Lengyi Shotlist

设计电影级分镜视频提示词。输入一段文本(小说片段、一句剧情或几个主题词),拆分或扩写成一段 8-15s 视频的顺序镜头脚本,自动确定分镜数(4/6/9/12/16)与每镜时长,逐镜给出可直接复制到可灵/海螺/seedance 2.0/veo 的提示词——含镜号、场景、风格、主体、动作过程、镜头运动、剧情内容,并附"参考主体生图提示词"以锁定主体一致性。当用户要做视频分镜、分镜脚本、镜头脚本、shotlist、顺序分镜、文生视频/图生视频、AI视频镜头脚本时使用。

43 stars 8 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add woyin2024/lengyi-shotlist --skill lengyi-shotlist
Manual install
git clone --depth 1 https://github.com/woyin2024/lengyi-shotlist.git /tmp/lengyi-shotlist
mkdir -p ~/.claude/skills
cp -r /tmp/lengyi-shotlist/. ~/.claude/skills/lengyi-shotlist

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# lengyi-shotlist:分镜视频提示词设计

把一段输入文本设计成电影级的**顺序镜头脚本**:一段 8-15s 视频被切成 N 个**依次播放**的镜头(镜号 1→2→3…),每个镜头是**一条可独立复制**到视频模型的提示词,按时间顺序拼成完整叙事。

## 默认前提(硬性)

- **本 skill 只做图生视频(I2V)**,不做文生视频。
- 因此**每次都必须**先产出「参考主体文生图(T2I)提示词」用于生成参考主体图,再写各镜头提示词(见第 4、5 步)。
- 默认用中文写提示词(可灵/海螺/seedance 对中文支持好)。目标工具为 veo/Runway/Sora 或用户要求时,给英文或中英双语。

## 能力边界

- 输入三类,分别处理:①小说/散文片段→提炼浓缩最具画面感的节拍;②一句简短剧情→补全前因后果与转折;③几个主题词→创意扩写出"起承转合"微叙事,再拆镜头。
- 输出**必须明确**:分镜数(4/6/9/12/16 之一)、视频总时长(8-15s)、每镜时长区间。
- 分镜数由 skill 按剧情复杂度**自动决定**(见第 2 步);用户若显式指定则优先,但越界(非 4/6/9/12/16 或时长超 8-15s)时取最接近的合法值并一句话说明。

## 工作流程(严格按序执行)

### 第 1 步 · 解析与扩写输入

在心里列出并补全:

1. **主体**:有几个角色/物体/道具主体?各自外观线索(缺失则合理设定)。
2. **场景**:地点、时间、天气、空间。
3. **基调/类型**:从文本气质推断(科幻/悬疑/温情/热血/广告/武侠…)。
4. **叙事节拍(beat)**:拆成有先后的、可视化的画面瞬间。

> 节拍与镜头不必一一对应:​**最终镜头数以第 2 步为准**。镜头数 > 节拍数时,把关键节拍用不同景别/机位拆成相邻数镜(如"推近→特写"占两镜);镜头数 < 节拍数时合并次要节拍,只留最有画面感的。

### 第 2 步 · 自动确定分镜数与时长

按"节拍数 + 复杂度"对照下表选择,​**先定总时长,再分配到各镜**:

| 分镜数 | 适用情形 | 节拍数 | 建议总时长 | 每镜均值 |
| - | - | - | - | - |
| 4 | 单一场景/单一动作;主题词极简 | 2-4 | 8-10s | \~2.0-2.5s |
| 6 | 简单情节、一个小转折 | 4-6 | 9-12s | \~1.5-2.0s |
| 9 | 中等复杂,完整起承转合 | 6-9 | 10-13s | \~1.1-1.4s |
| 12 | 多动作/多机位、情绪递进 | 9-12 | 12-15s | \~1.0-1.25s |
| 16 | 复杂多场景、快剪节奏 | 12-16 | 13-15s | \~0.8-0.95s |

