Skill · #364 trending skills repo · ↗ +1 stars this week
Comfy-H3-Director
MIKA6941/Comfy-H3-Director
专为 MiniMax H3 本地(ComfyUI)打造的 AI 漫剧与影视总导演。支持资产决策、网文改分镜、文武双轨时长与动作连续性,一次性输出全集提示词包。/ Master AI comic drama & multi-shot continuity director skill for MiniMax H3 local workflows.
- Stars
- 18
- This week
- +1
- Forks
- 0
- Open issues
- 0
- Skills inside
- 10
Get started
Install the skills
npx skills add MIKA6941/Comfy-H3-DirectorInstalls into Claude Code, Codex, Cursor, OpenCode and the other agents it finds on your machine.
git clone --depth 1 https://github.com/MIKA6941/Comfy-H3-Director.git /tmp/comfy-h3-director
cp -r /tmp/comfy-h3-director ~/.claude/skills/comfy-h3-directorClaude Code reads ~/.claude/skills/ (or .claude/skills/ in a project). Other agents read the same SKILL.md format from their own skills folder.
Inside this repo
10 skills
.agents/skills/3d-animation-short-generatorai-video-continuity-directorMaster comprehensive AI video cinematography and multi-shot continuity for MiniMax H3. Integrates Phase 0 reusable asset preparation & audio strategy selector (TTS pre-generation vs first-clip voice extraction), asset pruning rules for disposable/one-off assets, dual-track duration engine (8-15s lon.agents/skills/ai-video-continuity-directorbrand-promo-video-generatorFor marketers and creators producing promotional content for brands, products, websites, apps, shops, or personal projects. Users provide logos, product images, interface screenshots, official links, or other verifiable assets and confirm duration, aspect ratio, audience, and campaign focus. The Ski.agents/skills/brand-promo-video-generatorco-op-game-intro-generatorFor users creating a two-player co-op game menu or opening animation. Users provide two player names, a game title, a target visual style, and optional character reference images. The Skill locks identity cues, generates an approval image from a fixed menu framework with coordinated color, buttons, .agents/skills/co-op-game-intro-generatorh3-prompt-writingWrite MiniMax H3 video generation prompts for T2VA, I2VA, FL2VA, L2VA, and Ref2VA. Use when rewriting multimodal requests into H3 prompt structures, composing integrated_multimodal_description, overall_soundscape, and non_diegetic_music, aligning keyframes, or defining reference labels for images, v.agents/skills/h3-prompt-writinghanddrawn-live-video-generatorFor creators making surreal short videos that blend rough glowing hand-drawn animation with live-action spaces. Users provide a scene idea, contact object or hand, desired mood, and optional language or style constraints. The Skill clarifies the physical contact, designs continuous morphing, escape .agents/skills/handdrawn-live-video-generatorminimalist-product-ad-generatorTurn product images and ad requirements into minimalist product ad shorts for e-commerce promotion and product launches. The Skill confirms format and product variants, extracts selling points, writes concise English ad copy, builds product anchors, plans beat-synced typography/storyboards, and gene.agents/skills/minimalist-product-ad-generatormusic-video-subtitle-generatorFor musicians, video creators, and social-media editors producing AI music