A 15-second vertical ad for a running shoe. A runner crosses a city bridge at dawn; each footstrike briefly lights the road beneath her. End on a close-up of the shoe.

## Role and objective

<instructions>
You are a video director and generation-prompt writer. Rebuild the supplied concept into a polished, runnable 15-second vertical advertisement for a running shoe, intended for direct pasting into a video-generation workflow. Preserve every confirmed fact while making the staging visually legible and economical. Produce the required cut table, per-cut generation prompts, exclusions, parameters, and joining instructions. The result is complete only when every cut has all six direction topics, camera movement, speed, framing, duration, ending state where needed, coherent audio, and durations summing to exactly 15 seconds.
</instructions>
<context>
The confirmed concept is provided below.
</context>

## Scope and given facts

<instructions>
In scope: a 15-second vertical ad; a runner crossing a city bridge at dawn; each footstrike briefly lighting the road beneath her; the final image being a close-up of the shoe; minimal cuts; direct generation with no preparation stage.

Out of scope: invented brand claims, product-performance statistics, named brands, unconfirmed technical specifications, additional characters, and unrelated locations or story beats. Do not invent the runner's appearance, shoe design, bridge identity, resolution, frame rate, music licence, or commercial-clearance status. Use these slots where relevant: `[FILL IN: shoe appearance and runner description]`, `[FILL IN: resolution]`, `[FILL IN: frame rate]`, `[FILL IN: music licence status]`, and `[FILL IN: clearance status for identifiable people, trademarks, or recognisable buildings]`. Fill each with the actual production decision or verified source before rendering; never replace it with a plausible detail.
</instructions>
<context>
A 15-second vertical ad for a running shoe. A runner crosses a city bridge at dawn; each footstrike briefly lights the road beneath her. End on a close-up of the shoe.
</context>

## Working rules

<instructions>
Direct the video rather than transcribing the concept. Use the fewest cuts that preserve readable action; because the total is 15 seconds, allocate time unevenly, giving the footstrike-light interaction enough duration to read and reserving the final beat for the shoe close-up. Keep these values fixed across cuts unless the input explicitly changes them: dawn atmosphere, city-bridge setting, vertical composition, runner identity, shoe appearance, restrained colour grade, and the same rule that each visible footstrike creates a brief road glow.

For every cut, state:
1. Subject.
2. Action.
3. Setting, including dawn conditions.
4. Visual style, including light direction, colour treatment, depth of field, and texture.
5. Camera movement, speed, framing, and ending state.
6. Sound, including ambience, footfalls, road-light effect, music, and any dialogue.

Where a value is not supplied, use `[RECOMMENDED: value — one-line reason]`; do not leave a direction topic empty. If the shoe or runner is not described, use `[FILL IN: shoe appearance and runner description]` verbatim in every cut rather than inventing identity details. If the requested footstrike effect is visually unclear in a draft, strengthen timing or framing before adding a cut. If a new cut would make the action shorter than readable, keep the existing shot and adjust its camera movement instead.

State all four camera values in every cut: movement, speed, framing, and duration plus end state. Use real seconds and make their sum exactly 15. Do not name any generation service or use slash commands or double-dash flags.
</instructions>
<context>
For US-facing commercial use, leave music and clearance status as explicit verification slots. Do not assert that any advertising, publicity, trademark, building, or music rule applies; record the relevant clearance item and mark it for verification.
</context>

## Output structure

<instructions>
Produce the following sections in this order.

1. **Cut table:** Use columns `Cut`, `Visual`, `Camera`, `Audio`, and `Duration`. Keep visual entries as concise noun phrases, but include the footstrike-light cause and effect. Use the minimum practical number of cuts, with durations totaling exactly 15 seconds. For every multi-cut sequence, state the frame state at each ending so the next cut can open continuously.

2. **Per-cut generation prompt:** Give one copy-paste-ready block per cut. Each block must contain the six direction topics in labelled order and repeat `[FILL IN: shoe appearance and runner description]` exactly wherever the subject appears. Quote the continuity sentence settled in cut 1 in every later cut. Include the selected camera values and the exact ending state.

3. **Exclusions:** Provide one uninterrupted noun-phrase list covering unwanted artefacts, unreadable footfalls, inconsistent identity, extra limbs, warped shoes, arbitrary text, logos unless supplied, unintended crowds, false weather, and unrelated locations.

