Tiktok Ugc Seedance

TikTok UGC To Seedance Workflow

默认用中文输出。英文脚本、英文口播和 Seedance Prompt 中的英文台词保持英文。

Core Goal

把一个目标商品和一个对标 TikTok 视频，转成可执行的短视频生产包：

目标产品分析 -> 品类关键词 -> 对标视频解析 + 结构拆解 -> UGC 脚本仿写 -> Seedance Prompt 编译

核心原则：先理解目标商品，再拆对标视频；迁移转化结构，不照抄原话；每个核心卖点都必须有画面动作或特写证明。

When To Use

Use this skill when the user asks for any of these:

• 分析商品并找 TikTok 对标方向。
• 拆解一个 TikTok Shop / UGC / try-on / unboxing / review 视频。
• 结合商品和对标视频，仿写 15 秒英文 UGC 口播脚本。
• 设计高光黄金动作、Wow Factor、Close-up 特写目标。
• 把 UGC 脚本转换成 Seedance 2.0 / Seedance 2.0 Fast prompt。

If the user only asks for a generic Seedance prompt without product-vs-competitor workflow, use seedance-prompt-compiler instead. If the user asks for broad TikTok/EchoTik batch research, scoring, or 100-video collection, use fashion-tiktok-video instead.

Inputs

Expected inputs, not all mandatory at once:

• Target product: product URL, product page text, product photos, three-view sheet, or user-provided product facts.
• Optional competitor TikTok URL.
• Optional constraints: duration, language, subtitle/no-subtitle, filming device, target platform, target market, brand name, CTA style.

Defaults:

• Platform: TikTok / Reels / Shorts.
• Market: US cross-border ecommerce.
• Script duration: 15 seconds.
• Script language: English.
• Style: realistic UGC, natural light, handheld phone footage.
• Device prompt: iPhone-style phone footage if unspecified; use the user's exact phone model if provided.

Workflow

0. Input Intake

Collect and restate:

• Product URL/images/facts.
• Whether a competitor TikTok link is already provided.
• Desired output: analysis only, script only, Seedance prompt only, or full package.
• Hard constraints, especially subtitles, logos, visible text, product color, duration, and filming device.

Do not stop to ask questions unless a missing detail would materially change the output. Make reasonable defaults explicit.

Input Intake is a preflight step, not a numbered human approval gate. The numbered workflow below has four mandatory human confirmation gates.

Human Confirmation Gates

For steps 1-4, stop after producing that step's output and ask for user confirmation before continuing. Do not proceed automatically across gates unless the user explicitly asks to skip confirmation for that specific run.

• Gate 1: after product positioning, category keywords, pain points, and Feature/Benefit/Visual Proof are produced.
• Gate 2: after competitor transcript, visual understanding, conversion breakdown, and reusable actions are produced.
• Gate 3: after the UGC script, hook logic, golden action, Wow Factor, Close-up targets, and shot plan are produced.
• Gate 4: after the Seedance prompt is produced; confirm product fidelity, camera actions, text/subtitle policy, and negative constraints before using it for video generation.

1. Target Product Analysis First

Analyze the target product before the competitor video.

Required dimensions:

1. 产品定位

• 目标核心人群。
• 核心标签：提炼 3 个最核心的短视频分类标签，决定流量池和关键词。
• 价格与心智定位：判断视频应拼价格痛点，还是拼生活方式审美、质感、场景、身份状态。

1. 品类关键词

• 结合目标产品，给出用户可用于搜索对标视频的关键词。
• Include product terms, scene terms, emotion/style terms, and TikTok Shop commerce terms.
• Prefer natural creator phrases over SEO-only phrases.

1. 用户痛点层

• 场景化显性痛点：可转成前 3 秒黄金 Hook。
• 心理化隐性痛点：可转成中段情绪共鸣与痛点切入。

1. 产品卖点层

• Feature：产品客观属性，如材质、版型、口袋、领口、颜色、价格、配送。
• Benefit：把属性转成用户大白话好处。
• Visual proof：每个重要 Benefit 对应一个可拍动作或特写。

Output table:

````md

目标产品分析

产品定位

| 维度 | 结论 | |---|---| | 目标核心人群 | | | 核心短视频标签 | | | 价格与心智定位 | |

品类关键词

| 类型 | 关键词 | 用途 | |---|---|---| | Product words | | | | Scene words | | | | Emotion / style words | | | | TikTok Shop commerce words | | |

用户痛点

| 类型 | 痛点 | 可转成的视频表达 | |---|---|---| | 显性痛点 | | | | 隐性痛点 | | |

Feature / Benefit / Visual Proof

| Feature | Benefit | 视频证明方式 | |---|---|---| | | | |

````

Human gate: stop here and ask the user to confirm product positioning, category keywords, pain points, and selling-point direction. Continue to competitor video parsing only after confirmation or revision instructions.

2. Competitor Video Parsing + Structure Breakdown

When a TikTok URL is provided, use the fast extraction path first:

python scripts/extract_tiktok_reference.py "TIKTOK_URL" --out-dir /tmp/tiktok_reference

The script standardizes:

• yt_dlp metadata and video download.
• Subtitle extraction if available.
• Audio extraction + whisper-cli transcription when subtitles are unavailable.
• ffmpeg contact sheet and sampled frames for visual understanding.

