feat: 新增 methodology/20 写实 AI 角色 / 网红一致性大师课

榨取 /Users/lank/ai-women-masterclass.md(20 个 YouTube 教学,累计 5M+ 播放),
作为仓库第 20 篇方法论。

## 命名调整

源文件名 "ai-women" 调整为中性专业的 "realistic-character-consistency"
(避免性别偏向 + SEO/国际化友好;源教程主题实际是 AI 网红/数字 Persona
跨工具流水线,女性只是案例选择)。

## methodology/20-realistic-character-consistency.md (NEW, 12 节)

### §1 角色一致性 5 种主流技术
- Character Sheet(Nano Banana Pro 多角度卡)
- Omni Reference(OpenArt Flux Kontext 4 图同时,Composition Reference 0.17)
- OpenArt 内置角色锁定(11 分钟入门)
- 3×3 角度网格(替代 2×2,稳定性+绕过人脸过滤双效)
- Face Swap(Krea AI 跨场景换脸)

### §2 写实人像 5 段 Prompt 公式(Dan Kieft Pack)
Subject → Pose & Framing → Camera & Style → Skin/Face → Aesthetic
关键技巧:相机型号(Fujifilm X100V/X-T30) + `unfiltered` + `pores visible`
+ `Instagram aesthetic` 一句话定调

### §3 真实 prompt 原文档案(6 条)
- OpenArt 完整角色 prompt
- 多角度旋转一致性 prompt
- GPT Image 2 新面部 prompt
- Dan Kieft 核心人像 prompt
- Starbucks 品牌合作图像 prompt
- **Kling 2.1 推近视频 prompt → 已入库为 kl-156**

### §4 工具生态对比(20 视频整合)
图像生成 7 个 / 后处理 3 个 / 视频驱动 5 个 / 声音口型 4 个 /
自动化发布 3 个的完整工具对比表

### §5 12 个写实图像 prompt 技巧(Style Fusion/Lighting/Texture/Punk It...)
### §6 Nano Banana 15 个 prompt 技巧索引
### §7 Arena Zero 4 人 AI 剧集案例(5000 次生成完成一整集)
### §8 完整制作流水线(综合 20 视频,5 个 stage)
### §9 已入库 prompt 对照
### §10 视频来源索引(20 视频,5M+ 播放)
### §11 与其他 methodology 的关系
### §12 关键启示(5 条)

## 新增 prompt(1 条)

- **kl-156 · AI 网红品牌合作视频 — Starbucks 推近镜头**
  来源:Dan Kieft Prompt Pack 原文,Kling 2.1
  category: product-commercial, params: 9:16 / 5s / 1080p

## Stats 同步

- 单模型 prompts: 429 → 430 (+1)
- 总 prompt 资产: 539 → 540 (+1)
- methodology: 19 → **20 篇** (+1)
- Kling: 71 → 72
- 累计 YouTube 教学:47+ → **67+**(累计播放 13M+)
- schema_version 1.12 → 1.13

涉及 12 文件:methodology/20 + 13 个 stats 同步位置。

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Author: cclank

README.en.md

@@ -9,24 +9,24 @@
 # 🎬 lanshu-awesome-ai-video-kit
 
 **An awesome curated kit for AI video prompt engineering.**
-**The most complete AI video prompt library on the web — 15 models (11 commercial + 4 open source) / 429 standalone + 110 cross-model = 539 prompts / 7 Claude Skills / 19 methodology docs**
+**The most complete AI video prompt library on the web — 15 models (11 commercial + 4 open source) / 430 standalone + 110 cross-model = 540 prompts / 7 Claude Skills / 20 methodology docs**
 
 🇨🇳 **[中文 README →](README.md)**
 
 <p>
   <img alt="Awesome" src="https://img.shields.io/badge/-Awesome-fc60a8?style=flat&logo=awesome-lists&logoColor=white">
   <img alt="Models" src="https://img.shields.io/badge/models-15-8b5cf6?style=flat">
-  <img alt="Prompts" src="https://img.shields.io/badge/prompts-429-f97316?style=flat">
+  <img alt="Prompts" src="https://img.shields.io/badge/prompts-430-f97316?style=flat">
   <img alt="Skills" src="https://img.shields.io/badge/skills-7-06b6d4?style=flat">
-  <img alt="Docs" src="https://img.shields.io/badge/methodology-19-34d399?style=flat">
+  <img alt="Docs" src="https://img.shields.io/badge/methodology-20-34d399?style=flat">
   <img alt="License" src="https://img.shields.io/badge/license-MIT-cccccc?style=flat">
   <img alt="PRs Welcome" src="https://img.shields.io/badge/PRs-welcome-9333ea?style=flat">
 </p>
 
