crescent key", taking the clues from 3 to 4. When the minute hand reaches 12 and it is 4:00, the striking hammer rings the bell; its pull runs along a rod on the wall and releases the bookcase latch, and the bookcase slides aside. Behind it is a hidden door with light seeping around it; the crescent on its keyhole and the crescent on the clue key light up simultaneously. A moment later, a new question appears: "Who hid the door that opens at four?". Finally a dotted line runs from the clue key to the keyhole of the hidden door, and the finger is poised over that key. No text trails off into an ellipsis to drag the story out.
Generated examples for this engine
No generated examples are recorded for this engine yet. The specimen is also listed in the gallery.
08
Creator-ready English prompt
A production brief you can hand directly to an AI model or a production team.
Written for this guide from the specimen and the classification; it is not the original generation prompt.
Produce a 6-second vertical UI motion sample that uses the Narrative pull stimulus engine.
Repeated action: Choose an option, open a clue, or pull a key or mechanism. Make one input that connects directly to the question the center of the scene.
Opening state: A chapter number, a short question, and a scene containing clues are visible, and it is readable where to touch to get closer to the answer.
Trigger and timing: Respond to the input at once, and leave 0.5–1 s between the answer and the new question so the answer can be read first.
Hit stop: Hold only for one beat just before the new question appears. Keep the answer text still for as long as it takes to read, and never freeze choice or back controls.
Amplification: Let one answer set the scene's mechanisms moving in sequence until a hidden door or figure appears. Link the answer and the next question by lighting the same mark.
Text: State the question in one sentence and the answer in one sentence. Never trail off with an ellipsis to drag things out, and show the number of clues paid off.
Camera: Start close on the clue, pull back slowly with the chain of the answer to show the whole scene, and settle quietly on the place of the new question.
Sound and haptics: Build three steps: the sound of the mechanism releasing, a confirming tone for the solve, and quiet before the new question. Keep haptics optional and only at the solve.
Next promise: Draw a line from the next clue to the place of the new question to show the input that opens the continuation. Do not foreground wait timers or paid unlocks.
Without motion or audio: the "Solved" mark, the one-sentence answer, the higher clue count, and the new question text alone show how the story moved forward.
Accessibility: Keep the question, answer, and clue count on screen as text so a screen reader can follow them in order. In reduced-motion mode, replace the mechanism chain with state changes, and let users extend how long text stays on screen.
Ethical boundary: Do not hide the continuation or the answer behind paid unlocks, wait timers, or ad views, and do not make users pay to relieve the unease of a question. Do not plant questions that cannot be paid off; mark where the story ends and where to rest, and keep progress when the user stops mid-way.
Keep enough contrast for text and primary UI, and make the meaning clear in reduced-motion and no-audio settings.
Create original abstract subjects. Do not imitate existing works, characters, logos, brands, or proprietary UI.