spuneiartur/scroll-stop-prompter

Scroll-Stop Prompter Skill

You generate a coordinated set of 3 prompts that work together to produce scroll-stopping video content: a clean product shot, its deconstructed version, and a video transition between them.

After generating the prompts, you deliver them in a beautiful HTML page with tabbed navigation, one-click copy buttons, and confetti animation — so the user can instantly copy-paste each prompt.

---

The Process

Step 1: Confirm the Object

Ask the user what object they want (if not already specified). Good scroll-stop objects include:

• Laptops, phones, headphones, cameras (tech products)
• Shoes, watches, bags (fashion/luxury)
• Cars, bikes, drones (vehicles)
• Food & beverages (smoothies, cocktails, plated dishes)
• Any product with interesting internal components or ingredients

Default to laptop if the user doesn't specify.

Step 2: Generate Prompt A — The Assembled Shot

This is the clean, hero product image. Generate a detailed prompt optimized for AI image generators.

Template:

PROMPT A — ASSEMBLED SHOT

Professional product photography of a [OBJECT] centered in frame, shot from a [ANGLE] angle.
Clean white background (#FFFFFF), soft studio lighting with subtle shadows beneath the object.
The [OBJECT] is pristine, brand-new, fully assembled and closed/complete.

Photorealistic rendering, 16:9 aspect ratio, product catalog quality. Sharp focus across the
entire object, subtle reflections on glossy surfaces. Minimal, elegant, Apple-style product
photography. No text, no logos, no other objects in frame.

Shot on Phase One IQ4 150MP, 120mm macro lens, f/8, studio strobe lighting with large softbox
above and white bounce cards on sides. Ultra-sharp detail, 8K quality downsampled to 4K.

Customize for the specific object:

• Adjust the camera angle (3/4 view works best for most products)
• Add material-specific details (brushed aluminum, matte plastic, leather texture, etc.)
• Specify the state (laptop closed vs open, watch face visible, shoe from side profile, etc.)
• Keep it on white — this is critical for the deconstruction to read well

Step 3: Generate Prompt B — The Deconstructed Shot

This is the exploded/disassembled version. It should feel like the object has been elegantly taken apart and each piece is floating in space (or for food/beverage, an explosion of ingredients).

Template:

PROMPT B — DECONSTRUCTED / EXPLODED VIEW

Professional exploded-view product photography of a [OBJECT], deconstructed into its individual
components, all floating in space against a clean white background (#FFFFFF).

Every internal component is visible and separated: [LIST 8-15 SPECIFIC COMPONENTS FOR THE OBJECT].
Each piece floats with even spacing between them, maintaining the general spatial relationship
of where they sit in the assembled product. The arrangement follows a vertical or diagonal
explosion axis.

Soft studio lighting with subtle shadows on each floating piece. Components are pristine and
detailed — you can see textures, screws, ribbon cables, circuit traces. The overall composition
maintains the silhouette/outline of the original object.

Photorealistic rendering, 16:9 aspect ratio, technical illustration meets product photography.
Shot on Phase One IQ4 150MP, focus-stacked for sharpness across all floating elements.
Same lighting setup as the assembled shot for visual continuity.

Component lists by object type:

For a laptop:

• Aluminum unibody shell (top lid)
• LCD display panel with ribbon cable
• Keyboard deck / top case
• Trackpad module with haptic engine
• Battery cells (individual cells visible)
• Logic board / motherboard with chips visible
• SSD / storage module
• Fan assembly with heat pipe
• Speaker modules (left and right)
• Hinge mechanism
• Bottom case panel
• Rubber feet and screws arranged neatly
• WiFi antenna array
• Camera module

For a phone:

• Glass back panel, battery, OLED display, logic board, camera module array, SIM tray, speaker grille, Taptic engine, Lightning/USB-C port assembly, antenna bands, frame/chassis, face ID sensor array, wireless charging coil

For a shoe:

• Outer sole, midsole/cushioning layer, insole, upper mesh/leather panels, tongue, laces, heel counter, toe cap, eyelets, stitching thread, branding elements

For food/beverages (smoothies, cocktails, etc.):

• Use "explosion" instead of "deconstruction" — the glass shatters, liquid erupts, ingredients fly outward in a dramatic freeze-frame. List every ingredient, garnish, and element (ice cubes, glass shards, liquid splashes, fruit pieces, herbs, etc.). Emphasize high-speed photography style (1/10000s freeze) for dramatic effect.

For other objects, research and list 8-15 real internal components.

Step 4: Generate Prompt C — The Video Transition

This prompt instructs a video model to animate between the two states.

Template:

PROMPT C — VIDEO TRANSITION (Start Frame → End Frame)

START FRAME: A fully assembled [OBJECT] sitting centered on a white background, product
photography style, soft studio lighting.

END FRAME: The same [OBJECT] elegantly deconstructed into an exploded view — every component
floating in space, separated along a [vertical/diagonal] axis, maintaining spatial relationships.

TRANSITION: Smooth, satisfying mechanical deconstruction animation. The object begins whole
and still. After a brief pause (0.5s), pieces begin to separate — starting from the outer
shell and progressively revealing inner components. Each piece lifts and floats outward along
clean, deliberate paths. Movement is eased (slow-in, slow-out) with slight rotations on
individual pieces to reveal their 3D form. The separation happens over 2-3 seconds in a
cascading sequence, not all at once. Final floating arrangement holds for 1 second.

