Video & Animation

Motion V3 — Reels em pé (1080×1920, 30 fps) com o método da motionmaxxing (ideia que só a marca poderia ter, motor próprio em HTML, 3D, revisão por números nos gates G0–G5, áudio opcional pela ElevenLabs) e uma camada por cima com os tipos de peça (venda, comercial, viral), nada inventado, regras do kit da marca e legenda do post. Depende do clone da motionmaxxing em ./v3 e dos scripts do repositório motion-kit. Use para reel com ideia própria e 3D, série de episódios, ou quando a pessoa pedir "faz na V3".

Edit a raw talking-head video into a vertical short (TikTok, Reels, Shorts). Cuts dead air, burns 2-3 word captions inside the platform safe zone, boosts the voice and mixes in quiet music. Use when the user drops a raw .mov/.mp4 and asks to edit it, cut it, caption it, or "make it look like a reel".

Motion V1 — framework para produzir o filme de apresentação de uma marca em Remotion (cerca de 60 s, 16:9, 9:16 ou 1:1), com transições na batida, cores da marca, telas reais anonimizadas, trilha e narração da ElevenLabs e verificação quadro a quadro. Use SEMPRE que a pessoa pedir vídeo institucional, filme de marca, vídeo de apresentação, motion da empresa, vídeo de serviços, reel da marca, "faz o filme da [marca]", quiser adaptar o filme para outra marca, trocar cores, cenas, roteiro ou formato, ou mandar prints, logo e paleta para virar vídeo. Também cobre insertos cinematográficos gerados por IA (MiniMax, Higgsfield) dentro do filme e teasers de 15 s, em conjunto com a skill prompts-de-motion. Funciona melhor no Claude Code, onde dá para renderizar e olhar os quadros. Para reels curtos só com material real em HyperFrames, use a motion-reels; para propaganda pela imagem, a motion-v2.

Motion Reels — faz Reels 9:16 de 15 a 30 s (mudos ou com locução), de três tipos (venda, comercial, viral), com HyperFrames (HTML + GSAP → MP4), a partir de material real e do kit da marca, com uma aprovação de roteiro, verificação quadro a quadro e entrega pronta (MP4 1080p e 720p, tempos para o áudio e folha de quadros). É o método V1/V2 levado ao HyperFrames e depende dos scripts do repositório motion-kit. Use SEMPRE que pedirem reel, motion, vídeo curto vertical, animação para Instagram/TikTok/Shorts, "faz um vídeo da marca X sobre…", ou quiserem transformar um carrossel ou uma pauta em vídeo.

Шаги 2–4 конвейера ролика — в worktree ролика произвести вертикальный космический Shorts (≤ 30 с, только EN) для The Universe Inside Out — факты, визуальный бриф, сценарий, шоты, hero-кадры и черновое видео на одобрение, затем финал через общую очередь двух рендер-ПК до release (MP4 + publish_en.md) и влитие ветки в main. Использовать при «делаем вариант N», «/make-video», «запускай ролик про X», «пакет к загрузке».

Шаг 1 конвейера ролика — завести worktree ролика и предложить 3 варианта следующего космического Shorts (≤ 30 с, EN) для The Universe Inside Out — объект, крючок, money shot, факты с источниками, план шотов, оценка работы и рендера — для выбора пользователем. Использовать, когда пользователь говорит «предложи следующее видео», «что снимаем дальше», «/next-video», «варианты ролика».

>- Write and lint the title and thumbnail for a YouTube video as one pairing, checking truncation, duplication and vagueness before publish. Use for "title ideas", "what should I call this", "thumbnail text", "my CTR is bad", packaging, or any request to rename or repackage an existing video.

Create or revise result-first AI/software tutorial Reels: genuine cursor-led app walkthroughs or source-footage-to-cinematic transformation demos. Use for AI Demo Reel and show the result then how I made it; not generic AI-topic videos, split-screen lectures or montage-only edits.

>- Madhav's "Case File" investigative explainer: a documentary "evidence room" look adapted for Reels / Shorts / YouTube (charcoal ground, white-glow photo cards with torn name tags, dark street-map "system maps" with a glowing red route, white clay 3D dioramas, testimony quote cards with real sources, red date stamps, footnote markers, serif chapter titles), driven by a calm narrator over a continuous dark-ambient score. Use when asked for a "case file", "evidence room", "investigation", "true-crime / documentary style" or "how X was discovered / hacked / leaked" video, for a tech incident, breach, outage, scandal or history story. Runs inside the animated-video skill's stages and approval stops. Not for upbeat concept explainers (vector-world-explainer), paper collage over presenter footage (paper-lab-reel) or plain footage edits (video-use).

Design Madhav's educational Reels with a large moving presenter below readable teaching visuals. Use to enlarge his face or HeyGen avatar, replace a tiny inset, or reuse this split-screen teaching style; not a blanket restyle of approved videos.

Turn arbitrary text — an article, notes, a topic, a brief — into a faceless explainer video: there is no site or footage to capture, so the visuals are invented per scene (typography, abstract graphics, diagrams, data-viz). Use for topic explainers, concept breakdowns, how-tos, listicles. Not a video built from a website (/product-launch-video — promo or tour). Unclear → /hyperframes.

