System Requirements Document for shiny-video

1. Introduction

shiny-video is a cinematic AI video project: a seven-scene restoration narrative in which a person discovers an abandoned Lamborghini deep in a forest and brings it back to brand-new condition. The product intent is to deliver a complete, coherent, ready-to-generate prompt set for that video — the overall concept, the seven scene prompts in sequence, the visual style directives, the ASMR/sound directives, and the negative-prompt exclusions — presented as a first-party editorial experience.

The audience is the Video Creator: someone who watches and produces satisfying transformation and hyper-real vehicle-restoration content, is patient and detail-hungry, and is drawn to before/after drama, moody atmosphere, and the emotional payoff of decay turned into gleam. The experience is cinematic and editorial — film-poster gravitas, not a utility app.

2. System Overview

shiny-video is a custom-UI, browser-delivered experience with four first-party pages: Landing, Concept, Scenes, and Directives. All four are anonymously reachable (access_requirement: none); no account, sign-in, or identity establishment is required or implied anywhere in the product.

The current delivery covers:

• The overall cinematic concept: discovery of the abandoned Lamborghini, the decision to restore it, the time-lapse restoration process (cleaning, repairing, replacing parts, polishing, repainting), and the final brand-new, shiny car driving out of the forest.
• The seven scene prompts in fixed sequence — Discovery, Close-up Damage, Cleaning Begins, Repair Process, Interior Restoration, Final Reveal, Driving Away — each with its full prompt text.
• The visual style directives: cinematic, realistic, 4K, ultra detailed, dramatic lighting, film grain, shallow depth of field.
• The sound/ASMR directives: satisfying ASMR sounds, cleaning sounds, mechanical tools, engine roar, ambient forest sounds.
• The negative-prompt directives (avoid): cartoon, anime, blurry, low quality, distorted, extra limbs, watermark, text.

The only active human actor is the Video Creator. There are no other personas, no provider-owned surfaces, no external destinations, and no system-only capabilities that a human interacts with. The product is a presentation and reference surface for the accepted prompt set; it does not generate, render, host, or edit video itself, and no such capability is claimed.

2a. Product Interpretation and Delivery Boundary

Everything in shiny-video is delivered as first-party custom UI in the browser, anonymously. The four pages are the complete information architecture: Landing introduces the concept, Concept states the overall narrative and transformation, Scenes carries the seven numbered scene prompts in order, and Directives carries the style, sound, and negative-prompt rules. There is no protected area, no login, no dashboard, and no account lifecycle, because no accepted journey requires durable actor-specific state, a commitment, an entitlement, or a value transfer bound to a participant.

The current horizon is the prompt set and its presentation. Anything beyond that — actually generating, rendering, or publishing the video — is outside the accepted scope and is not represented as a current page, module, or capability.

2b. Source Content Inventory

Not applicable. No reference directive in this project declares content_source; all product facts come from the authoritative user requirement thread and the accepted Planning Scope.

2c. Page Content and Component Coverage

Landing

• Information/state: The public entry. A full-bleed cinematic still of the abandoned Lamborghini half-swallowed by forest fog fills the viewport, desaturated and dark. Over the left third, a massive Playfair Display headline — RESTORED — set at 56px mobile to 128px desktop, running past the right edge of its text column as a graphic gesture. Below it, a 1px hairline rule and a 14px Space Grotesk line: "A seven-scene cinematic restoration." A single red 4px underline slab sits under the word RESTORED.
• Primary action: Enter the concept — continue into the narrative.
• Supporting actions: Jump directly to the scene sequence; jump directly to the directives.
• Domain entities: The video concept; the seven-scene structure; the restoration narrative.
• Component responsibilities: Full-bleed hero image frame (sharp corners, no radius); oversized display headline; hairline rule; red underline slab; entry links to Concept, Scenes, Directives.
• States: Loading — hero image resolves with a slow crossfade; type reveals by line with a 400ms upward mask. Empty — not applicable (static editorial content). Success — hero and headline fully composed, entry links reachable. Error/recovery — if the hero image fails to load, the black ground (#0B0B0C) remains with the headline, hairline, and entry links intact and fully readable; no readable text or control is ever covered or cropped.

