before-after-ad

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Before-After Ad

前后对比广告

Purpose

目的

Create approval-ready before-and-after ad concepts and shot scripts. Make the starting state inspectable, show a plausible product-use behavior, and resolve with controlled visual evidence rather than an unsupported transformation claim.

Produce a

Before-After Grammar Plan

, shot script, one approval checkpoint, and direct downstream handoff after approval.

创建可获批的前后对比广告概念和分镜脚本。明确展示初始状态，呈现合理的产品使用行为，并以可控的视觉证据收尾，而非无依据的转型宣称。

产出一份

Before-After Grammar Plan

、分镜脚本、一个审批检查点，以及获批后的直接下游交接内容。

Workflow

工作流程

1. Ground The Brief

1. 明确brief核心

Read the smallest useful source context first:

product facts, approved claims, usage instructions, and substantiation
audience, persona, concern, desired visible result, and duration
product, creator, scene, audio, and style references
prior reference-video analysis or benchmark notes when relevant

If the user supplies raw reference videos without an analysis, use

video-analysis

first. Do not infer style from filenames, thumbnails, or a single frame.

Lock these before writing:

```
starting_state
```
```
desired_result_state
```
```
pressure_carrier
```
```
product_capability
```
```
approved_proof
```
```
claim_boundary
```
```
persona_and_concern_map
```
```
product_appearance
```
```
duration_and_format
```
```
reference_duties
```

Ask only when a missing fact would change the product, proof, claim, identity, or workflow. Do not invent treatment timelines, dosage, medical causality, guaranteed outcomes, or product appearance.

优先阅读最有价值的源上下文：

产品事实、已批准的宣称、使用说明及佐证材料
受众、用户画像、关注点、期望可见效果及周期
产品、创作者、场景、音频及风格参考
相关的过往参考视频分析或基准笔记

若用户提供未做分析的原始参考视频，需先使用

video-analysis

工具。请勿从文件名、缩略图或单帧画面推断风格。

在开始撰写前锁定以下内容：

```
starting_state
```
（初始状态）
```
desired_result_state
```
（期望结果状态）
```
pressure_carrier
```
（压力载体）
```
product_capability
```
（产品能力）
```
approved_proof
```
（已批准证据）
```
claim_boundary
```
（宣称边界）
```
persona_and_concern_map
```
（用户画像与关注点映射）
```
product_appearance
```
（产品外观）
```
duration_and_format
```
（时长与格式）
```
reference_duties
```
（参考职责）

仅当缺失的信息会改变产品、证据、宣称、品牌标识或工作流程时才询问用户。请勿编造治疗 timeline、剂量、医学因果关系、保证效果或产品外观。

2. Design The Proof Grammar

2. 设计证据语法

Read

references/style-grammar.md

Produce a

Before-After Grammar Plan

with these fields:

```
core_human_state
```
```
pressure_carrier
```
```
baseline_evidence
```
```
comparison_control
```
```
product_entrance
```
```
product_gesture
```
```
result_evidence
```
```
camera_language
```
```
framing_rules
```
```
movement_rules
```
```
edit_rhythm
```
```
sound_mode
```
```
text_mode
```
```
claim_boundary
```
```
continuity_locks
```
```
generator_risks
```

Choose one primary proof route:

```
direct comparison
```
: baseline → use → matched result → optional split view
```
discovery
```
: external trigger → inspection → use → result
```
routine montage
```
: concern → repeated normal use → result portrait

Do not copy a reference creator, exact scene, wording, music, brand marks, or recognizable personal details. Transfer only reusable camera, movement, edit, proof, sound, and text behavior.

阅读

references/style-grammar.md

文件。

产出包含以下字段的

Before-After Grammar Plan

：

```
core_human_state
```
（核心用户状态）
```
pressure_carrier
```
（压力载体）
```
baseline_evidence
```
（基准证据）
```
comparison_control
```
（对比控制）
```
product_entrance
```
（产品出场方式）
```
product_gesture
```
（产品使用动作）
```
result_evidence
```
（结果证据）
```
camera_language
```
（镜头语言）
```
framing_rules
```
（构图规则）
```
movement_rules
```
（运镜规则）
```
edit_rhythm
```
（剪辑节奏）
```
sound_mode
```
（音效模式）
```
text_mode
```
（文本样式）
```
claim_boundary
```
（宣称边界）
```
continuity_locks
```
（连贯性锁定项）
```
generator_risks
```
（生成风险）