把一句话选题写成短视频文案。默认只输出文案与发布文案，分镜表按需再给。按平台/品类/调性预设组合，带计时模型与标点规范。当用户说"写个短视频文案/口播稿/脚本"时使用。

Use this skill when the user requests AI digital human talking-head videos, voice cloning with consent, HeyGen avatar video generation, short-video scripts, subtitles, video packaging, or a repeatable topic-to-MP4 production pipeline. Works in Codex and Claude Code; prefer external executable tools over claiming model itself can edit video.

>- Turn a raw recording's transcript into an edit decision list - dead air, filler cues and retakes, with timecodes. Use for "edit this", "cut the dead space", "tighten this video", "I rambled", or any request to shorten footage from a transcript.

>- Audit a YouTube channel end to end - packaging, consistency, the first fifteen seconds, and what to fix first. Use for "audit my channel", "why isn't my channel growing", "review my videos", or a pasted channel URL.

>- Write YouTube chapters from a transcript, validated against YouTube's own rules so they actually render. Use for "add chapters", "timestamps", "break this video into sections".

把「我想拍点什么」变成一条想清楚的片子——先定主题（出 5 个候选、标明类型、给出推荐与淘汰理由），再定结构与分段剧本，最后切成镜头级内容分镜。支持两种入口：一句话主题（需要发散选题）与已有拆解剧本（跳过选题，先问复刻需求与复刻强度再判定改写）。当用户提到选题、定主题、短视频策划、写剧本、写大纲、分镜、镜头脚本、不知道拍什么、内容怎么排、科普干货、剧情软广、钩子怎么写、复刻别人的剧本时使用。只管内容层（讲什么、怎么讲），不碰执行层；内置的内容红线只防明显自伤，不替代合规审查。

把已定稿的分镜表翻成可直接执行的两份清单——资产清单（要准备哪些图、每张图的图像提示词）与生成清单（每段视频的时长、模式、素材引用与提示词）。覆盖画面化、资产编号、切分生成单元、Seedance 视频提示词写法、首帧方案、一致性与失败诊断。当用户说「把分镜变成提示词」「写视频提示词」「这条片子要哪些图」「生成清单」「某个镜头生成不出来」时使用。只做翻译与清单，不重新编排内容，不直接出图出片。

把已过 G2 的生成清单落成真实视频片段——按段建 KIE Seedance 任务、轮询、下载、媒体体检、抽帧质检、听声核台词，并把每段的状态与请求指纹写进生成台账。三种模式互斥（文生 / 首帧图生 / 首尾帧图生）。旁白由模型直接念，台词照稿写进提示词，不走独立 TTS 配音。当用户提到视频生成、出片段、跑 Seedance、首帧图生、首尾帧、原生音频、台词念错、任务查询、片段下载、生成失败重试时使用。只管「把已定的画面和台词变成成段素材」，不重新设计视觉、不改台词文本、不做字幕、不做剪辑。

>- 把任意内容（PDF / PPT / DOCX / Markdown / 网页 / 视频 / 一个主题）制作成“代码渲染”的可视化讲解视频： 先做教学结构，再写分镜与旁白，然后用 Canvas/HTML 的 render(t) 画每一帧，headless Chrome 逐帧截图， FFmpeg 编码，可选 TTS 旁白、合成配乐、软字幕、质检报告。 支持 16:9 / 9:16 / 1:1 / 4:5，内置 5 套视觉风格与 8 类通用镜头，也支持自定义镜头与品牌。 当用户说“把这个 PDF/PPT/文章做成讲解视频”“做一个科普/产品/教程短片”“可视化教学”“代码做视频” “explainer video”“video as code”时使用。 不用于：直接调用 Sora/Veo 等视频生成模型、剪辑已有实拍素材、或只写文案不做视频。

一条短视频从策划到发布的流程总控：建制作目录、依次走「剧本分镜 → 提示词 → 生成 → 交付 → 发布」五段、每段设门禁、记录流程状态、支持跨会话续做。当用户要开一条新片子、问「下一步做什么」、想接着做某条没做完的片子、要把分镜变成提示词、要整理素材与成片、要准备发布包时使用。本技能只规定流程、目录与产物标准，不重复各环节的写作规范。

Create a new editing-style skill from 2-3 reference reels of one creator or style (e.g. "make me a skill that edits like @creator"). Downloads the reels, measures them frame by frame, has one agent break down each reel, merges the breakdowns into a style guide plus a measurable style spec, clones the proven editing engine into a new /<style>-edit skill, then locks the style with a match score, pixel font matching, a golden recreation test, one critic round and the user's ear check. Free local tools only. Use when someone says "/new-edit-style", "make an edit style skill", "copy this creator's editing style", "I want to edit like X", or drops several reels from one creator and asks for their style.

为视频生成准备可用的图——把资产清单变成能直接喂给图生视频的素材：排生产顺序、定资产级别、出候选图、做一致性验收、按编号落盘、回写清单状态。支持三种来源：实拍、AI 生成、官方素材，不强制实拍。当用户说「准备素材」「出资产图」「首帧图」「角色三视图」「产品图跑一下」「资产对不上」「这条要哪些图」时使用。只管把图做出来并验收，不写分镜，不直接生成视频。

Produce a narrated, captioned 1080p demo video (3 minutes or less) for any software project, hackathon entry or product, as code, with Playwright and ffmpeg-static. Records the real app, replays real terminal proof output, tours real source code, draws diagrams, adds speech (Windows built-in voice by default, ElevenLabs optional) and burns in captions. Use when asked to make, redo or re-time a demo, pitch or submission video, or to prove on video that a project works.