Concept

• Information/state: The overall cinematic AI video concept. A person discovers an abandoned Lamborghini deep in a forest. The car has shattered windows, flat tires, rust, dirt, and a completely wrecked interior — torn seats, cracked dashboard, broken steering wheel, leaves and mud everywhere. The person decides to restore it. The restoration is a time-lapse process: cleaning, repairing, replacing parts, polishing, repainting. The final shot: the Lamborghini looks brand new, shiny, and drives out of the forest. Cinematic lighting, 4K, realistic detail, satisfying transformation.
• Primary action: Continue into the seven-scene sequence.
• Supporting actions: Return to Landing; jump to Directives.
• Domain entities: The abandoned Lamborghini; the person; the forest; the restoration process stages (cleaning, repairing, replacing parts, polishing, repainting); the final brand-new state.
• Component responsibilities: Asymmetric editorial text blocks occupying 5–7 columns offset from an edge; oversized Playfair Display section headline; 12px uppercase tracked eyebrow labels; hairline rules as the only structural chrome; red underline slab under key words on scroll; full-bleed imagery breaking the grid on the opposite side of its text.
• States: Loading — text blocks and imagery reveal one-shot on scroll, unhurried. Empty — not applicable. Success — the full concept reads as a before/after editorial spread. Error/recovery — if imagery fails, the text blocks remain complete and readable on the black ground.

Scenes

• Information/state: The seven required scene prompts in sequence, each with its full prompt text, presented as a vertical film strip with oversized scene numerals (01–07) pinned left and the prompt text set in a narrow measure beside it. A vertical film-strip rail down the left edge carries the seven scene numbers, the active one marked by the red slab and off-white type.
  • Scene 1 — Discovery: Cinematic shot, a person walking through a dense forest, suddenly stops and finds an abandoned Lamborghini covered in dirt, leaves, and moss. Shattered windshield, flat tires, rusted body. Moody lighting, fog, realistic, 4K.
  • Scene 2 — Close-up Damage: Close-up of a wrecked Lamborghini interior. Torn leather seats, cracked dashboard, broken steering wheel, shattered glass on the floor, mud and leaves inside. Dusty, abandoned, cinematic lighting, realistic detail, 4K.
  • Scene 3 — Cleaning Begins: A person starts cleaning the abandoned Lamborghini in the forest. Removing leaves, dirt, and trash from the interior. Satisfying ASMR-style cleaning, cinematic lighting, realistic, 4K.
  • Scene 4 — Repair Process: Time-lapse of a Lamborghini being repaired. Replacing broken glass, fixing the engine, repairing the interior, sanding and repainting the body. Tools, sparks, mechanical work. Cinematic, realistic, 4K.
  • Scene 5 — Interior Restoration: Restoring the Lamborghini interior. New leather seats, clean dashboard, new steering wheel, polished details. Satisfying transformation, close-up shots, cinematic lighting, 4K.
  • Scene 6 — Final Reveal: The fully restored Lamborghini, shiny and brand new, parked in the forest. Golden hour sunlight, reflections on the paint, cinematic slow pan around the car. Realistic, 4K.
  • Scene 7 — Driving Away: The restored Lamborghini drives out of the forest onto an open road. Engine roaring, dust behind, cinematic drone shot, sunset, realistic, 4K.
• Primary action: Read each scene prompt in sequence; hover a scene card to swap the wreckage frame for the restored frame.
• Supporting actions: Move between scenes via the film-strip rail; return to Concept; jump to Directives.
• Domain entities: The seven scenes; each scene's prompt text; each scene's wreckage and restored imagery; scene numerals 01–07.
• Component responsibilities: Oversized Playfair Display scene numerals pinned to the left margin; scene cards with before/after hover crossfade (600ms); full-bleed scene imagery breaking the grid opposite its text; hairline rules; red slab marking the active scene number; 12px uppercase tracked scene metadata labels.
• States: Loading — scene imagery crossfades in slowly; scroll-triggered reveals are one-shot. Empty — not applicable (all seven scenes are always present). Success — all seven prompts fully readable, sequence legible as a film index. Error/recovery — if a scene image fails, the prompt text and scene numeral remain complete and readable; the hover swap degrades to the default frame without hiding text.

Directives

• Information/state: The required production directives, in three groups.
  • Visual style: cinematic, realistic, 4K, ultra detailed, dramatic lighting, film grain, shallow depth of field.
  • Sound/ASMR: satisfying ASMR sounds, cleaning sounds, mechanical tools, engine roar, ambient forest sounds.
  • Negative (avoid): cartoon, anime, blurry, low quality, distorted, extra limbs, watermark, text.
• Primary action: Read and apply the directives alongside the scene prompts.
• Supporting actions: Return to Scenes; return to Concept; return to Landing.
• Domain entities: Visual style directives; sound/ASMR directives; negative-prompt directives.
• Component responsibilities: Three clearly separated directive groups; oversized Playfair Display group headings; 12px uppercase tracked labels; hairline rules between groups; red underline slab marking the negative-prompt group as the exclusion set; asymmetric text blocks with deliberate empty black columns.
• States: Loading — one-shot scroll reveal. Empty — not applicable. Success — all three directive groups fully readable and unambiguous. Error/recovery — static text; no failure mode that hides content.

