Plain Prompt to Art Card Showcase

A silent, result-first image-model showcase places a striking finished art card on a dark stage, then types a plain-language request beside it. The card flips through visually different results inside the same designed layout, implying that a short text prompt can yield directed artwork (inferred from the post and frames).

Why it works

• The finished cyan artwork is visible at 0:00, before any explanation; the viewer can judge the output instantly.
• The first type-on line builds from 0:00.25 to 0:01.5 while the result stays visible, connecting ordinary words to a polished picture.
• A card change at 0:01.8 and two more at 0:03.8 and 0:06.0 provide three visual rewards within 8.1 seconds, without changing the surrounding layout.

Hook formula: Result first + visual pattern interrupt: {{HIGH_QUALITY_RESULT}} appears immediately, then {{PLAIN_LANGUAGE_PROMPT}} types in beside it.

Retention map: Fresh results at 0:01.75, 0:03.8, and 0:06.0 · payoff starts in frame one and repeats at each swap · no explicit CTA · final art recalls the opening, making a visual loop (inferred).

Mix: 100% designed composite. There are no camera cuts or presenter shots. The street photograph and illustrations are assets inside moving vertical cards.

Rhythm: One 8.1-second scene; three detected picture changes at 1.8, 3.8, and 6.0 seconds. Between changes, typed text supplies the smaller motion.

Rules this editor follows (inferred from the sheets):

• Keep the dark stage, peripheral floating props, and left-side prompt area fixed as the art changes.
• Make the newest image large and center-right; let a previous card remain visible behind or to its right.
• Use radically different image subjects and palettes across examples so a card handoff proves range immediately.

Full shot list

• Look: Near-black background, approximately #0e0e0e. Main teal card around #078b91 with cream artwork; later cards introduce concrete gray, vivid green, and violet. Peripheral soft-focus donut-like shapes in mint, cyan, tan, and lavender sit partly outside frame. One sharper pink textured prop anchors the lower middle. These props belong to this execution; invent your own motif.
• Captions / prompt text: There are no spoken-word captions. Use white #f2f2f2 IBM Plex Mono or Space Mono Regular at roughly 13–16 px on 1280×720, left aligned around x=140, y=340. Show 1–2 short lines, type on over about 1.2 s, then clear. A small magenta #e62d9a right-facing triangle precedes the line.
• Graphics and overlays: Main cards are roughly 110×150 px in the 1540-wide sheet tiles, equivalent to about 365×500 px at full 1280×720 playback. Keep a slightly offset secondary card visible at 25–45% of the main card's area. Tiny brand marks and microtype live inside the artwork; do not add a large end slate.
• B-roll: One top-down, sunlit pedestrian photo functions as an art-card example, not a full-screen footage cut. Hold each main image around two seconds.
• Transitions: Card layer, slide, and pivot swaps at about 1.8, 3.8, and 6.0 seconds; no hard cuts. Exact easing is inferred from sampled frames.
• Branding: The reference's first and last card share the same teal branded poster. For a new video, put {{BRAND_MARK}} discreetly on your own card system.

Measured: No audio track, spoken words, WPM, pitch, or pauses.

Text delivery: Treat the typed request as the narrator. Reveal at a steady, readable typewriter pace, hold briefly, then clear as the next card takes focus. Use plain, concrete wording that describes the desired image.

• Music: None in the reference.
• Sound effects: None in the reference.
• Silence and space: The format is fully legible while muted.
• Mix targets: Not applicable; the source has no audio track. A remake can remain silent.

• Title: Benefit-led feature claim; template: {{PRODUCT}} turns plain words into {{VISUAL_OUTCOME}}.
• Cover / thumbnail: Select the opening teal-style hero frame: one tall high-contrast art card center-right, black negative space left, colorful floating accents, no large headline.
• Description: The post frames the feature as plain-language prompting and emphasizes art direction and light. For a remake, state the benefit in one sentence and link to {{PRODUCT_URL}}; use original wording.
• Audience signals: 1,072 views, 21 likes, one comment, and three reposts at capture. No most-replayed data or comment content was supplied, so these do not establish a particular audience reaction.