 <p>
   <a href="#-quick-start">🚀 Quick Start</a> ·
-  <a href="prompts/README.md">📚 429 Prompts</a> ·
-  <a href="methodology/README.md">📖 19 Methodology Docs</a> ·
+  <a href="prompts/README.md">📚 430 Prompts</a> ·
+  <a href="methodology/README.md">📖 20 Methodology Docs</a> ·
   <a href="skills/README.md">🛠️ 7 Skills</a> ·
   <a href="tools/prompt-browser/index.html">🌐 Web Browser</a> ·
   <a href="RESOURCES.md">🔗 Resources</a>
@@ -53,8 +53,8 @@ Structured into 4 categories of practical resources:
 
 | Resource | Content | Best for |
 |---|---|---|
-| 📚 [prompts/](prompts/) | **429 battle-tested prompts** across 16 models / 29 scenarios, with official samples + recommended params | Copy & tweak |
-| 📖 [methodology/](methodology/) | **19 methodology SOPs**: advanced formula / storyboarding / emotion externalization / 5 model-specific guides + 6-model quick reference + **4 open source quick reference** + cross-model comparison + 12 pitfalls | Learn to "direct shots" |
+| 📚 [prompts/](prompts/) | **430 battle-tested prompts** across 16 models / 29 scenarios, with official samples + recommended params | Copy & tweak |
+| 📖 [methodology/](methodology/) | **20 methodology SOPs**: advanced formula / storyboarding / emotion externalization / 5 model-specific guides + 6-model quick reference + **4 open source quick reference** + cross-model comparison + 12 pitfalls | Learn to "direct shots" |
 | 🛠️ [skills/](skills/) | **7 Claude Code Skills** — Seedance ×3 + HappyHorse + Kling + **model-selector** (15-model shopping advisor) + **prompt-translator** (cross-model converter) | Let Claude pick/write/fix/translate |
 | 🌐 [tools/prompt-browser/](tools/prompt-browser/) | Single-page HTML browser (15-model rainbow filter + URL state sharing + keyboard nav + Drawer details) | If you don't want to read markdown |
 
@@ -130,7 +130,7 @@ Filter by scenario / model / tags in the sidebar → hover cards for preview →
 
 - [Seedance 2.0 · 114 prompts](prompts/seedance/README.md) — commercials / social viral / cinematic / action / sports / nature / ASMR / comedy / craft
 - [HappyHorse 1.0 · 90 prompts](prompts/happyhorse/README.md) — 8s tight beats / product / portrait / action / cinematic / nature
-- [Kling 3.0 · 71 prompts](prompts/kling/README.md) — product / cinematic / action / **viral transform** / music video / experimental / gaming / **image-to-video specialty**
+- [Kling 3.0 · 72 prompts](prompts/kling/README.md) — product / cinematic / action / **viral transform** / music video / experimental / gaming / **image-to-video specialty**
 - [Sora 2 · 20 prompts](prompts/sora/README.md) — incl. OpenAI Cookbook official Shot List & parameterized templates
 - [Veo 3.1 · 20 prompts](prompts/veo/README.md) — incl. Google DeepMind official native audio dialogue templates
 - **Runway Gen-4.5 / Aleph · 12 prompts** — JSON `rw-*` · Aleph editing ★ #1 ELO overall
@@ -189,6 +189,7 @@ Read [methodology/](methodology/) in order:
 | 17 | [HappyHorse Masterclass](methodology/17-happyhorse-masterclass.md) ⭐ | 14 YouTube tutorials: 6-element formula + 20-word single-shot + emotion-to-physics + @tag Omni Reference + Storyboard 45s + 11 pitfalls + 33 prompts | 12 min ⭐ |
 | 18 | [Kling Masterclass](methodology/18-kling-masterclass.md) ⭐ | 25 YouTube channels (6M+ views): 5-element + Constraint Sandwich (O1) + Multi-Shot 6 shots/15s + Motion Brush/Library + 3×3 grid + Character Sheet + OmniEdit + 56 prompts | 18 min ⭐ |
 | 19 | [Seedance Round 3](methodology/19-seedance-masterclass-round3.md) ⭐ | 8 new YouTube tutorials: 3-type reference + CapCut Video Studio + AI Anime 4-step + 3×3 grid + video extend 30s + 19 prompts | 10 min ⭐ |
+| 20 | [Realistic Character Consistency](methodology/20-realistic-character-consistency.md) ⭐NEW | 20 YouTube tutorials (5M+ views): realistic portrait 5-part formula + Character Sheet + Omni Reference + 3×3 grid + Face Swap + Dan Kieft Prompt Pack + MidJourney→Nano Banana→Kling/VEO full pipeline | 15 min ⭐ |
 