3. Functional Requirements

FR-1 — Overall concept presentation (explicit) As a Video Creator, I should be able to read the complete overall concept of the cinematic AI video — a person discovers an abandoned Lamborghini deep in a forest, the car has shattered windows, flat tires, rust, dirt, and a completely wrecked interior (torn seats, cracked dashboard, broken steering wheel, leaves and mud everywhere), the person decides to restore it, the restoration proceeds as a time-lapse of cleaning, repairing, replacing parts, polishing, and repainting, and the final shot shows the Lamborghini brand new, shiny, and driving out of the forest — so that I understand the full narrative before assembling the scene prompts. Lifecycle: Trigger — the Video Creator opens the Concept page. Observable result — the complete concept is presented as readable editorial text. Access — anonymous. Failure/recovery — if supporting imagery fails to load, the concept text remains complete and readable. Continuation — the Video Creator proceeds to the seven-scene sequence.

FR-2 — Scene 1 prompt (explicit) As a Video Creator, I should be able to read the full Scene 1 — Discovery prompt: cinematic shot, a person walking through a dense forest, suddenly stops and finds an abandoned Lamborghini covered in dirt, leaves, and moss; shattered windshield, flat tires, rusted body; moody lighting, fog, realistic, 4K. Lifecycle: Trigger — the Video Creator opens the Scenes page. Observable result — Scene 1 is presented as the first item in the sequence with its full prompt text. Access — anonymous. Failure/recovery — if the scene image fails, the prompt text remains complete. Continuation — the Video Creator moves to Scene 2.

FR-3 — Scene 2 prompt (explicit) As a Video Creator, I should be able to read the full Scene 2 — Close-up Damage prompt: close-up of a wrecked Lamborghini interior; torn leather seats, cracked dashboard, broken steering wheel, shattered glass on the floor, mud and leaves inside; dusty, abandoned, cinematic lighting, realistic detail, 4K. Lifecycle: Trigger — the Video Creator reaches Scene 2 in the sequence. Observable result — Scene 2 is presented with its full prompt text. Access — anonymous. Failure/recovery — prompt text remains complete if imagery fails. Continuation — the Video Creator moves to Scene 3.

FR-4 — Scene 3 prompt (explicit) As a Video Creator, I should be able to read the full Scene 3 — Cleaning Begins prompt: a person starts cleaning the abandoned Lamborghini in the forest; removing leaves, dirt, and trash from the interior; satisfying ASMR-style cleaning, cinematic lighting, realistic, 4K. Lifecycle: Trigger — the Video Creator reaches Scene 3. Observable result — Scene 3 is presented with its full prompt text. Access — anonymous. Failure/recovery — prompt text remains complete if imagery fails. Continuation — the Video Creator moves to Scene 4.

FR-5 — Scene 4 prompt (explicit) As a Video Creator, I should be able to read the full Scene 4 — Repair Process prompt: time-lapse of a Lamborghini being repaired; replacing broken glass, fixing the engine, repairing the interior, sanding and repainting the body; tools, sparks, mechanical work; cinematic, realistic, 4K. Lifecycle: Trigger — the Video Creator reaches Scene 4. Observable result — Scene 4 is presented with its full prompt text. Access — anonymous. Failure/recovery — prompt text remains complete if imagery fails. Continuation — the Video Creator moves to Scene 5.

FR-6 — Scene 5 prompt (explicit) As a Video Creator, I should be able to read the full Scene 5 — Interior Restoration prompt: restoring the Lamborghini interior; new leather seats, clean dashboard, new steering wheel, polished details; satisfying transformation, close-up shots, cinematic lighting, 4K. Lifecycle: Trigger — the Video Creator reaches Scene 5. Observable result — Scene 5 is presented with its full prompt text. Access — anonymous. Failure/recovery — prompt text remains complete if imagery fails. Continuation — the Video Creator moves to Scene 6.

FR-7 — Scene 6 prompt (explicit) As a Video Creator, I should be able to read the full Scene 6 — Final Reveal prompt: the fully restored Lamborghini, shiny and brand new, parked in the forest; golden hour sunlight, reflections on the paint, cinematic slow pan around the car; realistic, 4K. Lifecycle: Trigger — the Video Creator reaches Scene 6. Observable result — Scene 6 is presented with its full prompt text. Access — anonymous. Failure/recovery — prompt text remains complete if imagery fails. Continuation — the Video Creator moves to Scene 7.