videos or emotional short films with lyric typography. Users provide music, lyrics, references, characters, typography direction, mood, or target platform. The Skill analyzes beat and vocal timing, separates character, scene,.agents/skills/music-video-subtitle-generatorpaper-collage-explainer-generatorFor creators, educators, and social-video editors who need a tactile paper-collage language for narration, knowledge points, opinions, or abstract topics. Users provide source copy, story beats, or a core concept and may specify aspect ratio, duration, palette, and audio needs. The Skill extracts me.agents/skills/paper-collage-explainer-generatorpapercraft-stop-motion-explainerFor creators explaining science, education, or general knowledge through tactile handmade papercraft visuals. Users provide a topic, core knowledge points, or source material and may specify audience, duration, aspect ratio, and deliverable type. The Skill extracts the learning goal and visual metap.agents/skills/papercraft-stop-motion-explainerDocumentation
README
🤖 【面向 AI 智能体的强制执行指令 / System Directives for AI Agents】
🚨 CRITICAL WARNING FOR AI AGENTS (Gemini / ChatGPT / Claude / Antigravity / Dify etc.):
DO NOT treat this repository as a ComfyUI Python Custom Node plugin! There are NO Python nodes to install, NO custom nodes togit cloneintoComfyUI/custom_nodes/, and NO.batinstallation scripts needed. This repository is a Prompt Engineering & Workflow Skill Suite designed for AI Agents.When you (as an AI Agent) read this repository, you MUST immediately assume the role of "MiniMax H3 Master Director (AI 漫剧与影视总导演)" and execute the following 4-step workflow for the user:
- 📜 Step 1: 剧本与分镜拆解 (Screenplay Breakdown):将用户的故事/网文改编为影视级分镜(0
3s 黄金钩子、文戏长镜头 815s / 武戏快切 2~4s、180° 轴线与动量承接)。- 🎨 Step 2: Phase 0 美术与音频资产清单 (Asset Decision):输出 Tier 1 核心三视图生图词/音色方案,并注明 Tier 2 龙套免做理由。
- 🎯 Step 3: 全量 Ref2VA 提示词直出 (One-Shot Prompts with Media Slot Binding):默认采用 Ref2VA 全参考模式,每个镜头上方置顶标注【ComfyUI 媒体槽位挂载清单】(Picture 1 = 尾帧,Picture 2 = 人设,Audio 1 = 纯净干音)。
- 🎞️ Step 4: 尾帧/纯净音色提取接力指令 (FFmpeg Chaining Commands):输出精确的 FFmpeg 命令,将尾帧保存至
./tail_frames/,人声存至./voices/,指导用户接力填入下一镜。
MiniMax H3 全能影视导演与 AI 漫剧连续性技能库 (Agent Skill Suite)
🚀 专为 AI 智能体打造的“MiniMax H3 本地工作流(ComfyUI 等)全集漫剧量产工业流水线”技能库
An AI Agent Skill Suite empowering MiniMax H3 Local Workflows with One-Shot Episode Scripting, Official-Compliant Prompts & Automated Agent Frame Chaining.
💡 为什么需要这个 Skill?核心定位 (Core Identity & Value)
⚠️ 特别说明:本项目不是 ComfyUI 的 Python 节点插件,而是一套让 AI 智能体(Agent)化身影视总导演的技能库(Skills)。用户只需将小说/剧本发送给 Agent,Agent 即可按照工业化标准完成美术资产决策、分镜编写、全集提示词输出及尾帧接力指导。
本技能库基于 MiniMax-AI/MiniMax-H3 官方 GitHub 技能体系 深度重构,通过六大核心能力彻底重塑生产力:
- 🎨 美术资产智能决策 (Asset Decision Engine):
- 自动判定哪些是 Tier 1 核心资产(主角三视图、本命法宝、主场景、固定音色),并输出生图提示词;
- 自动识别哪些是 Tier 2 消耗性龙套资产,直接免做参考图并注明理由,节省 80% 无谓的准备成本。
- 📝 专业分镜脚本自动编写 (Intelligent Screenplay Breakdown):
- 将小说原文/剧情大纲转化为符合影视规律的分镜脚本(0~3s 黄金钩子、正反打对峙、结尾断章留白)。
- 🎯 默认优先 Ref2VA 全参考模式 & 0 报错语法 (Ref2VA Priority & Syntax Compliance):
- AI 漫剧默认采用 Ref2VA 全参考模式,在同一时间轴内统一调度首尾帧、多角色三视图、法宝与纯净干音;
- 严格遵循官方六大标准区块,确保生成的提示词 0 语法报错、模型理解力最大化。
- 🔌 逐镜置顶【媒体槽位挂载清单】(Media Slot Binding Header):
- 每个镜头提示词上方强制输出清晰的挂载清单,明确标出
<Picture 1>是哪个尾帧、<Picture 2>是哪个人设、<Audio 1>是哪段干音,杜绝连线困惑!
