huashu-douyin-script

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

抖音爆款脚本创作

Creating Viral Douyin Video Scripts

从竞品视频拆解到脚本生成的完整工作流。将「凭感觉模仿爆款」变成「AI系统拆解+结构化复制」。
A complete workflow from competitor video breakdown to script generation. Transform "imitating viral videos by feel" into "AI-powered systematic breakdown + structured replication".

环境要求

Environment Requirements

  • uv
    (Python包管理器)
  • yt-dlp
    (抖音视频下载,需
    pip install yt-dlp
    brew install yt-dlp
  • GEMINI_API_KEY
    环境变量(用于Gemini视频分析,从 https://aistudio.google.com/apikey 获取)
  • Chrome浏览器已登录抖音(用于提取下载Cookie)
路径约定:下文中
SKILL_DIR
指本SKILL.md所在目录的绝对路径。运行脚本前,先用
dirname
或 Glob 工具确定SKILL.md的实际位置,替换
SKILL_DIR
  • uv
    (Python package manager)
  • yt-dlp
    (for downloading Douyin videos; install via
    pip install yt-dlp
    or
    brew install yt-dlp
    )
  • GEMINI_API_KEY
    environment variable (for Gemini video analysis; obtain from https://aistudio.google.com/apikey)
  • Chrome browser logged into Douyin (to extract download cookies)
Path Convention: In the following text,
SKILL_DIR
refers to the absolute path of the directory where this SKILL.md is located. Before running the script, use
dirname
or Glob tool to determine the actual location of SKILL.md and replace
SKILL_DIR
.

6步工作流

6-Step Workflow

Step 1: 收集输入

Step 1: Collect Inputs

向用户收集以下信息:
必填
  • 抖音视频链接(1-5个对标/竞品视频)
  • 产品信息(名称、核心卖点、目标人群、价格)
  • 脚本类型:种草视频 or 千川投放素材
选填
  • 产品图片(用于脚本中的视觉参考)
  • 目标时长(种草默认30秒,千川默认15秒)
  • 品牌调性要求
  • 投放目标(直播间引流/商品详情页/关注转化)
如果用户只提供了部分信息,主动询问缺失项。

Collect the following information from the user:
Required:
  • Douyin video links (1-5 benchmark/competitor videos)
  • Product information (name, core selling points, target audience, price)
  • Script type: Grass-planting video or Qianchuan advertising material
Optional:
  • Product images (for visual reference in the script)
  • Target duration (30 seconds by default for grass-planting videos, 15 seconds for Qianchuan material)
  • Brand tone requirements
  • Promotion goals (live stream引流, product detail page, follow conversion)
If the user only provides partial information, proactively ask for the missing items.

Step 2: 下载视频

Step 2: Download Videos

运行下载脚本:
bash
uv run SKILL_DIR/scripts/download_douyin.py \
  --urls "URL1" "URL2" \
  --output-dir _temp/douyin-downloads
参数说明
  • --urls
    :支持短链、页面链接、分享文本(自动提取URL)
  • --output-dir
    :默认当前工作目录下
    _temp/douyin-downloads/
  • --cookies-browser
    :默认chrome,可选edge/firefox
下载失败时
  1. 提示用户确认Chrome已登录抖音
  2. 建议手动下载视频放入
    _temp/douyin-downloads/
  3. 如果有douyin-mcp-server可用,尝试通过MCP下载

Run the download script:
bash
uv run SKILL_DIR/scripts/download_douyin.py \
  --urls "URL1" "URL2" \
  --output-dir _temp/douyin-downloads
Parameter Explanation:
  • --urls
    : Supports short links, page links, and share texts (automatically extracts URLs)
  • --output-dir
    : Defaults to
    _temp/douyin-downloads/
    in the current working directory
  • --cookies-browser
    : Defaults to chrome; edge/firefox are optional
When Download Fails:
  1. Prompt the user to confirm Chrome is logged into Douyin
  2. Suggest manually downloading videos to
    _temp/douyin-downloads/
  3. If douyin-mcp-server is available, try downloading via MCP

Step 3: Gemini视频分析(可并行)

Step 3: Gemini Video Analysis (Parallel Execution Supported)