FR-8 — Scene 7 prompt (explicit) As a Video Creator, I should be able to read the full Scene 7 — Driving Away prompt: the restored Lamborghini drives out of the forest onto an open road; engine roaring, dust behind, cinematic drone shot, sunset, realistic, 4K. Lifecycle: Trigger — the Video Creator reaches Scene 7. Observable result — Scene 7 is presented with its full prompt text as the final item in the sequence. Access — anonymous. Failure/recovery — prompt text remains complete if imagery fails. Continuation — the Video Creator moves to the directives.

FR-9 — Ordered scene sequence (explicit) As a Video Creator, I should be able to read the seven scene prompts in their required order — Discovery, Close-up Damage, Cleaning Begins, Repair Process, Interior Restoration, Final Reveal, Driving Away — so that the prompt set reads as a coherent film sequence rather than an unordered list. Lifecycle: Trigger — the Video Creator opens the Scenes page. Observable result — the seven scenes appear in fixed sequence with oversized numerals 01–07 pinned left and a film-strip rail marking the active scene. Access — anonymous. Failure/recovery — the sequence remains ordered and legible even if imagery fails. Continuation — the Video Creator reads through to Scene 7.

FR-10 — Visual style directives (explicit) As a Video Creator, I should be able to read the required visual style directives — cinematic, realistic, 4K, ultra detailed, dramatic lighting, film grain, shallow depth of field — so that I can apply them to the prompt set. Lifecycle: Trigger — the Video Creator opens the Directives page. Observable result — the visual style directives are presented as a distinct, fully readable group. Access — anonymous. Failure/recovery — static text; no failure mode that hides content. Continuation — the Video Creator reads the sound/ASMR directives.

FR-11 — Sound/ASMR directives (explicit) As a Video Creator, I should be able to read the required sound/ASMR directives — satisfying ASMR sounds, cleaning sounds, mechanical tools, engine roar, ambient forest sounds — so that I can apply them to the prompt set. Lifecycle: Trigger — the Video Creator reaches the sound/ASMR group on the Directives page. Observable result — the sound/ASMR directives are presented as a distinct, fully readable group. Access — anonymous. Failure/recovery — static text; no failure mode that hides content. Continuation — the Video Creator reads the negative-prompt directives.

FR-12 — Negative-prompt directives (explicit) As a Video Creator, I should be able to read the required negative-prompt directives — avoid cartoon, anime, blurry, low quality, distorted, extra limbs, watermark, text — so that I can exclude them from generation. Lifecycle: Trigger — the Video Creator reaches the negative-prompt group on the Directives page. Observable result — the negative-prompt directives are presented as a distinct, fully readable exclusion set. Access — anonymous. Failure/recovery — static text; no failure mode that hides content. Continuation — the Video Creator has the complete prompt set.

FR-13 — Before/after scene interaction (explicit) As a Video Creator, I should be able to hover a scene card and see the wreckage frame crossfade to the restored frame over 600ms, so that the transformation itself becomes the interaction. Lifecycle: Trigger — the Video Creator hovers a scene card. Observable result — the wreckage image crossfades to the restored image; on pointer exit it returns. Access — anonymous. Failure/recovery — if either image is unavailable, the card shows its default frame and the prompt text remains fully readable. Continuation — the Video Creator continues through the sequence.

FR-14 — Anonymous access to all pages (explicit) As a Video Creator, I should be able to reach Landing, Concept, Scenes, and Directives without creating an account or signing in, so that I can read the complete prompt set immediately. Lifecycle: Trigger — the Video Creator opens the experience. Observable result — all four pages are reachable anonymously; no identity establishment is requested. Access — none required on any page. Failure/recovery — not applicable; there is no protected state to recover. Continuation — the Video Creator reads the full prompt set.

FR-15 — Readable text and controls at every viewport (explicit) As a Video Creator, I should be able to read every headline, wordmark, label, number, card text, and control entirely inside the viewport and its container at 375px, 768px, and 1280px, with text wrapping or scaling to fit and no other element covering any part of it. Lifecycle: Trigger — the Video Creator views any page at any supported viewport. Observable result — all readable text and controls stay whole; imagery and decoration may crop, bleed, or overlap freely as long as they cover no readable text or control. Access — anonymous. Failure/recovery — where the direction asks readable text to run off an edge, the gesture is carried by imagery or decoration instead. Continuation — the Video Creator reads the full prompt set at any viewport.

4. User Personas

Video Creator

Product context. The Video Creator is producing a cinematic AI video of an abandoned Lamborghini discovered deep in a forest and restored to brand-new condition. They work in the register of satisfying transformation and hyper-real vehicle-restoration content: patient, detail-hungry, drawn to before/after drama, moody atmosphere, and the emotional payoff of decay turned into gleam. They are not looking for a utility dashboard; they are looking for a film-title-grade presentation of a prompt set they can take straight into generation.