选择一种核心证据呈现路径：

```
direct comparison
```
（直接对比）：基准状态→产品使用→匹配结果→可选分屏展示
```
discovery
```
（发现式）：外部触发→状态检查→产品使用→结果展示
```
routine montage
```
（流程蒙太奇）：痛点展示→日常重复使用→结果特写

请勿复制参考创作者的身份、具体场景、文案、音乐、品牌标识或可识别的个人细节。仅借鉴可复用的镜头、运镜、剪辑、证据呈现、音效及文本表现手法。

3. Write The Shot Script

3. 撰写分镜脚本

Read

references/shot-script-rubric.md

Return a shot table with exactly these columns:

time | visual | framing/composition | camera motion | action | sound | subtitle/text | proof duty | purpose

After the table, return

Production Locks

with:

```
persona_lock
```
```
concern_map_lock
```
```
product_lock
```
```
scene_lock
```
```
before_after_match_lock
```
```
audio_lock
```
```
claim_lock
```
```
post_only_elements
```

Keep each shot physically executable. Use one clear action per shot. Treat captions, arrows, circles, split-screen assembly, end cards, disclaimers, and comparison labels as post-production elements rather than generated scene content.

After the script, ask whether the direction is approved. Stop before media generation unless the user approves or explicitly asks to continue into production in the same request.

阅读

references/shot-script-rubric.md

文件。

返回包含以下固定列的分镜表格：

时间 | 视觉内容 | 构图/布局 | 镜头运动 | 动作 | 音效 | 字幕/文本 | 证据职责 | 目的

表格后需返回

Production Locks

（生产锁定项），包含：

```
persona_lock
```
（用户画像锁定）
```
concern_map_lock
```
（关注点映射锁定）
```
product_lock
```
（产品锁定）
```
scene_lock
```
（场景锁定）
```
before_after_match_lock
```
（前后对比匹配锁定）
```
audio_lock
```
（音频锁定）
```
claim_lock
```
（宣称锁定）
```
post_only_elements
```
（仅后期制作元素）

确保每个镜头在实际中可执行。每个镜头仅包含一个清晰动作。将字幕、箭头、圈注、分屏拼接、片尾卡片、免责声明及对比标签视为后期制作元素，而非生成场景的内容。

脚本完成后，询问用户是否批准该方案。除非用户批准或在同一请求中明确要求进入制作环节，否则在生成媒体内容前停止操作。

4. Handoff After Approval

4. 获批后交接

Read

references/downstream-handoff.md

and hand off directly:

Use
```
image-batch-runner
```
first when persona, product lifestyle, concern-map, matched before/after, first-frame, or per-clip scene images are needed.
Use
```
video-batch-runner
```
when the approved script and required hosted assets already exist and the user wants renders. Pass the approved plan, script, locks, facts, claim boundary, source basis, and reference duties without a redundant brief or second confirmation.

Let the downstream skill own model selection, request schema, upload rules, render execution, polling, and manifests.

阅读

references/downstream-handoff.md

文件并直接交接：

当需要用户画像、产品生活化场景、关注点映射图、匹配的前后对比画面、首帧或单片段场景图片时，优先使用
```
image-batch-runner
```
当已获批脚本及所需托管资产已存在，且用户需要渲染内容时，使用
```
video-batch-runner
```
无需重复提交brief或二次确认，直接传递已获批方案、脚本、锁定项、事实、宣称边界、源依据及参考职责。

下游工具将负责模型选择、请求 schema、上传规则、渲染执行、轮询及清单管理。

Quality Bar

质量标准

Make the problem inspectable before a normal product-use gesture.
Match proof-critical before and after framing, posture, light, and concern area.
Put the emotional shift after evidence and preserve natural texture.
Use one restrained text system, hard cuts, and subject-driven motion.
Separate observed state, testimony, and substantiated product facts.

在常规产品使用动作前，清晰展示问题状态
确保证据关键的前后画面在构图、姿势、光线及痛点区域上保持一致
情绪转变需在证据展示之后，保留自然质感
使用简洁的文本系统、硬切转场及主体驱动的运镜
区分观察状态、用户证言及已佐证的产品事实

Stop Conditions

停止条件

Stop when required product appearance lacks an approved reference or creation permission.
Stop when the requested result or timeline exceeds the approved claim boundary.
Stop when comparison cannot be honest or references contain no transferable grammar.

当所需产品外观缺乏已批准的参考或创作权限时，停止操作
当请求的结果或周期超出已批准的宣称边界时，停止操作
当无法进行诚实对比或参考内容无可复用的语法逻辑时，停止操作