对每个下载成功的视频,调用 Gemini 视频分析:
bash
uv run SKILL_DIR/scripts/analyze_video.py \
  --video "_temp/douyin-downloads/douyin-1-xxx.mp4" \
  --prompt "PROMPT_BELOW" \
  --model flash \
  --resolution medium \
  --output "_temp/douyin-downloads/analysis-1.md"
使用flash模型 + medium分辨率(短视频足够,节省成本)。
多个视频可并行分析(用
run_in_background=true
)。
爆款分析Prompt(传入
--prompt
参数):
作为资深抖音电商视频分析师,请对这个视频进行7维度深度拆解。
特别注意:前3秒是最关键的部分,需要逐帧级别的详细拆解。
For each successfully downloaded video, call Gemini for video analysis:
bash
uv run SKILL_DIR/scripts/analyze_video.py \
  --video "_temp/douyin-downloads/douyin-1-xxx.mp4" \
  --prompt "PROMPT_BELOW" \
  --model flash \
  --resolution medium \
  --output "_temp/douyin-downloads/analysis-1.md"
Use flash model + medium resolution (sufficient for short videos, cost-effective).
Analyze multiple videos in parallel (set
run_in_background=true
).
Video Analysis Prompt (passed via
--prompt
parameter):
As a senior Douyin e-commerce video analyst, please conduct a 7-dimensional in-depth breakdown of this video.
Special note: The first 3 seconds are the most critical part and require frame-by-frame detailed analysis.

1. 钩子分析(前3秒)⭐ 最关键

1. Hook Analysis (First 3 Seconds) ⭐ Most Critical

前3秒 = 文案钩子 + 视觉钩子,两条线都要详细分析。
First 3 seconds = Copy hook + Visual hook; analyze both lines in detail.

1a. 文案钩子

1a. Copy Hook

  • 钩子类型(反常识/痛点/数据/场景/悬念/利益)
  • 钩子文案完整转录
  • 文案与产品的关联度(强关联/弱关联)
  • Hook type (counter-intuitive/pain point/data/scenario/suspense/benefit)
  • Full transcription of the hook copy
  • Relevance between copy and product (strong/weak)

1b. 视觉钩子(逐帧描述)

1b. Visual Hook (Frame-by-Frame Description)

  • 第0秒画面:暂停在最开始,描述这一帧的构图、色调、主体
  • 视觉钩子类型(视觉奇观/产品堆叠/画面突变/动作冲击/文字弹出/第一人称POV)
  • 具体画面描述:景别、色调、运镜方式、特效、人物表情和动作
  • 视觉信息密度:画面中同时出现了多少信息元素
  • Frame 0: Pause at the start, describe the composition, color tone, and subject
  • Visual hook type (visual spectacle/product stacking/sudden frame change/action impact/text pop-up/first-person POV)
  • Detailed frame description: shot type, color tone, camera movement, special effects, character expressions and actions
  • Visual information density: number of information elements appearing simultaneously in the frame

1c. 文案×视觉配合

1c. Copy × Visual Coordination

  • 文案和画面是否同步强化同一个信息
  • 声音设计:前3秒有无音效/BGM变化/人声语气变化
  • 综合钩子强度评分(1-10)及理由
  • Whether copy and frame reinforce the same message synchronously
  • Sound design: presence of sound effects/BGM changes/vocal tone changes in the first 3 seconds
  • Comprehensive hook strength score (1-10) with reasons

2. 分镜结构

2. Storyboard Structure

用表格拆解每个镜头: | 时间 | 景别 | 画面内容 | 旁白/字幕 | 转场方式 |
Break down each shot using a table: | Time | Shot Type | Frame Content | Voiceover/Subtitles | Transition Method |

3. 节奏设计

3. Rhythm Design

  • 快切片段时间点
  • 节奏变化模式(快-慢-快/渐进加速/脉冲式)
  • BGM与画面的配合度
  • Time points of quick-cut segments
  • Rhythm change pattern (fast-slow-fast/gradual acceleration/pulsed)
  • Coordination between BGM and frames

4. 视觉元素

4. Visual Elements

  • 色调和滤镜风格
  • 字幕样式和出现时机
  • 产品展示方式
  • 构图特点
  • Color tone and filter style
  • Subtitle style and appearance timing
  • Product display method
  • Composition characteristics

5. 转化设计

5. Conversion Design

  • CTA类型和出现时机
  • 价格锚定策略
  • 紧迫感营造手段
  • 转化路径(视频→?→?)
  • CTA type and appearance timing
  • Price anchoring strategy
  • Urgency-building methods
  • Conversion path (Video → ? → ?)