 ---
 
@@ -217,7 +218,7 @@ See [skills/README.md](skills/README.md) for the decision tree.
 | [Cross-Model Matrix](tools/cross-model/index.html) | 10 scenarios × 11 commercial models = 110 cross-model prompts + video slots | `tools/cross-model/` |
 
 Features:
-- **Hero stats**: 429 / 16 / 7 / 19 live counts
+- **Hero stats**: 430 / 16 / 7 / 20 live counts
 - **15-model rainbow filter** — each model has its own glowing color dot
 - **3D filtering**: model + category + tags (multi-select)
 - **Keyword search** — title / prompt body / tags / notes
@@ -241,7 +242,7 @@ lanshu-awesome-ai-video-kit/
 ├── awesome.md                      # awesome-list submission entry
 ├── LICENSE                         # MIT
 │
-├── prompts/                        # 429 prompts (16 models)
+├── prompts/                        # 430 prompts (16 models)
 │   ├── data/all-prompts.json       # Single source of truth (web tools consume this)
 │   ├── data/cross-model-matrix.json # 110 cross-model prompts (10 × 11)
 │   ├── seedance/README.md          # 64 Seedance prompts index
@@ -251,7 +252,7 @@ lanshu-awesome-ai-video-kit/
 │   ├── veo/README.md               # 20 Veo prompts index
 │   └── (other 10 models)           # rw-* pk-* hl-* hy-* wn-* jm-* lt-* mo-* cg-* hg-*
 │
-├── methodology/                    # 19 methodology SOPs
+├── methodology/                    # 20 methodology SOPs
 │   ├── 01-基础公式.md ~ 08-避坑12问.md   # General + Seedance
 │   ├── 09-kling-公式.md ~ 12-veo-公式.md # Kling/Sora/Veo + cross-model
 │   ├── 13-六大模型公式速查.md      # Runway/Pika/Hailuo/Hunyuan/Wan/Jimeng + 11 selection

README.md

@@ -2,15 +2,15 @@
 
 <!-- Hero Banner -->
 <picture>
-  <source media="(prefers-color-scheme: dark)" srcset="https://capsule-render.vercel.app/api?type=waving&color=gradient&customColorList=12,2,30&height=200&section=header&text=lanshu-awesome-ai-video-kit&fontSize=42&fontColor=ffffff&fontAlignY=38&desc=AI%20Video%20Prompt%20Engineering%20Kit%20·%2016%20Models%20·%20539%20Prompts&descSize=15&descAlignY=62&descColor=cccccc">
-  <img src="https://capsule-render.vercel.app/api?type=waving&color=gradient&customColorList=12,2,30&height=200&section=header&text=lanshu-awesome-ai-video-kit&fontSize=42&fontColor=ffffff&fontAlignY=38&desc=AI%20Video%20Prompt%20Engineering%20Kit%20·%2016%20Models%20·%20539%20Prompts&descSize=15&descAlignY=62&descColor=ffffff" alt="banner" width="100%">
+  <source media="(prefers-color-scheme: dark)" srcset="https://capsule-render.vercel.app/api?type=waving&color=gradient&customColorList=12,2,30&height=200&section=header&text=lanshu-awesome-ai-video-kit&fontSize=42&fontColor=ffffff&fontAlignY=38&desc=AI%20Video%20Prompt%20Engineering%20Kit%20·%2016%20Models%20·%20540%20Prompts&descSize=15&descAlignY=62&descColor=cccccc">
+  <img src="https://capsule-render.vercel.app/api?type=waving&color=gradient&customColorList=12,2,30&height=200&section=header&text=lanshu-awesome-ai-video-kit&fontSize=42&fontColor=ffffff&fontAlignY=38&desc=AI%20Video%20Prompt%20Engineering%20Kit%20·%2016%20Models%20·%20540%20Prompts&descSize=15&descAlignY=62&descColor=ffffff" alt="banner" width="100%">
 </picture>
 
