Chore/site citation ci

.github/workflows/citations.yml

@@ -0,0 +1,27 @@
+name: Verify citations
+
+on:
+  pull_request:
+  workflow_dispatch:
+  schedule:
+    - cron: "17 4 * * 1"      # weekly, catches venue/ID drift
+
+permissions:
+  contents: read
+
+jobs:
+  citations:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - name: Check reference titles, venues and years
+        run: python3 scripts/verify_citations.py --json citation-findings.json
+      - if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: citation-findings
+          path: citation-findings.json
+          if-no-files-found: ignore

.github/workflows/pages.yml

@@ -0,0 +1,41 @@
+name: Deploy site
+
+on:
+  push:
+    branches: [main]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pages: write
+  id-token: write
+
+concurrency:
+  group: pages
+  cancel-in-progress: false
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+          cache: pip
+      - run: pip install -r requirements-docs.txt
+      - name: Build
+        run: python3 scripts/build_site.py --strict
+      - uses: actions/upload-pages-artifact@v3
+        with:
+          path: _site
+
+  deploy:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    steps:
+      - id: deployment
+        uses: actions/deploy-pages@v4

.gitignore

@@ -10,3 +10,8 @@ __pycache__/
 dist/
 build/
 .worktrees/
+_site/
+.mkdocs-build/
+_to_delete/
+.learnings/
+.workbuddy/

README.md

@@ -2,12 +2,40 @@
 
 _Created by Codex and <a href="https://vinthony.github.io/">Xiaodong Cun (Corresponding Author)</a>, from <a href="https://gvclab.github.io/">GVC Lab, Great Bay University</a>_
 
+[![Site](https://img.shields.io/badge/read-gvclab.github.io%2FVideo--Generation--101-1f6feb)](https://gvclab.github.io/Video-Generation-101/)
+[![Chapters](https://img.shields.io/badge/chapters-40%2B-informational)](docs/taxonomy.md)
+[![Evidence snapshot](https://img.shields.io/badge/evidence%20snapshot-2026--08-success)](docs/timeline.md)
+[![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)
+
+> 📖 **在线阅读（带全文搜索）：<https://gvclab.github.io/Video-Generation-101/>**
+
+
 一份面向初学者、研究者、工程师与创作者的视频生成知识体系。它不以堆叠模型和榜单为目标，而是试图建立一条从**生成原理、系统能力、世界理解到真实应用与验证**的完整学习路径。
 
 > 资料更新时间：**2026-08**
 
 ---
 
+
+<details>
+<summary><strong>English summary</strong></summary>
+
+**Video Generation 101** is a structured knowledge base on video generation, written for learners,
+researchers, engineers and creators. Rather than ranking models, it builds one continuous path from
+generative principles, through system capabilities and world understanding, to real applications and
+the evidence needed to believe them.
+
+Five layers organise the material: **generation foundations**, **system capabilities**,
+**controllable creation**, **world understanding**, and an **evidence framework**.
+
+Every chapter carries an explicit evidence snapshot date and separates author-reported numbers from
+independently reproduced ones. The prose is in Chinese; paper titles, venues and links are in English,
+and reference metadata is checked in CI.
+
+</details>
+
+---
+
 ## 🎯 项目定位
 
 视频生成正在从“根据条件合成一段画面”，走向能够持续保持状态、接受复杂控制、理解动作后果并服务真实任务的通用视频系统。本仓库围绕五个层面组织内容：
@@ -41,7 +69,7 @@ _Coding Agent 扩展研究覆盖面，研究者对结论与证据负责。_
 
 仓库以“原理 → 系统 → 能力 → 世界 → 应用”为主轴，并让评测与治理贯穿每一层。
 
-![图 001：Video Generation 101 知识全景](assets/imagegen-diagrams/001/创建一张用于中文学术教程的高质量横向信息图_主题_Video_2026-08-30T17-01-58.png)
+![图 001：Video Generation 101 知识全景](assets/imagegen-diagrams/001/diagram.png)
 ### 内容导航
 
 | 板块 | 核心入口 | 延伸专题 |

assets/js/mathjax.js

@@ -0,0 +1,23 @@
+// Material for MkDocs swaps page content without a reload when
+// navigation.instant is on, so MathJax has to be re-run per navigation via the
+// document$ observable rather than once on DOMContentLoaded.
+window.MathJax = {
+  tex: {
+    inlineMath: [["\\(", "\\)"]],
+    displayMath: [["\\[", "\\]"]],
+    processEscapes: true,
+    processEnvironments: true,
+  },
+  options: {
+    ignoreHtmlClass: ".*|",
+    processHtmlClass: "arithmatex",
+  },
+};
+
+document$.subscribe(() => {
+  if (!window.MathJax || !window.MathJax.typesetPromise) return;
+  MathJax.startup.output.clearCache();
+  MathJax.typesetClear();
+  MathJax.texReset();
+  MathJax.typesetPromise();
+});