6. 合规检查

6. Compliance Check

  • 是否存在绝对化用语
  • 功效表述是否合规
  • 有无潜在审核风险
  • Presence of absolute terms
  • Compliance of efficacy claims
  • Potential audit risks

7. 可复制要素

7. Replicable Elements

  • 可直接复制的:钩子句式、分镜结构、视觉风格、CTA话术
  • 不可复制的:独特场景、特定达人、品牌势能
请用中文输出,尽量详细。

---
  • Directly replicable: hook sentence patterns, storyboard structure, visual style, CTA scripts
  • Non-replicable: unique scenarios, specific influencers, brand momentum
Please output in English and be as detailed as possible.

---

Step 4: 爆款公式提炼 ← 用户确认节点

Step 4: Extract Viral Formulas ← User Confirmation Node

有竞品视频时 → 汇总分析结果:
读取所有
analysis-*.md
文件,同时读取
references/proven-formulas.md
中的已验证公式作为对照,由Claude汇总提炼:
  1. 品类爆款共性:多个视频共同的成功要素
  2. 钩子公式:最有效的钩子类型和句式模板
  3. 最佳分镜结构:适合该品类的标准分镜流程
  4. 节奏模板:推荐的节奏设计方案
  5. 转化路径:最优的CTA策略
  6. 风险提示:需要避免的合规问题
无竞品视频时 → 直接使用已验证公式:
读取
references/proven-formulas.md
,根据产品品类选择最接近的品类公式作为基线。提醒用户:无竞品分析的脚本可能缺少品类针对性,建议后续补充竞品视频。
保存汇总
_temp/douyin-downloads/[产品名]-爆款分析汇总.md
展示给用户确认
  • 确认分析是否准确
  • 确认脚本类型(种草/千川)
  • 补充产品信息(如果Step 1未完整提供)
公式沉淀:分析完成后,将新品类的公式追加到
references/proven-formulas.md
,持续积累。

With competitor videos → Summarize analysis results:
Read all
analysis-*.md
files, and cross-reference with verified formulas in
references/proven-formulas.md
, then use Claude to summarize and extract:
  1. Category-wide viral commonalities: Shared success factors across multiple videos
  2. Hook formulas: Most effective hook types and sentence templates
  3. Optimal storyboard structure: Standard storyboard process suitable for the category
  4. Rhythm template: Recommended rhythm design scheme
  5. Conversion path: Optimal CTA strategy
  6. Risk warnings: Compliance issues to avoid
Without competitor videos → Use verified formulas directly:
Read
references/proven-formulas.md
, select the closest category formula as the baseline based on the product category. Remind the user: Scripts without competitor analysis may lack category relevance; it is recommended to supplement competitor videos later.
Save summary to
_temp/douyin-downloads/[Product Name]-Viral Analysis Summary.md
Present to user for confirmation:
  • Confirm accuracy of the analysis
  • Confirm script type (grass-planting/Qianchuan)
  • Supplement product information (if not fully provided in Step 1)
Formula accumulation: After analysis, append formulas for new categories to
references/proven-formulas.md
for continuous accumulation.

Step 5: 脚本+分镜生成

Step 5: Generate Script + Storyboard

基于爆款公式 + 产品信息 + 口语风格样本生成脚本。
生成前准备
  1. 读取
    references/script-style-samples.md
    获取口语风格范例
  2. 读取
    references/analysis-dimensions.md
    中的种草/千川区别表
  3. 读取Step 4的爆款分析汇总
脚本生成Prompt模板
Generate scripts based on viral formulas + product information + colloquial style samples.
Preparations before generation:
  1. Read
    references/script-style-samples.md
    to get colloquial style examples
  2. Read the grass-planting vs Qianchuan comparison table in
    references/analysis-dimensions.md
  3. Read the viral analysis summary from Step 4
Script Generation Prompt Templates:

种草视频脚本

Grass-Planting Video Script

你是抖音电商短视频编导,擅长带货类种草内容。
You are a Douyin e-commerce short video director, specializing in product promotion grass-planting content.