STYLE: Photorealistic, white background throughout, consistent studio lighting. No camera
movement — locked-off tripod shot. The only motion is the object deconstructing. Satisfying,
ASMR-like mechanical precision. Think Apple product reveal meets engineering visualization.

DURATION: 4-5 seconds total.
ASPECT RATIO: 16:9
QUALITY: High fidelity, smooth 24fps or higher, no artifacts.

Variations to offer:

• Reverse version: Start deconstructed, assemble together (equally compelling)
• Loop version: Assemble → pause → deconstruct → pause → repeat
• Slow-mo version: Same animation but 8-10 seconds, ultra-smooth

Step 5: Build and Open the HTML Prompt Page

After generating all 3 prompts, deliver them in a gorgeous HTML page. This is the key deliverable that makes the skill feel premium and easy to use.

How to build it:

1. Read the HTML template from assets/prompt-page-template.html (in this skill's directory)

2. Replace these placeholders with the actual content:

   • {{OBJECT_NAME}} — the object name for the page title (e.g. "Tropical Smoothie Explosion")
   • {{HEADING_LINE1}} — first word(s) of the heading (e.g. "SMOOTHIE")
   • {{HEADING_LINE2}} — second word(s), displayed faded (e.g. "EXPLOSION")
   • {{TAB_A_NAME}} — name for tab A (e.g. "Assembled Shot")
   • {{TAB_A_SHORT}} — short mobile label for tab A (e.g. "Assembled")
   • {{TAB_B_NAME}} — name for tab B (e.g. "Explosion Shot" or "Deconstructed")
   • {{TAB_B_SHORT}} — short mobile label for tab B (e.g. "Explosion")
   • {{PROMPT_A}} — the full text of Prompt A (plain text, no HTML)
   • {{PROMPT_B}} — the full text of Prompt B (plain text, no HTML)
   • {{PROMPT_C}} — the full text of Prompt C (plain text, no HTML)

3. Important: When inserting prompt text, escape any <, >, and & characters as HTML entities (<, >, &) so they render correctly in the browser.

4. Write the completed HTML to a file called prompts.html in the user's current working directory (or wherever they're working).

5. Open the file in the browser:

   • If a local server is running, open via localhost URL
   • Otherwise, use open prompts.html (macOS) or xdg-open prompts.html (Linux) to open the file directly

The HTML page features:

• Tabbed A/B/C interface — click a tab, the prompt appears. No scrolling needed.
• One-click Copy button on each prompt
• Confetti animation fires every time you copy
• VoltFlow design: Space Grotesk + Archivo + JetBrains Mono, #BFF549 lime accent, #02040a dark bg
• Glass-morphism cards, floating background orbs, subtle grid
• Keyboard shortcuts: 1/2/3 to switch tabs, Cmd+C to copy active prompt
• Fully responsive on mobile

Step 6: Also Present Prompts in Chat

After building the HTML page, also present the prompts in chat as a fallback, using this format:

## Your Scroll-Stop Prompt Set: [OBJECT]

### PROMPT A — Assembled Shot
[paste into your image generator, set to 16:9]

{prompt A}

---

### PROMPT B — Deconstructed Shot
[paste into your image generator, set to 16:9, optionally reference Prompt A's output as input]

{prompt B}

---

### PROMPT C — Video Transition
[paste into your video model, upload Prompt A output as start frame and Prompt B output as end frame]

{prompt C}

---

### Recommended Settings
- **Image generator**: 16:9 aspect ratio, highest quality/resolution available
- **Video model**: 16:9, 4-5 seconds, highest quality
- **Tip**: Generate the assembled shot first, then reference it when generating the deconstructed
  version for visual consistency (same color, lighting, angle)

---

Best Practices

1. Consistency is key — The assembled and deconstructed versions must look like the same object. Same materials, same colors, same lighting direction, same camera angle.
2. White background always — This makes the scroll-stop website build (Skill 2) much easier and cleaner. It also makes the video transition cleaner.
3. Component accuracy matters — Don't make up parts. Use real components for the object type. This sells the realism.
4. The video prompt is model-agnostic — Write it descriptively enough that it works in Runway, Kling, Pika, Higgsfield, or any other video model. The user just uploads start/end frames.
5. Offer the reverse — Sometimes the assembly animation (parts coming together) is even more satisfying than the deconstruction.
6. The HTML page is the primary deliverable — The chat output is secondary. The page should always be generated and opened automatically.

---

Error Recovery

| Issue | Solution | |---|---| | Image gen produces inconsistent lighting | Add "match exact lighting direction and intensity from reference image" to Prompt B | | Deconstruction looks random, not organized | Emphasize "maintain spatial relationships" and "explosion along single axis" in Prompt B | | Video transition is too fast/jerky | Increase duration to 6-8 seconds, emphasize "smooth eased motion" and "cascading sequence" | | Components don't look realistic | Add specific material descriptions (brushed aluminum, matte black plastic, green PCB with gold traces) | | White background isn't pure white | Add "pure white #FFFFFF background, no gradient, no vignette" explicitly | | HTML page doesn't open | Fall back to open <filepath> command, or tell user the file path to open manually | | Prompt contains HTML special chars | Always escape <, >, & when inserting into the HTML template |