docs/applications.md

@@ -26,7 +26,7 @@
 
 **图 1：部署是带硬门槛的证据链。** 四条领域证据并不是四个排行榜，而是说明同一个生成模型进入不同场景时，必须换一套成功标准。创作关心可控、连续与可改；数字人增加同意、身份与音画同步；交互世界要求动作响应、状态记忆和 deadline；Physical AI 最终要看反事实、闭环成功和安全。图中没有性能数字，避免把示意值误读成 benchmark 结果。
 
-![图 002：从使用合同到部署与回滚](../assets/imagegen-diagrams/002/创建一张用于中文学术教程的高质量横向信息图_主题_从使用合同_2026-08-30T17-01-57.png)
+![图 002：从使用合同到部署与回滚](../assets/imagegen-diagrams/002/diagram.png)
 顺序化文字替代：先写用户、决策和伤害，再写允许使用的文字、图像、视频、音频或动作条件；把基础模型与编辑、音频、安全和版本工具组装成工作流；人工选择和修订后，以多个随机种子、分项指标和压力测试验收。任何硬门槛失败都停止上线。通过后仍需监测服务等级、成本、隐私、来源、事故和分布漂移，并保留回滚入口。
 
 ## 2. 五种证据对象不能混用
@@ -83,7 +83,7 @@ NIST AI 600-1 把生成式 AI 风险管理组织为面向生命周期的 govern
 
 专业制作的核心循环是：
 
-![图 003：创意视频的可回滚制作循环](../assets/imagegen-diagrams/003/创建一张用于中文学术教程的高质量横向信息图_主题_创意视频的_2026-08-30T17-01-57.png)
+![图 003：创意视频的可回滚制作循环](../assets/imagegen-diagrams/003/diagram.png)
 最低报告项：
 
 - 每个 shot 生成多少候选、采用率和返工次数；

docs/evaluation.md

@@ -28,7 +28,7 @@
 
 下面是同一逻辑的可编辑、可搜索版本：
 
-![图 004：视频生成评测证据链](../assets/imagegen-diagrams/004/创建一张用于中文学术教程的高质量横向信息图_主题_视频生成评_2026-08-30T17-01-58.png)
+![图 004：视频生成评测证据链](../assets/imagegen-diagrams/004/diagram.png)
 顺序化文字替代：声明任务和失败代价；按开放生成、开放集个性化、编辑、推理或 World Model 分流；先压力测试自动指标，再用冻结人类样本校准；完成人工盲测与部署门槛；最后发布分项结果、置信区间、失败类型和质量—速度—成本 Pareto。
 