任务

Task

为以下产品生成一个{时长}秒的种草视频脚本。
Generate a {duration}-second grass-planting video script for the following product.

产品信息

Product Information

{产品名称、核心卖点、目标人群、价格、品牌信息}
{Product name, core selling points, target audience, price, brand information}

爆款公式(基于竞品分析)

Viral Formulas (Based on Competitor Analysis)

{Step 4提炼的爆款公式}
{Viral formulas extracted in Step 4}

口语风格要求

Colloquial Style Requirements

参考以下真实脚本的风格特征:
  • 极度口语化:用说话的方式写字,短句为主(不超过15字)
  • 具体数据锚点:用具体数字,不用「很多」「大量」
  • 感官描述:调动视觉/触觉/味觉/嗅觉
  • 情感触发:身份认同、焦虑激活、场景共鸣
  • 品牌背书简化:一句话带过
  • 价格锚定+紧迫感
真实脚本范例: {从script-style-samples.md选取同品类或风格接近的1-2个样本}
Refer to the style features of the following real scripts:
  • Extremely colloquial: Write like speaking, use short sentences (no more than 15 words)
  • Specific data anchors: Use specific numbers instead of "many"/"a large number"
  • Sensory descriptions: Engage visual/tactile/gustatory/olfactory senses
  • Emotional triggers: Identity recognition, anxiety activation, scenario resonance
  • Simplified brand endorsement: Mention in one sentence
  • Price anchoring + urgency
Real script examples: {Select 1-2 samples from script-style-samples.md that match the category or style}

输出要求

Output Requirements

Part A:完整脚本

Part A: Complete Script

  • 开头3秒必须有强钩子(参考分析中最有效的钩子类型)
  • 中间展示核心卖点(最多3个,多了观众记不住)
  • 结尾有明确CTA
  • 全文口语化,像朋友推荐
  • 不使用绝对化用语,符合抖音广告审核规范
  • 标注每段预估时长
  • Must have a strong hook in the first 3 seconds (refer to the most effective hook type in the analysis)
  • Highlight core selling points in the middle (max 3, as audiences can't remember more)
  • Clear CTA at the end
  • Fully colloquial, like a friend's recommendation
  • No absolute terms, comply with Douyin advertising review regulations
  • Mark estimated duration for each segment

Part B:分镜表格(前3秒必须详细描述视觉钩子)

Part B: Storyboard Table (Detailed description of visual hook required for first 3 seconds)

| 镜号 | 时长 | 景别 | 画面内容 | 旁白/字幕 | 拍摄建议 |
⚠️ 前3秒的分镜必须包含:
  • 视觉钩子类型(视觉奇观/产品堆叠/画面突变/动作冲击/文字弹出/第一人称POV)
  • 具体画面描述:构图、色调、运镜、特效、表情动作
  • 第一帧设计:暂停在第0秒时这一帧能否吸引点击
  • 文案×画面配合:声音和画面是否同步强化同一个信息
undefined
| Shot No. | Duration | Shot Type | Frame Content | Voiceover/Subtitles | Shooting Suggestions |
⚠️ The first 3 seconds of the storyboard must include:
  • Visual hook type (visual spectacle/product stacking/sudden frame change/action impact/text pop-up/first-person POV)
  • Detailed frame description: composition, color tone, camera movement, special effects, expressions and actions
  • First frame design: Can this frame attract clicks when paused at second 0?
  • Copy × frame coordination: Whether sound and frame reinforce the same message synchronously
undefined

千川投放素材脚本

Qianchuan Advertising Material Script

你是千川投放素材编导,擅长高转化信息流广告。
You are a Qianchuan advertising material director, specializing in high-conversion feed ads.

任务

Task

为以下产品写一条{时长}秒千川投放素材脚本。
Write a {duration}-second Qianchuan advertising material script for the following product.