 # 🎬 lanshu-awesome-ai-video-kit
 
 **做企业 AI 视频项目逼出来的开源工具包**
 
-539 实测 prompt · 16 模型 · 7 Claude Skill · 19 篇方法论 · GitHub Action 每周自动监控官方端点
+540 实测 prompt · 16 模型 · 7 Claude Skill · 20 篇方法论 · GitHub Action 每周自动监控官方端点
 
 [🇨🇳 中文](README.md) · [🇬🇧 English](README.en.md) · [🌐 **Live Demo**](https://lanshu-awesome-ai-video-kit.lank.workers.dev) · [⚡ awesome 投稿](awesome.md)
 
@@ -27,9 +27,9 @@
 
 <!-- Content stats badges -->
 [![Models](https://img.shields.io/badge/Models-15-8b5cf6?style=flat-square&logo=googlegemini&logoColor=white)](#-15-模型一览)
-[![Prompts](https://img.shields.io/badge/Prompts-539-fb923c?style=flat-square)](prompts/)
+[![Prompts](https://img.shields.io/badge/Prompts-540-fb923c?style=flat-square)](prompts/)
 [![Skills](https://img.shields.io/badge/Claude_Skills-7-d97757?style=flat-square&logo=anthropic&logoColor=white)](skills/)
-[![Methodology](https://img.shields.io/badge/SOP-19-34d399?style=flat-square)](methodology/)
+[![Methodology](https://img.shields.io/badge/SOP-20-34d399?style=flat-square)](methodology/)
 [![Categories](https://img.shields.io/badge/Categories-29-f472b6?style=flat-square)](prompts/data/all-prompts.json)
 [![Endpoints Monitored](https://img.shields.io/badge/Auto_Monitor-32_endpoints-06b6d4?style=flat-square&logo=githubactions&logoColor=white)](.github/workflows/model-version-monitor.yml)
 