 ### 1.1 历史范式为什么不断扩大

mkdocs.yml

@@ -0,0 +1,95 @@
+site_name: Video Generation 101
+site_description: 从生成原理、系统能力、世界理解到真实应用与验证的视频生成知识体系
+site_author: Xiaodong Cun, GVC Lab, Great Bay University
+site_url: https://gvclab.github.io/Video-Generation-101/
+repo_url: https://github.com/GVCLab/Video-Generation-101
+repo_name: GVCLab/Video-Generation-101
+edit_uri: edit/main/
+copyright: MIT License · GVC Lab, Great Bay University
+
+# The repository is browsed on GitHub as much as on the site, so the whole repo
+# is the docs root. That keeps every existing relative link (docs/x.md,
+# ../../assets/y.png) working identically in both places.
+docs_dir: .mkdocs-build/docs
+site_dir: _site
+
+not_in_nav: |
+  sources/*
+
+theme:
+  name: material
+  language: zh
+  icon:
+    repo: fontawesome/brands/github
+  features:
+    - navigation.instant
+    - navigation.tracking
+    - navigation.top
+    - navigation.indexes
+    - navigation.footer
+    - toc.follow
+    - search.suggest
+    - search.highlight
+    - search.share
+    - content.code.copy
+    - content.action.edit
+  palette:
+    - media: "(prefers-color-scheme: light)"
+      scheme: default
+      primary: indigo
+      accent: indigo
+      toggle:
+        icon: material/weather-night
+        name: 切换到深色模式
+    - media: "(prefers-color-scheme: dark)"
+      scheme: slate
+      primary: indigo
+      accent: indigo
+      toggle:
+        icon: material/weather-sunny
+        name: 切换到浅色模式
+
+plugins:
+  - search:
+      lang:
+        - zh
+        - en
+
+markdown_extensions:
+  - abbr
+  - admonition
+  - attr_list
+  - def_list
+  - footnotes
+  - md_in_html
+  - tables
+  - toc:
+      permalink: true
+      toc_depth: 3
+  - pymdownx.details
+  - pymdownx.highlight:
+      anchor_linenums: true
+  - pymdownx.inlinehilite
+  - pymdownx.snippets
+  - pymdownx.arithmatex:
+      generic: true
+  - pymdownx.superfences:
+      custom_fences:
+        # The chapters write display equations as ```math fences (GitHub style).
+        - name: math
+          class: arithmatex
+          format: !!python/object/apply:pymdownx.arithmatex.arithmatex_fenced_format
+            kwds:
+              mode: generic
+              tag: pre
+
+extra_javascript:
+  - assets/js/mathjax.js
+  - https://unpkg.com/mathjax@3/es5/tex-mml-chtml.js
+
+extra:
+  social:
+    - icon: fontawesome/brands/github
+      link: https://github.com/GVCLab
+    - icon: fontawesome/solid/flask
+      link: https://gvclab.github.io/

requirements-docs.txt

@@ -0,0 +1,4 @@
+mkdocs-material==9.7.6
+pymdown-extensions>=10.21
+jieba>=0.42          # Chinese tokenisation for the search index
+Pillow>=10           # scripts/optimize_assets.py

resources/datasets.md

@@ -67,7 +67,7 @@ N_{\mathrm{eff,source}}
 
 下面给出可编辑、可搜索、对读屏器更友好的确定性版本：
 
-![图 085：可审计的现代 Video Data Engine](assets/imagegen-diagrams/085/diagram.png)
+![图 085：可审计的现代 Video Data Engine](../assets/imagegen-diagrams/085/diagram.png)
 顺序化文字替代：
 
 1. 登记来源、权利证据与 opt-out 通道；通过权利门后再做不可变摄取。

resources/open-models.md

@@ -72,7 +72,7 @@ R(m)=(P,C,W,I,T,D,E,L,H),
 
 **图 1：开放模型不是一个下载按钮。** 主链把模型身份、版本、发布物、许可证、硬件、运行证据和最终 manifest 连起来；红色出口表示证据不足时的隔离路径。上图是教学总览，下方 Mermaid 是可编辑且适合读屏器的规范版本。
 
-![图 086：开放视频模型的可复现发布面证据链](assets/imagegen-diagrams/086/diagram.png)
+![图 086：开放视频模型的可复现发布面证据链](../assets/imagegen-diagrams/086/diagram.png)
 顺序化文字替代：确认作者身份；固定代码和权重 revision；分别核对九个发布面；求许可证交集；检查显存、磁盘、CUDA 和精度；用固定输入与随机种子跑最小样例；验证输出媒体契约；人工回读条件遵循和失败模式；最后保存 manifest、hash、命令、硬件和日志。任一关键门失败都应记录并隔离。
 
 ---
@@ -138,7 +138,7 @@ R(m)=(P,C,W,I,T,D,E,L,H),
 
 ## 4. 先选研究问题，再选模型
 
-![图 087：开放视频模型选择树](assets/imagegen-diagrams/087/diagram.png)
+![图 087：开放视频模型选择树](../assets/imagegen-diagrams/087/diagram.png)
 ### 4.1 按显存筛选时的正确读法
 
 | 官方入口级别 | 可先考虑的路线 | 关键限定 |
@@ -175,7 +175,7 @@ R(m)=(P,C,W,I,T,D,E,L,H),
 
 单个“最好看”的视频只能证明一次成功，不能证明模型稳定。最低限度应使用多个 prompt 与 seed，并统一输出合同：容器、编码、宽高、FPS、帧数、音轨和时长。FFmpeg 可检查媒体规格，但不能替代语义和物理评测 [[30]](#ref-30)。
 
-![图 088：开放模型最小复现链](assets/imagegen-diagrams/088/diagram.png)
+![图 088：开放模型最小复现链](../assets/imagegen-diagrams/088/diagram.png)
 ---
 
 ## 6. 训练与微调：先问“开放到哪一层”