产品信息

Product Information

{产品名称、核心卖点、目标人群、价格、投放目标}
{Product name, core selling points, target audience, price, promotion goals}

爆款公式(基于竞品分析)

Viral Formulas (Based on Competitor Analysis)

{Step 4提炼的爆款公式}
{Viral formulas extracted in Step 4}

口语风格要求

Colloquial Style Requirements

{同种草脚本的风格要求}
{Same as grass-planting script style requirements}

千川素材特殊要求

Special Requirements for Qianchuan Materials

  • 前3秒必须有强钩子(数据型或痛点型,不要用问句开头)
  • 信息密集,卖点直给
  • CTA要硬:明确引导点击/进直播间/下单
  • 不使用绝对化用语,不做未经验证的功效承诺
  • 符合千川审核规范
  • Must have a strong hook in the first 3 seconds (data or pain point type; avoid question openings)
  • Information-dense, direct selling points
  • Hard CTA: Clearly guide clicks/enter live stream/place order
  • No absolute terms, no unproven efficacy claims
  • Comply with Qianchuan review regulations

输出

Output

Part A:完整脚本(标注时长)

Part A: Complete Script (with duration marks)

Part B:分镜表格

Part B: Storyboard Table

| 镜号 | 时长 | 画面内容 | 旁白/字幕 | 素材建议 |

**保存脚本**到 `_temp/douyin-downloads/[产品名]-种草脚本.md` 或 `[产品名]-千川素材脚本.md`

---
| Shot No. | Duration | Frame Content | Voiceover/Subtitles | Material Suggestions |

**Save script** to `_temp/douyin-downloads/[Product Name]-Grass-Planting Script.md` or `[Product Name]-Qianchuan Material Script.md`

---

Step 6: AI味审校 ← 用户确认节点

Step 6: AI-Style Proofreading ← User Confirmation Node

对生成的脚本进行视频脚本专用审校,5项检查:
Conduct dedicated proofreading for video scripts with 5 checks:

检查1:口语化程度

Check 1: Colloquial Level

参考
视频脚本口语化审校
skill的规则:
  • ❌ 「接下来我们将进行...」→ ✅ 「我们来...」
  • ❌ 「进行操作」→ ✅ 「点击」
  • ❌ 「综上所述」→ ✅ 直接总结
  • 短句为主(不超过15字)
  • 大声读一遍,卡顿的地方重写
Refer to the rules of the "video script colloquial proofreading" skill:
  • ❌ "Next, we will proceed to..." → ✅ "Let's..."
  • ❌ "Perform operation" → ✅ "Click"
  • ❌ "In summary" → ✅ Directly summarize
  • Use short sentences (no more than 15 words)
  • Read aloud; rewrite any awkward parts

检查2:钩子强度

Check 2: Hook Strength

  • 前3秒是否有足够吸引力
  • 是否与目标人群匹配
  • 是否制造了好奇心/紧迫感/共鸣
  • Whether the first 3 seconds are attractive enough
  • Whether it matches the target audience
  • Whether it creates curiosity/urgency/resonance

检查3:节奏与时长

Check 3: Rhythm and Duration

  • 每段时长标注是否合理
  • 总时长是否在目标范围内
  • 信息密度是否适中(不过载也不空洞)
  • Whether the duration marks for each segment are reasonable
  • Whether the total duration is within the target range
  • Whether the information density is moderate (neither overloaded nor empty)

检查4:广审合规检查(重点!)

Check 4: Advertising Compliance Check (Key!)