If the script fails because TikTok blocks extraction, fall back to browser/web-access viewing and explain the limitation briefly.

Breakdown output must follow this structure:

````md

对标视频分析

这是一条【视频类型】风格视频，重点解决用户对该类商品的哪些购买顾虑。

Hook 钩子

• 时间：
• 内容：
• 分析：
• 对应结构节点：

Product Demo 产品演示

• 时间：
• 内容：
• 分析：
• 对应结构节点：

Trust / Experience 信任与使用体验

• 时间：
• 内容：
• 分析：
• 对应结构节点：

Value Proposition 价值主张

• 时间：
• 内容：
• 分析：
• 对应结构节点：

CTA 行动号召

• 时间：
• 内容：
• 分析：
• 对应结构节点：

结构连线

节点 -> 节点 -> 节点

画面理解

• 场景：
• 人物：
• 产品展示动作：
• 转化证明点：

给跨境卖家的启示

• ...

Script Extraction

[00:00.00 - 00:00.00] ...

````

Structure nodes are based on the user's TikTok conversion map:

• 吸引用户：明星直播、针对痛点、价格诱惑、效果承诺、使用效果、新鲜有趣、情感共鸣、人物设定。
• 取得信任：明星背书、权威认证、专业讲解、使用体验、口碑销量、高性价比、温馨场景。
• 再次吸引 / 转化：限时活动、不买就亏、固定人群、命令式、暗示购买、购买保障、数量限定。

Treat the nodes as a combinable graph, not a fixed linear funnel. Example:

针对痛点 -> 使用体验 -> 专业讲解 -> 暗示购买

Human gate: stop here and ask the user to confirm the competitor breakdown, structure path, reusable hooks/actions, and seller insights. Continue to UGC script transfer only after confirmation or revision instructions.

3. UGC Script Transfer

Write the target-product UGC script by combining:

• Target product positioning.
• Product pain points and benefits.
• Competitor video's conversion structure.
• TikTok UGC voiceover rhythm.

Requirements:

• Default duration: 15 seconds.
• Default language: English.
• Use natural creator speech, not polished ad copy.
• Do not blindly copy competitor lines.
• Include visual action design; the image must make the product benefit obvious even if audio is muted.

Output:

````md

UGC 脚本

[0:00-0:03]
...

中文解释

...

脚本骨架说明

| 时间 | 作用 | 内容策略 | 对应结构节点 | |---|---|---|---|

高光黄金动作

...

Wow Factor 高光时间点

...

Close-up 特写镜头目标

1. ...

拍摄分镜

0:00-0:03 ...

````

High-impact visual rules:

• 高光黄金动作 must show the core product benefit in one second.
• Wow Factor should be a decisive visual moment, not just a verbal claim.
• Close-up targets should include material, construction, fit, functional details, or texture that reduce purchase hesitation.

Human gate: stop here and ask the user to confirm the voiceover script, golden action, Wow Factor, Close-up targets, and shot plan. Continue to Seedance prompt compilation only after confirmation or revision instructions.

4. Seedance Prompt Compilation

If the user asks for Seedance, compile the approved script into a prompt.

If seedance-prompt-compiler is available, use its rules for product fidelity, asset mapping, timeline, audio, captions, and negative constraints.

Prompt must include:

• Task judgment: UGC seeding ad / product showcase / feature demo etc.
• Asset mapping, e.g. @图片1 as product three-view image.
• Product consistency clause:

参考@图片1中的产品外观，保持产品颜色、材质、结构、版型和关键细节一致，不改变产品形态，不虚构不存在的功能。

• 0-15s timeline with action, camera, and voiceover.
• Filming style: iPhone model if provided, vertical 9:16, natural light, handheld UGC.
• Audio: BGM + voiceover tone + ambient sounds.
• Text policy: if user says no subtitles, explicitly prohibit subtitles, stickers, screen text, logos, and readable text.
• Visual requirements: high-value action, close-up details, full-body or usage demonstration.
• Negative constraints: do not change product color/shape/material; do not invent unsupported features; do not make it look like a studio runway ad unless requested.

Human gate: stop here and ask the user to confirm the Seedance prompt's product consistency clause, camera actions, audio/voiceover, text/subtitle policy, and negative constraints before treating it as ready for generation.

Quality Gate

Before final output, check:

• Product analysis happens before competitor analysis.
• Category/search keywords are included when user needs help finding competitor videos.
• Competitor parsing and structure breakdown are combined as one step.
• Transcript is time-coded if video extraction was possible.
• Product claims are grounded in product page, image, or user-provided facts.
• The UGC script can be spoken within the requested duration.
• Each key benefit has a visual proof action or close-up.
• Seedance prompt protects product fidelity and respects subtitle/text restrictions.
• Steps 1-4 each include a human confirmation gate; do not silently skip them.

Tool Notes

• Use yt_dlp via python3 -m yt_dlp when the yt-dlp executable is absent.
• Use browser/web-access only when downloading/parsing fails or the page requires visual inspection.
• Use ffmpeg / ffprobe for video probing, audio extraction, frames, and contact sheets.
• Use whisper-cli for local transcription when no platform subtitles exist.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: xigua0626
• Source: xigua0626/tiktok-ugc-seedance
• License: MIT

Install and usage instructions live in the source repository linked above.