Primary goal. Walk away with a complete, coherent, ready-to-generate prompt set for the seven-scene restoration video — concept, all seven scene prompts in order, the visual style directives, the ASMR/sound directives, and the negative-prompt exclusions — with nothing missing and nothing contradictory.

Distinct accepted responsibilities. The Video Creator defines and reads the overall concept (discovery, wrecked condition, decision to restore, time-lapse restoration, brand-new driving-away finale); reads each of the seven scene prompts in its required sequence and with its full prompt text; reads and applies the visual style directives; reads and applies the sound/ASMR directives; and reads and applies the negative-prompt exclusions. Their work is editorial curation and assembly, not operation of a tool — the value is in the completeness and coherence of the prompt set and in the drama of its presentation.

Relevant inputs or decisions. The Video Creator decides whether the concept is complete and coherent, whether each scene prompt carries its required detail (for example Scene 1's fog and moss, Scene 2's torn leather and shattered glass, Scene 4's tools and sparks, Scene 6's golden hour and reflections, Scene 7's drone shot and sunset), and whether the style, sound, and negative directives are all present and unambiguous. They also choose how they move through the material — sequentially through the film strip, or jumping directly to the directives.

Interactions with other accepted participants. None. The Video Creator is the only active human actor in this product; there is no counterparty, recipient, or beneficiary whose state changes, and no handoff or approval step. The experience is a single-participant editorial read.

Observable success. The Video Creator has read the complete concept, all seven scene prompts in order with their full text, and all three directive groups, and can apply them directly to generation — with every headline, label, number, and control fully readable at 375px, 768px, and 1280px.

5. Core User Flows

Flow 1 — Read the concept and understand the narrative

1. The Video Creator opens the experience anonymously and lands on Landing. A full-bleed cinematic still of the abandoned Lamborghini half-swallowed by forest fog fills the viewport, desaturated and dark. The oversized Playfair Display headline RESTORED spans from the left edge and runs past the right edge of its text column, its final letters cropped by the viewport edge as a graphic gesture; a 1px hairline rule and a 14px Space Grotesk line — "A seven-scene cinematic restoration." — sit below it, with a single red 4px underline slab under the word RESTORED.
2. The Video Creator selects the entry into the concept. The Concept page opens.
3. On Concept, the Video Creator reads the overall concept as asymmetric editorial text blocks: a person discovers an abandoned Lamborghini deep in a forest; the car has shattered windows, flat tires, rust, dirt, and a completely wrecked interior — torn seats, cracked dashboard, broken steering wheel, leaves and mud everywhere; the person decides to restore it; the restoration is a time-lapse of cleaning, repairing, replacing parts, polishing, and repainting; the final shot shows the Lamborghini brand new, shiny, and driving out of the forest; cinematic lighting, 4K, realistic detail, satisfying transformation.
4. Observable result: the Video Creator understands the full narrative arc from wreckage to gleam. Next step: continue into the seven-scene sequence.
5. Failure/recovery: if supporting imagery fails to load, the concept text remains complete and readable on the black ground; the Video Creator continues reading without loss.

Flow 2 — Read the seven scene prompts in sequence

1. From Concept, the Video Creator continues into Scenes. The page presents a vertical film strip with oversized Playfair Display numerals 01–07 pinned left and a film-strip rail down the left edge; the active scene number is marked by the red slab and off-white type.
2. The Video Creator reads Scene 1 — Discovery: cinematic shot, a person walking through a dense forest, suddenly stops and finds an abandoned Lamborghini covered in dirt, leaves, and moss; shattered windshield, flat tires, rusted body; moody lighting, fog, realistic, 4K.
3. The Video Creator continues to Scene 2 — Close-up Damage: close-up of a wrecked Lamborghini interior; torn leather seats, cracked dashboard, broken steering wheel, shattered glass on the floor, mud and leaves inside; dusty, abandoned, cinematic lighting, realistic detail, 4K.
4. The Video Creator continues to Scene 3 — Cleaning Begins: a person starts cleaning the abandoned Lamborghini in the forest; removing leaves, dirt, and trash from the interior; satisfying ASMR-style cleaning, cinematic lighting, realistic, 4K.
5. The Video Creator continues to Scene 4 — Repair Process: time-lapse of a Lamborghini being repaired; replacing broken glass, fixing the engine, repairing the interior, sanding and repainting the body; tools, sparks, mechanical work; cinematic, realistic, 4K.
6. The Video Creator continues to Scene 5 — Interior Restoration: restoring the Lamborghini interior; new leather seats, clean dashboard, new steering wheel, polished details; satisfying transformation, close-up shots, cinematic lighting, 4K.
7. The Video Creator continues to Scene 6 — Final Reveal: the fully restored Lamborghini, shiny and brand new, parked in the forest; golden hour sunlight, reflections on the paint, cinematic slow pan around the car; realistic, 4K.
8. The Video Creator continues to Scene 7 — Driving Away: the restored Lamborghini drives out of the forest onto an open road; engine roaring, dust behind, cinematic drone shot, sunset, realistic, 4K.
9. Observable result: all seven prompts have been read in their required order with their full text. Next step: move to the directives.
10. Failure/recovery: if a scene image fails to load, the scene numeral and full prompt text remain complete and readable; the sequence stays ordered and legible.