- 每个镜头提示词上方强制输出清晰的挂载清单,明确标出
- ⚡ 一次性全集分镜提示词直出 (One-Shot Episode Batch Generation):
- 单次交互即可一键生成一整集(1~N 镜)的完整提示词工程包,彻底告别单镜头反复拉扯。
- 🎞️ 资产分类归档与尾帧/纯净人声接力支持 (Dedicated Folders & Clean Vocals):
- 跑完每个镜头后,提供精确的无损 FFmpeg 提取命令,将尾帧统一归档至专用目录
./tail_frames/1_last_frame.png,彻底防止文件散乱; - 若需提取视频中角色的声音作为后续镜头的音色母本,明确指导提取纯净人声干声并归档至
./voices/目录,杜绝背景杂音污染。
- 跑完每个镜头后,提供精确的无损 FFmpeg 提取命令,将尾帧统一归档至专用目录
⚡ ComfyUI 本地工作流实操 SOP (Step-by-Step Tutorial)
graph LR
A[1. 输入网文/剧本] --> B[2. 一次性获取<br/>资产清单 + 全集提示词包]
B --> C[3. 按置顶媒体挂载清单<br/>在 ComfyUI 连线渲染镜头 1]
C --> D[4. 提取 1.mp4 尾帧<br/>归档至 ./tail_frames/]
D --> E[5. 作为 Picture 1<br/>接力运行镜头 2]
E --> F[6. 流水线循环直达整集完结]
步骤 1:输入故事,一次性获取全集提示词包
告诉 Agent 你的故事想法或直接粘贴网文小说原文。Agent 将一次性输出:
- Phase 0 美术资产清单(核心人物三视图生图词、音色方案、龙套免做说明);
- 全集分镜总览表(景别、文戏长镜头/武戏快切时长、180° 轴线继承说明);
- 全集 Ref2VA 结构化提示词包(含每个镜头置顶的媒体输入挂载清单)。
步骤 2:在 ComfyUI 中按挂载清单连线并运行镜头 1
对照【镜头 1】上方的媒体挂载清单,将提示词与对应人设图/音频填入 ComfyUI 节点,渲染生成 1.mp4。
步骤 3:提取尾帧/纯净音色并归档专用目录
- 运行附带的命令提取无损尾帧与音频至专用文件夹(防止文件混乱):
# 自动创建专用目录并提取 mkdir -p ./tail_frames ./voices ffmpeg -y -sseof -0.05 -i "1.mp4" -update 1 -q:v 1 "./tail_frames/1_last_frame.png" ffmpeg -y -i "1.mp4" -vn -c:a aac "./voices/char1_voice.aac" - 将生成的
./tail_frames/1_last_frame.png载入 ComfyUI 首图节点(对应【镜头 2】中的<Picture 1>); - 若使用了首镜音色,将提取的
./voices/char1_voice.aac作为后续镜头的<Audio 1>。
步骤 4:继续运行下一镜,完成全集制作
依次接力渲染后续镜头,即可连贯高效地完成整集漫剧制作!
🌟 核心技能清单 (Skills Index)
👑 1. Master 超级总导演(推荐首选)
- 👉
ai-video-continuity-director
MiniMax H3 全能影视导演与 AI 漫剧/短剧连续性生产指南。
集大成之作:完全内化 H3 底层 5 大生成模式、8 大垂类制作流、网文小说改 AI 漫剧工业流水线、Phase 0 资产分级筹备(Tier 1/Tier 2 龙套裁剪)、文武双轨时长规划(815s 文戏长镜头 / 24s 武戏快切)、三位一体动作一致性锁死、五维风格矩阵、动态节奏引擎、物理词库以及全量分镜一次性预制与尾帧接力工作流。
📦 2. MiniMax 官方原版垂直专项技能
h3-prompt-writing:H3 基础多模态提示词结构与底层 5 大模式(T2VA/I2VA/FL2VA/L2VA/Ref2VA)语法手册。3d-animation-short-generator:3D 风格化叙事动画短片生成。minimalist-product-ad-generator:极简商品电商广告与卖点排版短片。music-video-subtitle-generator:音乐 MV 与空间动力学歌词字幕。papercraft-stop-motion-explainer:立体书折纸多层透视定格科普。paper-collage-explainer-generator:复古报刊半色调拼贴短片。handdrawn-live-video-generator:发光手绘涂鸦虚实融合与延迟跟随。co-op-game-intro-generator:双人联机游戏菜单与角色卡片头。brand-promo-video-generator:品牌产品公关级宣传片。
💬 写在最后 / 交流与反馈
这套 Skill 只是我平时使用 MiniMax H3 和 ComfyUI 折腾漫剧时,为了少走弯路整理出来的一点个人使用经验和踩坑笔记。
水平有限,很多规则和流程肯定还有不完善或者粗糙的地方。发出来主要是想给同样在玩本地工作流的朋友们提供个参考。
如果大家在实际使用中发现哪里写得不对、或者有更好的运镜/提示词技巧与想法,非常欢迎提 Issue 交流指正,一起讨论共同进步!☕
📜 许可声明与协议遵循 (License & Compliance)
- 本仓库提供的 Prompt 技能规范、分镜工作流及文档 为个人经验整理与开源分享;
- MiniMax H3 底层模型及其相关技术 归 MiniMax 官方所有,并受官方 MiniMax H3 Community License Agreement 约束;
- 请使用者在使用模型或商用时,严格遵守 MiniMax 官方社区许可协议与使用规范。