Generates AI image and video prompts for scroll-stopping content. Creates three linked prompts: (1) a clean product/object shot on white background at 16:9, (2) an exploded/deconstructed version of the same object, and (3) a video transition prompt that animates between assembled and deconstructed states. Works with any AI image generator (Higgsfield, Midjourney, etc.) and any video model (Runway, Kling, Pika, etc.). Delivers prompts via a gorgeous HTML page with one-click copy buttons and confetti. Trigger when the user says "scroll-stop prompt", "deconstruction prompt", "exploded view prompt", "product animation prompt", or asks for prompts to create scroll-stopping video content from object imagery.
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.
Ask the user what object they want (if not already specified). Good scroll-stop objects include:
Default to laptop if the user doesn't specify.
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:
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:
For a phone:
For a shoe:
For food/beverages (smoothies, cocktails, etc.):
For other objects, research and list 8-15 real internal components.
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:
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:
Read the HTML template from assets/prompt-page-template.html (in this skill's directory)
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)Important: When inserting prompt text, escape any <, >, and & characters as HTML
entities (<, >, &) so they render correctly in the browser.
Write the completed HTML to a file called prompts.html in the user's current working
directory (or wherever they're working).
Open the file in the browser:
open prompts.html (macOS) or xdg-open prompts.html (Linux) to open
the file directlyThe HTML page features:
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)
| 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 |