@@ -68,10 +68,10 @@
 ## 🎯 数据规模(v0.9.0 · 2026-05)
 
 ```
-539 条 prompt   (429 单模型最佳实践 + 110 跨模型对照矩阵)
+540 条 prompt   (430 单模型最佳实践 + 110 跨模型对照矩阵)
  15 个 模型      (11 商业旗舰 + 4 开源 / 友好开源)
   7 个 Claude Skill  (含 model-selector + prompt-translator 跨模型核心)
- 19 篇 方法论 SOP    (+ HappyHorse / Kling / Seedance Round 3 三大 masterclass)
+ 20 篇 方法论 SOP    (+ AI 网红/角色一致性大师课 20 视频)
  29 个 场景分类      (产品 / 对话 / 物理 / I2V / 武侠 / 萌宠 / ...)
  32 个 监控端点      (每周一 09:00 北京时间自动巡检)
    3 个 Web 工具    (Liquid Glass · 零依赖单文件 HTML)
@@ -101,7 +101,7 @@ python3 serve.py 8000
 - [🚀 5 秒上手](#-5-秒上手)
 - [📦 15 模型一览](#-15-模型一览)
 - [🛠️ 7 个 Claude Code Skill](#%EF%B8%8F-7-个-claude-code-skill)
-- [📖 19 篇方法论 SOP](#-19-篇方法论-sop)
+- [📖 20 篇方法论 SOP](#-20-篇方法论-sop)
 - [🌐 3 个 Web 工具](#-3-个-web-工具)
 - [🤖 自动监控机制](#-自动监控机制)
 - [📁 目录结构](#-目录结构)
@@ -169,9 +169,9 @@ done
 
 ---
 
-## 📖 19 篇方法论 SOP
+## 📖 20 篇方法论 SOP
 
-按主题分七组:
+按主题分八组:
 
 | # | 类别 | 文档 |
 |---|---|---|
@@ -184,8 +184,9 @@ done
 | 17 | **HappyHorse Masterclass** ⭐ | 14 YouTube 教学:6 要素黄金公式 + 20 词单镜规则 + 情绪转物理动作 + @tag Omni Reference(9 张上限) + 多镜头 SHOT(0:00-0:05) + AUDIO 块 7 语言 lip-sync + Storyboard 45s + 11 条避坑 + 8 模板 + 33 条 prompt |
 | 18 | **Kling Masterclass** ⭐ | 25 YouTube 频道(6M+ 累计):5 要素公式 + Constraint Sandwich(O1) + Style Bible Line + Elements/Bind 跨镜头一致性 + Multi-Shot **6 shots/15s** + Motion Brush/Library/Transfer + 3×3 网格法 + Character Sheet + OmniEdit + 56 条 prompt |
 | 19 | **Seedance Masterclass Round 3** ⭐ | 8 个新 YouTube 视频:三类 reference(图/音/视频)+ CapCut Video Studio 三模式 + AI Anime 四步法 + 3×3 网格(替代 2×2)+ 视频扩展 30s + Advanced Guide 五段式 + BytePlus/ChatCut/Mitte 访问渠道 + 19 条 prompt |
+| 20 | **AI 网红/角色一致性大师课** ⭐NEW | 20 个 YouTube 教学(累计 5M+ 播放):写实人像 5 段公式 + Character Sheet + Omni Reference + 3×3 网格 + Face Swap + Dan Kieft Prompt Pack 原文 + MidJourney→Nano Banana→Kling/VEO 完整流水线 |
 
-> **如果只挑两篇必读**:[`02-进阶 8 要素`](methodology/02-进阶公式.md)(导演级写作框架,所有 prompt 都是它的变体)+ [`13-六大模型公式速查`](methodology/13-六大模型公式速查.md)(12 分钟拿到 6 个模型公式)。其他 17 篇按需查。
+> **如果只挑两篇必读**:[`02-进阶 8 要素`](methodology/02-进阶公式.md)(导演级写作框架,所有 prompt 都是它的变体)+ [`13-六大模型公式速查`](methodology/13-六大模型公式速查.md)(12 分钟拿到 6 个模型公式)。其他 18 篇按需查。
 
 ---
 
@@ -195,7 +196,7 @@ done
 
 | 工具 | 路径 | 卖点 |
 |---|---|---|
-| **Prompt Browser** | [`tools/prompt-browser/`](tools/prompt-browser/) | 539 prompt 浏览器 · 15 模型彩虹筛选 · URL 状态分享 · 键盘导航(`/`/`j`/`k`/`Enter`/`c`)|
+| **Prompt Browser** | [`tools/prompt-browser/`](tools/prompt-browser/) | 540 prompt 浏览器 · 15 模型彩虹筛选 · URL 状态分享 · 键盘导航(`/`/`j`/`k`/`Enter`/`c`)|
 | **Cross-Model Matrix** ★ | [`tools/cross-model/`](tools/cross-model/) | 10 场景 × 11 模型 = 110 横向对照 · `prompt-translator` 的查表基准 |
 | **Markdown Viewer** | [`viewer.html`](viewer.html) | 所有 .md 文件渲染成漂亮 web 文档 · 自动 TOC + 代码高亮 |
 
@@ -231,11 +232,11 @@ lanshu-awesome-ai-video-kit/
 ├── serve.py                        # 本地 dev server(UTF-8 + .md redirect)
 │
 ├── prompts/
-│   ├── data/all-prompts.json       # 单一数据源(539 条)
+│   ├── data/all-prompts.json       # 单一数据源(540 条)
 │   ├── data/cross-model-matrix.json # 110 条跨模型对照
 │   └── {seedance,happyhorse,kling,sora,veo}/README.md
 │
-├── methodology/                    # 19 篇方法论 SOP
+├── methodology/                    # 20 篇方法论 SOP
 │   ├── 01-基础公式.md ~ 08-避坑12问.md
 │   ├── 09-kling-公式.md ~ 12-veo-公式.md
 │   ├── 13-六大模型公式速查.md
@@ -256,7 +257,7 @@ lanshu-awesome-ai-video-kit/
 │   └── ISSUE_TEMPLATE/             # 4 个贡献表单
 │
 ├── tools/
-│   ├── prompt-browser/             # 539 prompt 浏览器
+│   ├── prompt-browser/             # 540 prompt 浏览器
 │   └── cross-model/                # 跨模型对照矩阵
 │
 ├── viewer.html                     # Markdown viewer