jimeng-storyboard
Use when 编写即梦分镜。按讲戏本剧情点映射母镜头、填7字段。
- License 许可
- MIT
- 最近更新
- 2026-11-11
- Stars
- ★ 1.6k
- Forks
- 468
01 前言 Front Matter
- name
- jimeng-storyboard
- author
- 即梦分镜师工作台
- license
- MIT
- version
- 1.1.0
- metadata
- { "codex": { "tags": [ "即梦分镜师", "分镜", "母镜头", "台词校验" ], "related_skills": [ "jimeng-director", "jimeng-review", "jimeng-panorama-blocking" ] } }
- platforms
- [ "windows", "linux", "macos" ]
- description
- Use when 编写即梦分镜。按讲戏本剧情点映射母镜头、填7字段。
02 资源文件 18
- 参考文档config/duration-rules.json2.1 KB
- 参考文档examples/jimeng-prompt-examples.md12.2 KB
- 脚本scripts/apply_corrections.py13.6 KB
- 脚本scripts/auto_correct.py4.7 KB
- 脚本scripts/auto_header.py4.8 KB
- 脚本scripts/calc_duration.py15.0 KB
- 脚本scripts/check_assets_change.py9.4 KB
- 脚本scripts/check_budget.py6.4 KB
- 脚本scripts/check_dialogue.py11.6 KB
- 脚本scripts/check_mother_id.py3.5 KB
- 脚本scripts/check_picture_field.py8.8 KB
- 脚本scripts/gen_clean.py2.3 KB
- 脚本scripts/gen_dialogue_list.py10.3 KB
- 脚本scripts/gen_episode_brief.py5.5 KB
- 脚本scripts/run_all_checks.py7.6 KB
- 脚本scripts/stage5_pipeline.py3.6 KB
- 参考文档templates/dialogue-shots.md6.3 KB
- 参考文档templates/jimeng-prompts-template.md7.8 KB
03 SKILL.md 正文
jimeng-storyboard — 即梦分镜编写(完整版)
工作区根:Codex 项目根(AGENTS.md 所在目录;下文所有
scripts/…、tools/…、config/…相对该根,可用环境变量JIMENG_ROOT显式指定)。
角色:即梦分镜师。
先读
<工作区根>/.codex/memory/<剧名>-storyboard-artist.md身份文件(缺失 → 按派遣契约 abort)。
输入
- 剧本原文
scripts/<剧名>/script/epXX.txt(台词字面唯一权威) - 台词本
outputs/epXX/epXX-dialogue-list.md(台词逐字 + 参考时长) - 讲戏本
outputs/epXX/01-director-analysis.md(剧情点/台词/预算) - 全景站位
outputs/epXX/00-panorama-blocking.md+ 资产清单epXX-asset-list.md(@引用名唯一权威) - 资产提示词
assets/{character,prop,scene}-prompts.md、全局风格config/global-style.md
结构
- 母镜头 = 一个剧情点,总时长 4-15s(最佳 12-15s);剧情点边界 1:1 映射(不重拆/不合并),超15s 勿自行拆,标注待阶段七
- 子镜头 = 母镜头内单个镜头,每母镜头内独立从 1 开始编号;子镜头时长之和 = 母镜头总时长
- 每个母镜头至少 2 个子镜头
时长铁律(关键)
- 台词时长唯一锁定:台词文本逐字复制剧本原文;每句分配时长 ≥ 台词本参考时长,有效语速 = 字数/分配时长 ≤11字/秒
- 子镜头时长下限:子镜头 ≥1.0s;完整动作链 ≥3s;复杂动作链(追逐打斗) ≥5s
- 反应镜头一般 ≥1s(0.5s 只够快速闪切,不够读清情绪);重大情绪反应可 1.5-2s
- 台词类子镜头时长 = 计算台词时长 + 动作余量:纯台词+0.5s / 台词+微小动作+1s / 台词+明显动作+2s
- 纯动作子镜头 1-3s;环境交代 2-3s;不机械复刻讲戏本 P 预算(导演非台词时长不精准),总时长由分镜师自控在 4-15s
- 同母镜头内景别、机位角度、运镜必须有变化
流程
- 读全部上游;建立素材对应表(@引用只给资产库中有的素材,群演用纯文字)
- 母镜头 1:1 映射剧情点;判定场景类型(对话/动作/情绪/揭示/过渡)
- 内联镜头设计:按场景类型套用镜头组合模板(对话"正反打+反应"、动作"全景交代→中景动作→特写反应"、情绪"环境→状态→细节→空镜"等)
- 母镜头结构顺序:标题
母镜头N(总时长:Xs秒)—— [Pxx 标题]→ 场景类型 → 分镜过渡(定调+连接)→ 全局风格设定 → 资产声明行 → 子镜头列表 - 填写 7 字段(按权重):【景别】【运镜】【机位】【画面】【微表情】【台词】【音效】
资产声明行格式
人物:@角色A 是角色A名, @角色B 是角色B名 | 场景:@场景名 是场景名 | 道具:@道具A 是道具A名
- 三个位置用
|分隔;人物/道具多人逗号分隔、无人或无道具写"无";场景必填且一母镜头一场景 - 只列本母镜头实际 @点名 出现的人物(未入画不得声明);群像点名化一致
7 字段要点
- 【景别】景别(8档),构图;同一母镜头内必须变化
- 【运镜】中文+英文括注(缓慢推近 push in);AI可执行性分级:第一级(固定/缓慢推近/缓慢拉远/跟拍)≥60%,第二级(摇/俯仰/环绕/手持/升降)≤30%且须"缓慢"修饰,第三级(甩镜/定格凝视/推拉变焦等)≤10%慎用
