Video & Animation

Translate video subtitles (SRT, VTT, TXT) between languages. Supports batch translation and maintains timing synchronization. Use when creating multilingual subtitles, expanding to international audiences, or translating video transcripts.

Add a measured-quality soundtrack to ANY video: use whenever a video is being created, generated, rendered or edited (Remotion, Hyperframes, Veo, Sora, Runway, HeyGen, ffmpeg, screen recordings, demos, trailers, reels) and has no music, or when the user asks for background music, BGM, a soundtrack, a loop, 配樂, 背景音樂, 做音樂. Composes offline (8 moods, no downloads), fits the exact video length, changes sections on scene cuts, self-evaluates and iterates, and ducks under narration. Optional neural tier (ACE-Step) for richer music. For voice/TTS see voicebox.

Edita un video vertical donde una persona habla a cámara (reel de Instagram o TikTok) y le agrega zoom en cada corte, un gancho legible desde el primer cuadro, títulos, subtítulos karaoke, un chat que se escribe solo, una lista tipo Finder, capturas en abanico, pantalla dividida con otro video y un botón de seguir al cierre. Úsala cuando digan "edita este video", "hazlo más dinámico", "ponle subtítulos", "agrégale zooms y efectos", "que se vea profesional", "edítalo para reels" o pasen un video vertical hablado. Renderiza con Remotion desde public/plan.json.

Use whenever a task involves designing or reviewing anything people read or interact with (UI, web, documents, video, social posts, forms), writing alt text or annotations, or any accessibility or compliance question, even when accessibility is not mentioned: legal dates changed in 2026 and are verified here. Contains: WCAG 2.2 A/AA with what the designer must decide per criterion, the 9 new 2.2 criteria, AAA worth adopting, inclusive design, cognitive accessibility (COGA), plain language, motion and flashing, captions and audio description, the alt-text decision tree, focus and keyboard specs, forms and errors, screen-reader annotation, laws (EU EAA and EN 301 549, ADA Title II 2027/2028, Section 508, UK, Canada, Brazil, Japan), tool coverage, and test checklists.

Investiga un hecho de actualidad (un aluvión, una tormenta, un anuncio, una noticia que se está moviendo) leyendo varias fuentes y devuelve una ficha de hechos con la fuente de cada dato, las cifras que no calzan entre medios y lo que todavía no se sabe. Es el primer paso para hacer un video explicativo o un post sin decir algo falso. Úsala cuando digan "investiga qué pasó con X", "averigua las causas de", "qué se sabe de", "hazme un video explicando lo que pasó" o "chequea este dato antes de publicarlo".

Operate a private Personal Operating Memory system using Codex plus a markdown vault. Use when asked to initialize, capture, remember, query, review, summarize, or hand off durable personal memory across emotion, relationships, work, finance, health, and life admin; ingest, summarize, classify, fact-check, and discuss articles, videos, reports, papers, books, links, transcripts, and reading notes as an OKF-style knowledge catalog with source digests, claims, ideas, topics, citations, and searchable indexes; provide warm user-prioritized relationship support; use Perplexity when current external evidence is needed; keep private facts in the vault, not the skill; and avoid replacing therapy, emergency, investment, tax, legal, or other professional advice.

Write, fix and review JMathAnim (Java) animation scenes. Use when code imports com.jmathanim, extends Scene2D, or the user wants an animated explainer or math/diagram video rendered with JMathAnim in Java. Covers layout, text, timing to narration, arrows, images, and the library's silent-failure bugs in 1.5.1. Not for Python Manim.

构造主义解说视频工作流的**一次性环境配置**技能。下载项目后首次使用：只读探测本机工具链（Node/Python/ffmpeg/Remotion/LMMS/中文字体），对缺失项检索下载源并补装，把解析出的绝对路径写成「路径契约」toolchain.json，再按契约去开源社区收割各工具的成熟 Agent Skill（拿来主义，harvest-skills.py），做端到端冒烟校验，最后询问是否禁用/归档/删除本技能。当用户刚下载本项目、说"配环境""装工具链""跑不起来/命令找不到"，或生产中引入新开源工具要找现成 skill 时使用。

Design 3D models together with the user inside Autodesk Fusion through its local MCP server (http://127.0.0.1:27182/mcp). Asks up front whether the project is a 3D-printed part (PLA/PETG/ASA and others, default printer Bambu Lab H2C), a showcase piece that will not be printed (acrylic keychains, exploded views for crowdfunding or video), or a clothing and brand mockup, then builds a parametric model from measurements, checks it, screenshots it and exports it. Use when the user wants to model, tune, review or export anything in Fusion.

生产「构造主义解说视频」的四阶段流水线：文稿 → TTS 解说音轨（时间主轴）→ LMMS BGM（用变奏标记论述层次）→ Remotion 渲染成片 + 封面。当用户要做解说视频/知识视频/口播视频，或要做"以解说配音为时间轴、BGM 随论述层次变奏"的成片时使用。前提：已由 ideaform-setup 配好环境并写出 toolchain.json 路径契约。

Remove backgrounds from animated GIFs, image sequences and short video, and export a transparent GIF + WebP + APNG. Use this skill whenever the user asks to "把背景透明", "扣图", "抠图", "去背景", "去白底", "透明背景 GIF/动图", "remove background from GIF", "make this GIF transparent", or complains that a cut-out still has 残留 / 噪点 / 白边 / 倒影 / 没扣干净. Trigger on "抠干净一点", "边缘有毛刺", "还有背景色". Do NOT use it to turn a photo into a cartoon sticker or Q-version meme (use gif-sticker-generator), to restyle or repaint the subject, or to add a background to something that already has transparency.