这是最容易导致素材被拒的环节。即使产品官方宣传语中有某些表述,投放时也可能被平台审核拒绝。
4a. 绝对化用语(必须删除):
  • ❌ 最好/第一/100%/绝对/唯一/顶级/全网最低
4b. 时间+功效承诺(高危!最常被卡):
  • ❌ 「60秒缓解敏感」→ ✅ 「用完嘴巴就舒服多了」
  • ❌ 「28天淡斑83%」→ ✅ 「坚持用,斑真的有变淡」
  • ❌ 「7天见效」→ ✅ 「用了一阵子就能看到变化」
  • 规则:具体时间 + 具体功效 = 必须改写。去掉时间或去掉功效承诺,用模糊化的体感描述代替
  • 即使数据来自产品官方/临床报告,投放素材中也不能直接使用
4c. 功效承诺措辞
  • ❌ 治好/根治/永久/彻底解决 → ✅ 缓解/帮助改善/感觉好多了
  • ❌ 美白/祛痘/抗衰 → ✅ 提亮/改善痘痘肌/让皮肤状态更好(化妆品类需特别注意)
  • ❌ 消炎/杀菌/抑菌 → ✅ 感觉清爽干净(非药品不能用医疗术语)
4d. 对比和贬损
  • ❌ 「比XX品牌好」「吊打所有竞品」
  • ❌ 「医院要花几千块」→ ⚠️ 可以用但不能暗示替代医疗
4e. 诱导和虚假
  • ❌ 「不买就亏了」「错过再等一年」(虚假紧迫感)
  • ❌ 未标注的演员扮演使用者/专家
改写原则:用第一人称体感代替第三人称数据。「我用了之后感觉...」比「经临床验证...」安全得多。
This is the most common reason for material rejection. Even if certain expressions are in the official product宣传语, they may be rejected by the platform during review.
4a. Absolute Terms (must be removed):
  • ❌ Best/First/100%/Absolutely/Only/Top/Lowest price online
4b. Time + Efficacy Claims (high risk! Most frequently rejected):
  • ❌ "Relieves sensitivity in 60 seconds" → ✅ "My mouth feels much better after using it"
  • ❌ "Fades spots by 83% in 28 days" → ✅ "With consistent use, spots really fade"
  • ❌ "Visible results in 7 days" → ✅ "You can see changes after using it for a while"
  • Rule: Specific time + specific efficacy = must be rewritten. Remove either the time or the efficacy claim, replace with vague sensory descriptions
  • Even if data comes from official product/clinical reports, it cannot be used directly in promotional materials
4c. Efficacy Claim Wording:
  • ❌ Cure/Eradicate/Permanent/Completely solve → ✅ Relieve/Help improve/Feel much better
  • ❌ Whitening/Anti-acne/Anti-aging → ✅ Brighten/Improve acne-prone skin/Improve skin condition (special attention for cosmetics)
  • ❌ Anti-inflammatory/Bactericidal/Antibacterial → ✅ Feel fresh and clean (medical terms cannot be used for non-pharmaceutical products)
4d. Comparison and Derogation:
  • ❌ "Better than XX brand" "Outperforms all competitors"
  • ❌ "Costs thousands in hospitals" → ⚠️ Can be used but cannot imply replacement of medical treatment
4e. Inducement and Falsehood:
  • ❌ "You'll lose out if you don't buy" "Wait another year if you miss this" (false urgency)
  • ❌ Unlabeled actors playing users/experts
Rewriting Principle: Use first-person sensory experiences instead of third-person data. "I felt... after using it" is much safer than "Clinically proven...".

检查5:AI味检查

Check 5: AI-Style Check

参考
ai-taste-proofreading
skill的6大AI腔规则:
  • ❌ 套话连篇:「在当今时代」「值得注意的是」
  • ❌ AI句式:「不是...而是...」「不仅...而且...」连续使用
  • ❌ 书面词汇:「显著提升」「充分利用」
  • ❌ 结构机械:过度使用「首先、其次、最后」
  • ❌ 态度中立:「具体取决于实际情况」
  • ❌ 细节缺失:大词多、具体细节少
审校后展示修改清单给用户确认,然后应用修改。

Refer to the 6 AI-style rules in the
ai-taste-proofreading
skill:
  • ❌ Clichés: "In today's era" "It is worth noting that"
  • ❌ AI sentence patterns: Overuse of "not...but..." "not only...but also..."
  • ❌ Written vocabulary: "Significantly improve" "Make full use of"
  • ❌ Mechanical structure: Overuse of "First, Second, Finally"
  • ❌ Neutral attitude: "Depends on actual situation"
  • ❌ Lack of details: Too many big words, too few specific details
After proofreading: Present the revision list to the user for confirmation, then apply the changes.