- 【机位】拍摄角度,焦段,位于@角色 X侧;同母镜头内角度要有变化;动作用中景、反应用特写
- 【画面】@场景名-内外 @角色 具体动作,环境光影;用具体动词禁止抽象;每个@人物引用必须命中姿态词(站/坐/蹲/躺/弯腰等;动态词 走/跑/跳 等亦有效,蹦起/跃起/纵身/弹起/蹿/站起/立起/挺直/俯下/蜷/撑起 均在词表)
- 【微表情】@角色名 具体面部肌肉描写(情绪标签:XX%)
- 【台词】@角色名 说:「原文」(情绪,语速XX,语调XX,情绪强度XX%);情绪强度与微表情一致;无台词写"无"
- 【音效】环境音:…, 拟音:…, 人物发声:…(禁BGM)
动作链铁律(最高优先级)
一个完整动作链 = 一个子镜头("接近→碰撞→反弹"一个镜头完成,不拆成 3 个子镜头)。
禁忌
禁否定句、禁角色直视镜头、禁纯正面证件照、禁母镜头超15s或低于4s、禁全同景别/运镜/机位、禁为不存在的@引用。
产出自检(四道全绿才算交稿)
tools\py.cmd tools/run_all_checks.py <剧名> epXX
子镜头数>剧本句数是拆分非缺失,勿合并回。改完必验,返回一行摘要:母镜头数=N | 子镜头数=N | 总时长=Ns | 0.5s镜头数=N | 反应最短=Ns | 校验=全绿/未全绿。
04 同仓库更多 Skills
deja-search
deja-vu memory — search the user's past AI coding sessions with the deja CLI. Holds what they already settled and never wrote into the repo (naming rules, banned or already-tried tools, required checks before merge, preferred values, who owns what). Use before your first edit in a task, before you call a change done or ready to merge, before you recommend a tool, library, design or reviewer, before re-debugging an error that may already be solved, when they say 'didn't we fix this before', 'what did we decide about X' or 'I already have that', and before telling them something on this machine does not exist.
deja-reflect
End-of-session reflection and memory curation with deja-vu. Review what was decided, what commands broke and what repaired them, and record durable project rules or decisions with deja remember so the next session starts with them.
revid
Make, edit, export and post AI videos with Revid (revid.ai) through the Revid MCP server. Use whenever the user wants a short-form or long-form video made (from an idea, a script, a URL, an article, a PDF, a tweet, a product page, their own clips, or a song), wants to find what is going viral on TikTok, YouTube, Instagram or X before making something, wants to edit or re-export an existing Revid project, generate a standalone AI image or clip, make a series episode, or publish and schedule a finished video to connected social accounts. Covers setup (hosted MCP with OAuth, or an API key), the 36 tools, credits, polling and the rules for publishing. Routes to the more specific revid-* skills when one fits.
review-study
Review a QuestionPunk survey before launch for wording, answer choices, branching, and missing settings. Use for check my survey or review this draft before publishing.
manage-study
Find QuestionPunk studies, check status, publish an explicitly requested draft, or retrieve its participant link. Use for find my survey, publish this study, or how many people responded.
create-study
Create a saved QuestionPunk survey draft from a research brief or edit specified questions. Use for requests such as make a customer feedback survey or shorten my onboarding study.