Video & Animation

Turn photographed or Markdown English reading worksheets into source-grounded narrated Remotion lesson videos through a user-confirmed teaching plan and narration script, followed by sentence teaching, exercise solutions, exact paragraph evidence, ASR-derived timing, reproducible local rendering, and visual QC. Use for repeatable English reading video production in this repository; not for ChatCut timelines, generic OCR, math lessons, or data-story videos.

Master editor for vertical talking-head short videos (Reels, TikTok, Shorts). From one raw clip it transcribes, removes bad takes and silences, detects the niche (24-genre matrix), asks the user to choose style options, then renders a smooth, professional edit with golden-ratio kinetic captions, meaning-driven motion graphics, licensed B-roll, music, SFX and fonts, color correction, and QA. Use it whenever the user gives a raw talking-head, podcast-clip, coach, founder, UGC or explainer video and wants it edited or "made viral", types /talking-head-short, or asks for captions, motion graphics, B-roll or a reel edit, even if they don't name the skill.

Runs a full review loop on any video. Puts it on a private, frame.io-style review page where the user leaves notes pinned to frames, whole-video notes, ranges and drawings; then reads the notes back, says what can and cannot be done, re-edits the video, checks the result and puts up the next version with a what-changed list to tick off, until a version is approved. Use whenever the user wants to review, critique, mark up or leave notes on a video, edit, cut, ad, reel or rough cut; says "put this up for review", "review link" or "frame.io"; says "apply my notes", "apply the changes", "notes are in", "I've submitted my notes", or pastes a list of timecoded notes; gives feedback on a cut in chat ("it's a bit slow", "the music is too loud"); says a cut is approved or final; or a comment from the review page asks to apply notes, go ahead with a plan or show a file. Also use right after you edit or render any video for the user, to put it up for review instead of asking for feedback in...

Plan and produce product-ad storyboards and video from a consistent white-background product reference, with shot-by-shot prompts and an explicit review checkpoint. Use for product commercial photography, image-to-video advertising, or multi-angle product motion tests.

Edit footage and make videos that look professionally finished, then prove it with measured checks. Editing: word-level transcripts, cutting pauses, fillers and retakes, jump-cut punch-ins, captions (clean, karaoke, short-form punch), reframing to 9:16/1:1, voice clean-up, music ducked under speech, platform loudness and exports. Making: brand intros and logo reveals (stings), product launch spots and social ads, kinetic typography, rendered frame-by-frame from HTML with designed motion and synthesised sound. Use it for any request to edit, trim, tighten, caption, reformat or export a video; to turn a talking head, tutorial, podcast or screen recording into a YouTube video, Short, Reel or TikTok; or to create a brand intro, logo animation, launch video, promo, ad, title sequence or motion graphics. Also for checking a video before it ships (loudness, black frames, safe areas).

把会动的东西按固定时间一帧一帧录下来，做成左右对比的 GIF 或逐帧对照图，给人验收、对比改动前后、放进说明或 PR：网页的 CSS 过渡和动画、element.animate()、requestAnimationFrame 写的 JS 动画、canvas 游戏、three.js / VRM 角色、PixiJS / Live2D 角色都适用。附脚本：record.mjs（无头 Chrome + 假时钟，页面不用改，渲染多慢时间轴都准，两次录的逐帧一样）、make_gif.py（几个变体并排、标签、共用调色板）、clock-shim.js。还讲先逐帧记数值再录 GIF、开关前后怎么做变体、怎么取景，以及实测过的坑（Chrome 虚拟时间不可靠、等加载时别推时钟、Vite 首次打开会刷新页面、Live2D 运行库要先加载、无头 Chrome 用完要关、别人模型的许可）。只要用户想「录个 GIF」「做个动图给我看」「对比一下改之前和改之后的动画」「把动作录下来」「验收一下动画效果」，或者要给动画、动效、角色动作拿出能看的证据，就用这个 skill，即使没提 GIF。Also use for English requests to record an animation as frames or a GIF, make a before/after comparison of motion, or capture deterministic frames from a web page, canvas, three.js, VRM or Live2D scene.