Flow 3 — Compare wreckage and restored states on a scene card

1. On Scenes, the Video Creator hovers a scene card. The card shows the wreckage frame by default.
2. The wreckage frame crossfades to the restored frame over 600ms. On pointer exit, it crossfades back.
3. Observable result: the transformation itself becomes the interaction — the before/after drama is felt directly on the card. Next step: the Video Creator continues through the sequence or moves to the directives.
4. Failure/recovery: if either image is unavailable, the card shows its default frame and the prompt text remains fully readable; no readable text or control is covered.

Flow 4 — Read and apply the production directives

1. From Scenes, the Video Creator moves to Directives. Three clearly separated groups are presented, each with an oversized Playfair Display heading, 12px uppercase tracked labels, and hairline rules between groups.
2. The Video Creator reads the visual style directives: cinematic, realistic, 4K, ultra detailed, dramatic lighting, film grain, shallow depth of field.
3. The Video Creator reads the sound/ASMR directives: satisfying ASMR sounds, cleaning sounds, mechanical tools, engine roar, ambient forest sounds.
4. The Video Creator reads the negative (avoid) directives: cartoon, anime, blurry, low quality, distorted, extra limbs, watermark, text — marked by the red underline slab as the exclusion set.
5. Observable result: the Video Creator holds the complete prompt set — concept, seven scene prompts, style, sound, and exclusions. Next step: apply the prompt set to generation.
6. Failure/recovery: the directives are static text with no failure mode that hides content; the Video Creator can return to Scenes, Concept, or Landing at any point.

Flow 5 — Read the full prompt set at any viewport

1. The Video Creator opens the experience at 375px, 768px, or 1280px.
2. Every headline, wordmark, label, number, card text, and control stays entirely inside the viewport and its container, wrapping or scaling to fit; no other element covers any part of it. Imagery and decoration crop, bleed, and overlap freely as the direction asks, without covering readable text or controls.
3. Observable result: the complete prompt set is fully readable at every supported viewport. Next step: the Video Creator applies the prompt set to generation.
4. Failure/recovery: where the direction asks readable text to run off an edge, the gesture is carried by imagery or decoration instead, keeping the text whole.

6. Visuals Colors and Theme

Muse and headline. Tobias van Schneider — Editorial swagger for a machine brought back from the dead. A restoration story is fundamentally a before/after editorial spread: wreckage on one page, gleam on the next. Asymmetric grids and sharp-cornered, chrome-free layouts let the imagery and the seven scene prompts carry the drama instead of UI furniture.

Palette (dark mode).

No gradients. No blue anywhere. The red accent is used sparingly and always meaningfully.

Typography.

• Headings: Playfair Display — oversized display, tight leading (0.92–0.96), tight tracking (-0.02em), mixed case with occasional all-caps eyebrow labels set in the body face.
• Body: Space Grotesk.
• Scale: 1.5 modular display rhythm — 56/72/104/128px for display (56px mobile → 128px desktop via clamp); 28/20/16/14px for body and labels; 12px uppercase tracked labels for scene metadata.
• The extreme size contrast between display headlines and small 13–15px labels is itself the composition.

Shape language. Sharp corners only — no border-radius on cards, images, or buttons. Hard-edged full-bleed image frames. 1px hairline rules (#2A2A2D) as the only structural decoration. Thick red underline slabs (4–6px) as emphasis markers under key words and active scene numbers.

Layout. Asymmetric editorial grid: a 12-column base where text blocks occupy 5–7 columns offset from the left or right edge, leaving deliberate empty black columns as breathing room. Full-bleed scene imagery breaks the grid on the opposite side of its text. The scene sequence runs as a vertical film strip with oversized scene numerals (01–07) pinned left and the prompt text set in a narrow measure beside it. Generous vertical rhythm — sections separated by 120–180px of black.

Imagery style. Large-scale cinematic photography: fog-choked forest discovery, macro interior decay (torn leather, cracked dash, shattered glass), tool-and-spark repair frames, and a golden-hour full-car reveal. Monochrome-leaning treatment with the red accent reserved for graphic overlays, not the photographs. Images are full-bleed and generously cropped, often cut by the viewport edge.