**时长算术(唯一硬约束)**:各镜时长之和 = 选定总时长,且总时长 ∈ \[8,15]s。"每镜均值"仅供参考,各镜不必等长。
**节奏规则**:开场镜略长(建立场景)→ 高潮/转折用更短的镜快切 → 结尾镜留收束。每镜标明时间区间(如 `0.0–2.0s`),自检末镜结束时间 = 总时长。

### 第 3 步 · 建立主体档案(一致性核心)

1. 为**每个主体**写**锚点描述**——固定不变的视觉特征:年龄、发型发色、瞳色、脸型、服装/材质、标志性道具。措辞一旦确定,全程不改。
Read the full skill on GitHub

Details

Skill name
lengyi-shotlist
License
MIT
Repo updated
16 June 2026

FAQ

What does the Lengyi Shotlist skill do?

设计电影级分镜视频提示词。输入一段文本(小说片段、一句剧情或几个主题词),拆分或扩写成一段 8-15s 视频的顺序镜头脚本,自动确定分镜数(4/6/9/12/16)与每镜时长,逐镜给出可直接复制到可灵/海螺/seedance 2.0/veo 的提示词——含镜号、场景、风格、主体、动作过程、镜头运动、剧情内容,并附"参考主体生图提示词"以锁定主体一致性。当用户要做视频分镜、分镜脚本、镜头脚本、shotlist、顺序分镜、文生视频/图生视频、AI视频镜头脚本时使用。

How do I install the Lengyi Shotlist skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add woyin2024/lengyi-shotlist --skill lengyi-shotlist. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Lengyi Shotlist popular?

It lives in woyin2024/lengyi-shotlist, which has 43 GitHub stars and 8 forks.

Related skills

Oobo

43

ooboai/oobo

Context beyond the diff - captures AI sessions, tokens, code attribution, semantic code search, and engineering memory via MCP. Use when the user asks about commit history with AI context, session history, code attribution, token usage, code search, or engineering memory.

AI, LLMs & Agents

Gtm Pipeline:signal Search

77

keinsaasforever/gtm-pipeline-skills

Find buying intent signals for target companies and score them for purchase intent. Runs a Python script (signal_search.py) that orchestrates Parallel web search (always on), Firecrawl crawl (opt-in), Parallel structured enrichment (opt-in), and a Signal Assessment LLM that scores 1-100. Universal templates live in the script; client-specific prompts come from the working directory's context/ files. Runs standalone or in either pipeline workflow — does NOT require ICP scoring as input. Also triggers on "signal search", "find signals for", "buying intent".

AI, LLMs & Agents

Agent Media

43

yuvalsuede/agent-media-skill

AI UGC video production from the terminal using the `agent-media` CLI.

AI, LLMs & Agents

AI Ds Composer

73

cris-achiardi/claude-skills

Strategic guide for consuming design system metadata to compose components intelligently. Use when generating or modifying UI components and layouts to ensure component reuse, proper variant selection, and adherence to design system patterns. Helps Claude understand project structure, read hierarchical metadata indexes, apply anti-pattern rules, and flag when existing components don't fit a requirement.

AI, LLMs & Agents

Capafy Publisher

49

Capafy/Capafy-skills

Creator-side Capafy workflow for publishing, updating, packaging, uploading, re-shipping, or managing an Agent/Skill. Also handles creator earnings, payouts, statistics, refunds, certification/KYC, review links, Agent continuation, and publisher-account switching. Generic buyer login, balance, orders, purchases, subscriptions, and instances belong to capafy-user.

AI, LLMs & Agents10

Adopting Jev

49

shiftynick/jev-axi

Finds where TypeSafe's Jev model fits in a codebase and designs the integration. Use when the user asks where or how to use Jev or TypeSafe in their project, or wants classification cheaper or faster than the LLM calls or keyword rules they run today.

AI, LLMs & Agents12

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.