在网页或桌面应用里让 Live2D（Cubism，PixiJS）和 VRM（three.js + @pixiv/three-vrm）角色动起来：加载顺序、程序姿势和动作片段互相覆盖、在物理之前叠加、VTube Studio 模型补齐（没有 Idle、没眨眼、头角度是物理输出、8K 贴图、包围盒被特效带偏）、动作通道和定住、表情存取和道具冲突、three.js AnimationMixer 的坑（不写没变的值、LoopOnce 提前关掉）、程序「带动层」让手 K 和 AI 动作不僵、动捕片段只取上半身、弹簧骨骼延迟和跟节拍、裙子穿模、让 AI 生成表情和动作的数据格式与校验规则，以及逐帧记录的验证方法。只要用户在做虚拟形象、桌宠、VTuber 工具、数字人，提到 Live2D、Cubism、model3.json、motion3、exp3、physics3、VTube Studio、VRM、VRMA、three-vrm、Mixamo、动捕、SpringBone，或者抱怨「动作一顿一顿的」「动作僵」「手放下又抬起来闪一下」「头不跟着动」「定不住」「表情看不出来」「做动作脸上没表情」「手指僵、直板手」「托腮够不到下巴」「眼睛呆、一直盯着」「腿站不稳、脚滑」，就用这个 skill。Also use for English requests about animating Live2D or VRM avatars, procedural idle motion, physics, expressions, or AI-generated character motions.

Use when the user wants clips, highlights, shorts or reels cut from a gameplay recording, stream VOD or a folder of recordings. Runs the local clip-anything CLI: previews the picks, renders 9:16 clips, explains each pick from report.json. Not for editing a clip by hand or for uploading.

Make animated, narrated tech explainer videos ("what actually happens when you type X into Google", "how does HTTPS work", "what happens when you git push", "explain this code") as 1080p MP4s with hand-drawn line-art animation, local TTS narration (a built-in voice or the user's own cloned voice), a ducked music bed and sound effects synced to the visuals. Use this whenever the user wants an explainer, walkthrough, animated diagram video, "make a video about how X works", a video from their code, a social clip for X/Twitter, YouTube, Reels or Shorts about a technical concept, or wants to change narration, voice takes, timing, music or SFX of an existing explainer project, even if they don't say "skill" or "template".

口播动效剪辑管线。把已经录好的中文口播素材剪成 3:4（1080×1440）动效成片，走知识类口播常见的"真人段 ↔ 图形段"交替剪法：剪口误、剪停顿、对字幕走基座 cangshu-koubo-wudongxiao，再找脸裁 3:4、macOS Vision 抠像，按逐词时间戳写 plan.json，由 Remotion 模板渲染真人段的毛玻璃数据卡、逐字升起的大字、顶部进度导航，以及逐镜头手写的图形段（一个主角演一条链、元素状态联动、停稳再推拉的 3D 镜头；白卡或 9 套动画片主题可选，图形段背后可加品牌字墙），只在关键联动处配偏亮的实录音效；渲完一条命令质检（每秒 5 帧分镜查穿模越界、纯音效音轨逐个测响度）。也管这条片子配套的 BGM 试听连播、用户给的字幕（重点词放大标黄）、风格样片、系列固定片头、3:4 封面，也回答"3:4 口播怎么拍"。用户说「走口播剪辑管线」「口播动效剪辑」「把这条素材剪成 3:4 动效版」「口播加动效 / 加 B-roll / 做图形段」「出一版 3:4 成片 / 预览」「给这条配几个 BGM 试试」「换几个动效风格看看」「给这条做个 3:4 封面」「3:4 口播怎么拍」时使用。除拍摄建议外，输入必须是已录好的视频；不用于写稿、改稿、选题、录制前的提词稿和演示页；只剪口误和长停顿、不要 B-roll 和 BGM 出 9:16 竖屏的用 cangshu-koubo-wudongxiao。只在 macOS（Apple 芯片）上跑。

制作作品集级动态图形视频(15/30/60/自定秒):纪录片证据流 + 液态玻璃设计系统 + 数据驱动时间轴 + TTS 旁白 + 分块渲染与完整验收。当用户要求 动态图形视频 / 作品集视频 / motion graphics / portfolio video / 带旁白的宣传片 / 模型发展史视频 时使用。开工前必须询问用户三件事:需求主题、主模型、TTS 模型, 推荐使用Qwen系列模型。

製品・アプリ・機能の紹介動画（プロモ動画、PV、デモ動画、ローンチ動画、説明動画、モーショングラフィックス）を、コードだけで作る。HTML で組んだ場面を 1 フレームずつ撮影して MP4 にし、曲と効果音も合成する。Remotion・GSAP・動画編集ソフト・素材ファイルは不要。「紹介動画を作って」「このアプリの PV / プロモ / デモ動画がほしい」「リリース告知の動画」「機能を説明する動画」「make a product / intro / launch / demo / explainer video」「animate our UI for a video」など、製品やサービスを見せる動画を頼まれたら、動画という言葉が出た時点で必ずこのスキルを使う。このスキルの雛形で作った動画の手直し（場面の追加、文言や曲の変更、英語版・別の尺）にも使う。実写の編集、既存の動画ファイルの切り貼り、画面録画の加工には使わない。

Makes ready-to-post 1080x1920 Instagram Reels / TikToks / Shorts from one prompt. It writes the script (or uses the user's), generates the voiceover, animates every scene in a proven motion-design style, reviews its own frames, and exports an mp4. Use whenever the user asks for a reel, short, TikTok, explainer video, faceless video, animated video for social, or says /ai-video.