7. Signature Design Concept

The public entry is a single full-bleed cinematic still of the abandoned Lamborghini half-swallowed by forest fog, filling the viewport, desaturated and dark. Over the left third of the image, a massive Playfair Display headline — RESTORED — set at 56px mobile to 128px desktop, spans from the left edge and deliberately runs past the right edge of the text column, its final letters cropped by the viewport edge as a graphic gesture. Below it, a 1px hairline rule and a 14px Space Grotesk line: "A seven-scene cinematic restoration." A single red 4px underline slab sits under the word RESTORED.

There is no centred stack, no subheadline-and-button pattern, and no gradient blob — the photograph and the oversized type are the entire first screen. The gesture of the headline running past the column edge is carried by the type's own graphic placement; the readable word itself stays whole and inside the viewport at every breakpoint, with the crop applied to the composition's negative space rather than to any readable character. Entry links to Concept, Scenes, and Directives sit below the hairline rule, fully readable and uncovered.

8. Interaction Model & Motion Direction

Interaction Model: Static Motion Tempo: restrained Hero Dimensionality: flat

Landing Hero Motion Brief.

• Focal subject: the full-bleed cinematic still of the abandoned Lamborghini half-swallowed by forest fog, desaturated and dark, with the oversized Playfair Display headline RESTORED composed over its left third.
• Input → transformation → outcome thesis: as the page settles, the hero photograph resolves with a slow crossfade while the headline reveals by line with a 400ms upward mask; the red 4px underline slab slides in under the word RESTORED. The outcome is a composed first frame in which the photograph and the oversized type are the entire screen — no bounce, no parallax theatrics, no particles.
• Motion vocabulary: slow crossfades, line-by-line type reveal with a 400ms upward mask, one-shot unhurried scroll-triggered reveals, and the 600ms wreckage-to-restored crossfade on scene-card hover.
• Composed first frame: the fog-choked Lamborghini still filling the viewport; RESTORED spanning from the left edge past the right edge of its text column; the hairline rule and the 14px Space Grotesk line beneath it; the red underline slab under RESTORED.
• Reduced-motion state: with prefers-reduced-motion, all crossfades and type reveals resolve immediately to their final composed state; the scene-card hover swap shows the restored frame without animation; nothing moves, and every readable element remains whole and uncovered.

9. Non-Functional Requirements

NFR-1 — Visual fidelity directives (explicit) The prompt set must carry the visual style directives exactly: cinematic, realistic, 4K, ultra detailed, dramatic lighting, film grain, shallow depth of field. Rationale: these are explicit hard constraints in the authoritative source and define the look of the generated video.

NFR-2 — Sound/ASMR directives (explicit) The prompt set must carry the sound/ASMR directives exactly: satisfying ASMR sounds, cleaning sounds, mechanical tools, engine roar, ambient forest sounds. Rationale: explicit hard constraint in the authoritative source.

NFR-3 — Negative-prompt exclusions (explicit) The prompt set must carry the negative-prompt directives exactly: avoid cartoon, anime, blurry, low quality, distorted, extra limbs, watermark, text. Rationale: explicit hard constraint in the authoritative source; these are exclusions to be avoided in generation, not a prohibition on any product capability.

NFR-4 — Readable text and controls at every viewport (explicit) Headlines, wordmarks, labels, numbers, card text, and controls must stay entirely inside the viewport and their container at 375px, 768px, and 1280px, wrapping or scaling (for example font-size: clamp(...) with its mobile size) to fit, with no other element covering any part of them. Imagery, decoration, and motion may be cropped, bled off an edge, rotated, overlapped, or cut as the direction asks, as long as they cover no readable text or control. Where the direction asks readable text to be cropped, clipped, covered, or run off an edge, the text stays whole and the gesture is carried by imagery or decoration instead. Rationale: explicit user design constraint; this rule takes precedence for readable text and controls.

NFR-5 — Reduced-motion support (explicit) With prefers-reduced-motion, motion stops and content shows whole: crossfades and type reveals resolve immediately, and any moving or scrollable content stops and shows whole items, wrapping into rows or sitting in a horizontally scrollable row (overflow-x: auto) whose further items are reached by scrolling. Rationale: explicit user design constraint.

NFR-6 — Palette and typography fidelity (explicit) The dark-mode palette (#0B0B0C background, #151517 surface, #F2EFE9 text/primary, #E4401F accent, #8A857E muted, #2A2A2D hairline), the Playfair Display heading face, the Space Grotesk body face, the 1.5 modular display rhythm, and the sharp-corner shape language must be preserved. No gradients, no blue anywhere, no rounded-corner cards or pill buttons, no Inter/Roboto/Arial/Helvetica/Open Sans/Lato/Poppins/system-ui as display or body faces, no centred headline-plus-subtext-plus-button hero, no gradient blobs, no grid of identical hover-lift cards, no bright saturated colour grounds, and no decorative particles, bouncy easing, or playful illustrations. Rationale: explicit creative direction and user design constraints; the generic indigo/blue-on-white SaaS template is forbidden for this project.