The reference has zero spoken words. The following is an on-screen text and image script.

[0:00–0:01.75 · HERO] Show {{RESULT_IMAGE_1}} immediately. Type “{{PROMPT_1}}” beside it (aim for 6–9 words).
[0:01.75–0:03.8 · VARIATION] Bring {{RESULT_IMAGE_2}} forward. Clear the first line; type “{{PROMPT_2}}” (aim for 5–8 words).
[0:03.8–0:06.0 · RANGE] Bring {{RESULT_IMAGE_3}} forward; keep a smaller former card visible.
[0:06.0–0:08.1 · RETURN] Restore {{RESULT_IMAGE_1}} as hero with {{BRAND_MARK}}. End on a composition close to frame one.

Script prompt (paste into an AI assistant with this file attached):

Create a silent 8.1-second, 16:9 prompt-to-result showcase for {{PRODUCT}} about {{TOPIC}}. Write two original plain-language image requests, each 5–9 words, and describe three visually distinct original results that still share one art direction. Map them to the four beats at 0:00, 0:01.75, 0:03.8, and 0:06.0. The first finished image must be visible at frame one. Give me only the two on-screen text lines, image briefs, and timed shot list. Avoid the reference brand, artwork, and slogans.

Voice prompt: No voice. The type-on text carries the idea.

Visual prompts:

• Hero art card: Create original vertical campaign artwork for {{PRODUCT}} inspired by {{PROMPT_1}}. Bold single motif, deliberate composition, crisp microtype placeholders, high contrast, palette {{PALETTE_1}}. No existing brand marks.
• Second art card: Create an original top-down or otherwise strongly framed image for {{PROMPT_2}}. Clear light direction and one readable subject. Palette {{PALETTE_2}}.
• Third art card: Create another original vertical art treatment for {{TOPIC}}, distinct from the first two in subject and palette, with similarly polished composition. Palette {{PALETTE_3}}.

Caption spec: IBM Plex Mono Regular, 13–16 px at 1280×720, white, x≈140/y≈340, no box, one or two lines, typewriter entrance. A 12–15 px magenta triangle sits 15–20 px to the left.

Music prompt: None; export silent to match the reference.

1. Design three original vertical result cards; ensure each reads instantly at roughly 350–400 px wide on a 1280 px canvas.
2. Set a near-black 1280×720 stage. Add a few soft-focus colored accent objects around the frame perimeter, leaving the left middle clear for text.
3. Place the first card center-right at frame zero and offset a secondary card behind it.
4. Animate two short left-side prompt lines with a typewriter reveal. Clear each line when the corresponding card changes.
5. At 1.75, 3.8, and 6.0 seconds, slide/pivot the next card into the front position. Keep transitions short, about 0.25–0.4 s (inferred production target).
6. Return to the first card for the final two seconds. Export 30 fps, 8.1 s, without an audio track.

• A finished result is readable at 0:00.
• First prompt is complete by roughly 0:01.5; first major change occurs around 0:01.75.
• Three internal card changes occur near 1.8, 3.8, and 6.0 seconds; no hard cuts are required.
• Prompt type remains legible at phone size and does not overlap the art.
• Three cards differ visibly in subject and palette, while the stage and card system remain consistent.
• Final frame recalls the opening; audio is absent if matching the reference.

• The creator's signature (don't copy): The specific liquid-word poster, embedded loud-creative identity, candy-like props, brand palette, and individual generated images.
• The transferable format (do copy): A result visible in frame one; plain-language request typing beside it; a fixed dark stage; three rapid card handoffs; a visual return to the hero result.
• Use your own artwork, brand marks, and image rights.

Source: https://x.com/magnific/status/2108188333633622450 · measured with dissect 1.1.0 on 2026-10-08 · production targets marked “inferred” are judgment calls.