4. **Parameters:** Give `Aspect ratio: 9:16`. Give `Resolution: [FILL IN: resolution]` and `Frame rate: [FILL IN: frame rate]` unless a recommendation is explicitly requested; add that the service's plan may not offer the chosen value and to step down one level if unavailable. Give `Total length: 15 seconds`.

5. **How to run it:** Explain how to generate the cuts separately and join them in numerical order, preserving the final frame of each cut as the opening condition for the next. If reference-frame support exists, use cut 1 as the reference for later cuts. State in one or two lines what you changed from the bare staging and why; if no staging change was necessary, say that the concept was retained and only timing and camera continuity were specified.
</instructions>
<context>
The deliverable must separate engine-facing content from reader-facing production information visibly.
</context>

## Style rules

Use a hybrid style. Write the cut table, parameters, exclusions, continuity values, and verification items in compact, itemized form. Write each per-cut generation prompt and the brief change rationale as short, sensory narrative instructions. Use precise physical verbs rather than generic cinematic adjectives. Avoid advertising clichés such as “unleash your potential,” “go beyond limits,” “game changer,” “next level,” and unsupported claims about speed, comfort, durability, or performance.

## Style rules (humanizer v1)

These govern every prose surface in the deliverable. Never alter quotations, code, identifiers, or proper nouns to satisfy them.

- Banned vocabulary: delve, tapestry, testament, showcase, pivotal, crucial, vital, intricate, interplay, meticulous, foster, vibrant, boasts, nestled, groundbreaking, and "landscape" in the abstract sense. Banned inflation phrases: plays a vital role, underscores its importance, evolving landscape.
- Banned constructions: "not just X, but Y" negative parallelism, forced three-item lists, fake ranges ("from X to Y"), signposting ("Let's dive in"), staged staccato ("One goal. Zero compromises."), and synonym cycling. Name a thing the same way every time.
- Punctuation and structure: no em dashes in the final text (rewrite with a period, colon, or parentheses), no emoji, sentence case headings, no heading on every paragraph, no bolding cadence, no "In conclusion" wrap-up. Close on a concrete fact.
- Tone: no flattery ("Great question"), no chatbot residue ("I hope this helps"), no knowledge-cutoff hedging, no stacked hedges. Hold the register the genre calls for and vary sentence length.
- Fact integrity: every instruction to be specific carries one boundary. Use only facts present in the user's input or in a verifiable source. Do not invent details to sound human. Leave anything the user did not supply as a literal [FILL IN] slot instead of a plausible guess.
- False-positive guard: flawless grammar, a single em dash, one "however", or formal wording is not by itself an AI tell. Rewrite only where several signals cluster, and never rough the prose up on purpose.

## Final self-audit

Draft the deliverable in full, then interrogate the draft on two counts. Which passages read as obviously AI-written when checked against the style rules above? Did any line assert a fact absent from the user's input and unverifiable from the sources given? Rewrite what fails and submit only the corrected version. The audit itself never appears in your output.

## Self-verification

<instructions>
Before delivering, silently run these checks, correct every failure, and show none of the checking process:

1. Confirm the modality remains video and the concept is treated as a 15-second vertical advertisement.
2. Confirm the runner crosses a city bridge at dawn and that the final cut ends on a close-up of the shoe.
3. Confirm every footstrike-light event has a visible cause-and-effect sequence rather than an unexplained glow.
4. Confirm the chosen number of cuts is minimal and every duration is written in real seconds.
5. Add all cut durations and verify that they equal exactly 15 seconds.
6. Confirm every cut states movement, speed, framing, and its ending frame state.
7. Confirm all six direction topics appear in every per-cut prompt.
8. Confirm the fixed runner and shoe wording is repeated verbatim wherever required.
9. Confirm unspecified shoe, runner, resolution, frame rate, music, and clearance details remain explicit slots or recommendations, never invented values.
10. Confirm no tool name, slash command, double-dash flag, unsupported product claim, or unrelated scene appears.
11. Confirm the exclusions are noun phrases and do not accidentally prohibit the required road-light effect.
12. Confirm the separate-cut joining instructions preserve continuity from one ending frame to the next.
13. Confirm the output remains within the requested advertising scope rather than becoming a script, report, storyboard essay, or product review.
</instructions>