NFR-7 — Anonymous delivery (explicit) All four pages are reachable without an account or sign-in. No identity establishment, session continuity, or permission control is required or implied. Rationale: the accepted page contract sets access_requirement: none on every surface, and no accepted journey requires durable actor-specific state, a commitment, an entitlement, or a value transfer bound to a participant.

NFR-8 — Content completeness (explicit) The seven scene prompts, the three directive groups, and the overall concept must all be present with their full source text; no scene, directive, or concept element may be omitted, merged, or paraphrased away. Rationale: the authoritative source enumerates each item explicitly.

10. Tech Stack

• Frontend: React (web), delivered in the browser. Rationale: custom UI is accepted (delivery_shape.custom_ui: true) and the experience is a first-party editorial surface.
• Styling: CSS with clamp()-based fluid type for the 56px → 128px display scale, and prefers-reduced-motion media queries. Rationale: required by NFR-4 and NFR-5.
• Fonts: Playfair Display (headings) and Space Grotesk (body), self-hosted or loaded as web fonts. Rationale: explicit creative direction.
• Backend: None required. The product presents a static, source-complete prompt set with no persistence, no accounts, and no server-side state. Rationale: no accepted journey requires durable state or a server.
• Storage: None required. Rationale: no accepted journey requires persistence.
• Deployment: Static hosting of the built frontend. Docker/docker-compose and Kubernetes are not required, because there is no backend service, no database, and no orchestration need. Rationale: deployment does not require them.

11. Assumptions and Constraints

Assumptions.

• A-1 (required_inference) — The four pages (Landing, Concept, Scenes, Directives) are the complete information architecture, taken exactly from the versioned page contract; no page is added, removed, merged, split, renamed, or reordered.
• A-2 (required_inference) — The Video Creator is the only active human persona; there is no counterparty, recipient, or beneficiary whose state changes, so no participant handoff or approval step exists.
• A-3 (required_inference) — All four pages are anonymously reachable; no identity establishment is needed because no accepted journey requires durable actor-specific state, a commitment, an entitlement, or a value transfer bound to a participant.
• A-4 (required_inference) — The product presents the prompt set; it does not generate, render, host, or edit video. No such capability is claimed or implied.
• A-5 (required_inference) — Scene imagery (wreckage and restored frames) is illustrative of the accepted scene content; where imagery is unavailable, the prompt text and scene numerals remain complete and readable.

Constraints.

• C-1 (explicit) — Negative-prompt directives: avoid cartoon, anime, blurry, low quality, distorted, extra limbs, watermark, text. These are exclusions applied to generation; they do not prohibit any product capability.
• C-2 (explicit) — Visual style must be cinematic, realistic, 4K, ultra detailed, dramatic lighting, film grain, shallow depth of field.
• C-3 (explicit) — Sound/ASMR must include satisfying ASMR sounds, cleaning sounds, mechanical tools, engine roar, ambient forest sounds.
• C-4 (explicit) — The seven scene prompts must appear in their required order: Discovery, Close-up Damage, Cleaning Begins, Repair Process, Interior Restoration, Final Reveal, Driving Away.
• C-5 (explicit) — The palette, typography, shape language, layout, and motion direction in Sections 6–8 are binding; the generic indigo/blue-on-white SaaS template is forbidden.
• C-6 (explicit) — Readable text and controls stay whole at 375px, 768px, and 1280px; imagery and decoration may crop, bleed, or overlap but must cover no readable text or control.

12. Glossary

• Video Creator — The single active human persona: the person assembling and applying the cinematic AI video prompt set.
• Scene prompt — One of the seven required prompt texts, in fixed order, from Discovery through Driving Away.
• Directives — The three required production rule groups: visual style, sound/ASMR, and negative-prompt exclusions.
• Negative prompt — The exclusion set applied to generation: cartoon, anime, blurry, low quality, distorted, extra limbs, watermark, text.
• ASMR — The satisfying, tactile sound register required for the cleaning and mechanical sequences.
• Before/after swap — The 600ms crossfade on a scene card from the wreckage frame to the restored frame on hover.
• Film-strip rail — The vertical left-edge rail carrying scene numbers 01–07, with the active scene marked by the red slab and off-white type.
• Red slab — The 4–6px #E4401F underline used as an emphasis marker under key words and active scene numbers; the only accent colour in the composition.
• Hairline — The 1px #2A2A2D rule used as the only structural decoration.