Turns timed words into on-screen captions for vertical video. Status: **implemented and tested**.

Turn a story, idea or voiceover script into a finished hand-animated 2D cartoon short (vertical 1080x1920 mp4, about 35-45 s) with character animation, a VoiceStudio voiceover, burned-in captions, sound effects and music, plus an Instagram caption. Use this whenever the user runs /animation-video or asks for an animated video, cartoon, animated short, reel, animated explainer, "episode", or "another one like the Mall Landlord / Link Jungle video", or pastes a logline, scene list or voiceover script and wants it animated, even if they never say the word "skill". Does everything end to end with no questions asked - script, voice, animation, sound, render.

Rancang video Remotion agar visualnya punya karakter, background dan gerak yang hidup, serta tidak terasa seperti slideshow generik. Gunakan saat membuat storyboard atau scene.

Edit raw video footage into finished videos and vertical shorts, locally with ffmpeg and Whisper. Use when the user wants to set up a video editing folder, edit a new recording, cut out mistakes and pauses, add captions or subtitles, make YouTube Shorts / Reels / TikToks from a longer video, fix audio levels, or check a finished video before posting. Also use when the user says "edit my new video", "make shorts", "set up my video studio", or drops footage in their studio inbox.

Paint scenes in code and turn them into video — oil-painting look built from real brush strokes (bristles, impasto, dry-brush, canvas weave), spectral pigment mixing, a paint-on reveal (the picture being painted stroke by stroke) and a "living painting" phase (wind, drifting clouds, shimmering water, parallax camera). Stack — WebGPU/WGSL instanced strokes in headless Chrome, WebCodecs + Mediabunny MP4 export, ffmpeg QA. Use when the user wants a painted / painterly / oil / hand-painted / "como pintado" animated scene, a speed-painting or timelapse-style video, an animated landscape or background, a painted look for a story/music/ambient video, or wants to study a reference painting's composition, values and palette. Triggers — "pintar", "pintura animada", "escena pintada", "óleo", "painterly", "paint a scene", "speedpaint", "living painting".

Test every function of a code repository for real, then record a narrated, captioned, chaptered 1080p walkthrough video (MP4 + SRT + per-chapter files + README) that explains the project to humans and the team. Terminal scenes are real commands recorded with VHS, the voice is macOS `say`, slides are drawn with Pillow, and ffmpeg assembles everything with chapter markers and a subtitle track. Use this skill whenever the user asks to record, film, screencast, or "take a video" of a repo or its features, wants an onboarding / demo / walkthrough / tour video of a codebase, wants to "test everything and show it", or wants a video explaining a project for a team, even if they don't name VHS or ffmpeg.

Explains any idea with simple shapes, moving dots and clean animation, as a 1080x1920 MP4. Use for /minimal-graphics with an idea or question, or any ask for a minimal explainer animation.

Turns an idea or question into a cartoon explainer acted out by stick figures, pets and props, as a 1080x1920 MP4. Use for /stick-figure or any ask for a stick-figure animation.

Makes a 3D map video with the real route drawing itself between two places (1080x1920 MP4). Use for /map-route with two places, or any ask to animate a route, trip or directions.

Turn an idea or topic into a finished vertical explainer video (YouTube Shorts / Reels / TikTok, 1080x1920 MP4 + subtitles + cover) in the dark-canvas, hand-lettered, one-continuous-diagram tech-explainer style (like KodeKloud Shorts) — locally and free (Remotion + Kokoro TTS, no paid APIs, no video models, no face). Use whenever the user asks to make/generate/create a short, reel, explainer video, or "video about <topic>", wants to explain a technical concept (how X works, X vs Y, system design, interview question) as a video, or says "idea to video".

Turns your numbers into an animated chart video that builds itself on screen (1080x1920 MP4). Use for /graph with numbers, a table or a CSV, or any ask to animate a chart.