文件保存规则

File Saving Rules

所有中间产物和最终输出保存在
写作/_temp/douyin-downloads/
_temp/douyin-downloads/
├── douyin-1-[ID].mp4             # 下载的视频
├── douyin-2-[ID].mp4
├── analysis-1.md                  # 视频分析结果
├── analysis-2.md
├── [产品名]-爆款分析汇总.md       # 多视频汇总
├── [产品名]-种草脚本.md           # 最终种草脚本
└── [产品名]-千川素材脚本.md       # 最终千川脚本
All intermediate products and final outputs are saved in
writing/_temp/douyin-downloads/
:
_temp/douyin-downloads/
├── douyin-1-[ID].mp4             # Downloaded videos
├── douyin-2-[ID].mp4
├── analysis-1.md                  # Video analysis results
├── analysis-2.md
├── [Product Name]-Viral Analysis Summary.md       # Multi-video summary
├── [Product Name]-Grass-Planting Script.md           # Final grass-planting script
└── [Product Name]-Qianchuan Material Script.md       # Final Qianchuan script

参考资源

Reference Resources

资源路径用途
分析维度框架
references/analysis-dimensions.md
7维度分析+种草vs千川
已验证爆款公式
references/proven-formulas.md
品类公式库,无竞品视频时的基线
口语风格样本
references/script-style-samples.md
脚本生成的few-shot范例
视频下载脚本
scripts/download_douyin.py
抖音视频下载(yt-dlp)
Gemini视频分析
scripts/analyze_video.py
视频理解和拆解(内置)
AI味审校规则Step 6中内嵌6大AI腔识别
口语化审校Step 6中内嵌口语化改写规则
ResourcePathPurpose
Analysis dimension framework
references/analysis-dimensions.md
7-dimensional analysis + grass-planting vs Qianchuan comparison
Verified viral formulas
references/proven-formulas.md
Category formula library; baseline when no competitor videos are available
Colloquial style samples
references/script-style-samples.md
Few-shot examples for script generation
Video download script
scripts/download_douyin.py
Douyin video download (yt-dlp based)
Gemini video analysis
scripts/analyze_video.py
Video understanding and breakdown (built-in)
AI-style proofreading rulesEmbedded in Step 66 AI-style identification rules
Colloquial proofreadingEmbedded in Step 6Colloquial rewriting rules

快速使用

Quick Start

有竞品视频时(「帮我拆解这几个抖音视频,生成一个脚本」):
  1. 收集链接和产品信息(Step 1)
  2. 下载+分析可并行执行(Step 2-3)
  3. 展示分析汇总,确认方向(Step 4)
  4. 生成脚本,确认后审校(Step 5-6)
  5. 将新品类公式追加到
    references/proven-formulas.md
无竞品视频时(「帮我写个XX产品的种草脚本」):
  1. 收集产品信息(Step 1,跳过视频链接)
  2. 跳过 Step 2-3
  3. references/proven-formulas.md
    选取品类公式(Step 4)
  4. 生成脚本,确认后审校(Step 5-6)
总耗时:有视频约15-20分钟,无视频约5-10分钟。

花叔出品 | AI Native Coder · 独立开发者 公众号「花叔」| 30万+粉丝 | AI工具与效率提升 代表作:小猫补光灯(AppStore付费榜Top1)·《一本书玩转DeepSeek》
With competitor videos ("Help me break down these Douyin videos and generate a script"):
  1. Collect links and product information (Step 1)
  2. Run download + analysis in parallel (Step 2-3)
  3. Present analysis summary and confirm direction (Step 4)
  4. Generate script, then proofread after confirmation (Step 5-6)
  5. Append formulas for new categories to
    references/proven-formulas.md
Without competitor videos ("Help me write a grass-planting script for XX product"):
  1. Collect product information (Step 1, skip video links)
  2. Skip Step 2-3
  3. Select category formula from
    references/proven-formulas.md
    (Step 4)
  4. Generate script, then proofread after confirmation (Step 5-6)
Total time: ~15-20 minutes with videos, ~5-10 minutes without videos.

Produced by Uncle Hua | AI Native Coder · Independent Developer Official Account "Uncle Hua" | 300,000+ followers | AI tools and efficiency improvement Masterpieces: Kitten Fill Light (Top 1 in AppStore Paid Charts) · Mastering DeepSeek in One Book