Work in the Paper design app (paper.design) through its local MCP server (the mcp__paper tools). Use when asked to design, mock up, or explore variants in Paper, edit artboards or slides on a Paper canvas, take screenshots or export PNGs, PDFs, or videos from Paper, move or clone Paper files, use custom fonts or local images in Paper, or turn a Paper design into code. Covers codemode quirks, result parsing, export and screenshot workarounds, canvas organization, and what the MCP can't do.

Specialized skill for discovering local smart TVs, projectors, and casting local video files or streams to the big screen via DLNA/UPnP. Triggers when the user asks to cast a video to TV/projector or mentions "投屏", "放电视上", "大屏播放".

Make AI short videos for TikTok Shop products with Kaloclip via the bundled `clip` CLI (Python, zero install): turn a product link (TikTok Shop, Kalodata, Amazon) or product images into a video script, then into a finished 8–30s vertical/horizontal video; estimate the credit cost first; break an existing TikTok / YouTube Shorts video down into shots, voice-over and camera moves. Use whenever the user wants a product video, UGC-style ad, 带货视频、商品视频、短视频脚本、AI 生成视频、仿写/拆解爆款视频脚本、分镜 — even if they never say "kaloclip".

Reelary, the 100% local AI media library for your own videos and photos on Mac and Windows: install and connect it, index folders and external drives, check indexing progress, remove folders, fix errors. Use when the reelary_* tools are missing or fail, or the user wants to organize footage. Trigger on "Reelary", "media library", "index my videos", "organize my footage", "catalog my clips", "scan my drive", "make my footage searchable", "video library app", "digital asset management for video", "local video search app". Not for cloud storage or renaming files.

Turn a script, outline, shot list, blog post or theme into a video cut from the user''s own real footage (not AI-generated): rough cut, assembly edit, string-out, script to video, montage, highlight reel, recap, sizzle reel. Picks the best clip per line with alternatives, saves an editable project and exports MP4. Local with Reelary on Mac or Windows. Trigger on "rough cut", "assembly edit", "first cut", "script to video", "turn this script into a video", "make a video from my blog post", "text to video with my footage", "montage", "highlight reel", "recap video", "best moments", "year in review video", "trip video from my clips", "wedding highlights", "edit my footage into a video". Not for AI video generation, color grading or effects.

Pull b-roll from the user''s own footage library for a voiceover, interview, podcast or talking-head video. Matches each line of narration to cutaway shots by meaning, 100% local on Mac or Windows. Trigger on "b-roll", "broll", "cutaways", "cover shots", "find b-roll for my voiceover", "illustrate this script with my clips", "inserts for my interview". Not for downloading stock footage or generating AI clips.

Search the user''s own videos and photos by meaning, 100% local with Reelary on Mac or Windows. Finds a shot by describing it (exact start and end times), photos by what is in them, the moment someone says something (transcript search, Pro), every clip of a person (face recognition, Pro) and footage by place from phone, DJI drone and GoPro GPS (Pro). Trigger on "find the clip where", "search my videos", "do I have footage of", "video search", "footage search", "find a photo of", "search my photos", "find where he says", "find the quote", "sound bite", "search transcripts", "every video with my daughter", "face search", "footage shot in Paris", "drone footage of", "geotagged videos". Not for stock footage sites, web image search or identifying strangers.

Search Pexels and Pixabay together for stock photos and stock video, and Pixabay for illustrations, vector art and animations, pick the best match, then download and optimize it for a web project (WebP for images, compressed MP4 plus poster for video). Use for hero images, background videos, card thumbnails, illustrations, or any stock media a page needs.

Turn a maths question, concept, rough script or channel brief into a complete, professional Manim explainer video. It researches the best solution, writes a hooked script that independent reviewers must approve before any code is written, generates the narration first so the animation is timed to the real voice, renders with the channel theme, runs an independent QA pass (pronunciation, theme, sync) and outputs separate video, separate audio, a combined file and a publish kit. Creates and gets approval for the visual theme on first run, then always reuses it. Use when the user says "make a maths video", "manim video", "animate this concept", gives a maths problem/concept, a rough script or a brief from channel/briefs/, or invokes /manim-video.

Make a source-faithful Morandi cinematic photo poster through original title design, restrained small type, negative space, scene-axis alignment and selective subject occlusion. Use for 莫兰迪电影/电影感海报; no decorative collage symbols or copied film branding.

双语阅读——将英文 PDF（技术书、教程、手册等）转换为逐段中英对照、图片 base64 内嵌、带左侧可折叠目录导航的多文件 HTML 阅读版。触发场景：用户提供英文 PDF 并希望双语对照阅读、中英对照翻译成 HTML 阅读版、逐段翻译做双语网页、或提到"英文 PDF 转双语 HTML / 双语阅读 / 中英对照阅读版"等需求。核心规则：原文逐段全译不概要、一章一文件、代码块保持英文、中文段浅蓝底左色条、每译完原书一页加页码标记、文件名小写连字符、🌐 按钮循环中英显示模式。

Post one video to several platforms with PostEverywhere, with the right title and caption for each. Use when the user wants to share a video, Reel, Short or TikTok across TikTok, Instagram Reels, YouTube Shorts, LinkedIn and Facebook Reels, or asks to "post this video everywhere".

Create a self-contained cinematic narrated HTML story animation for a customer story, market-entry journey, product launch, campaign, or learning tutorial using the Story Studio editorial design system with scene motion, captions, always-on narration, voice picker, and interactive playback. Use when the user says make a narrated animation, turn this case study into an animated story, create a customer story animation, make a narrated tutorial, or make a story demo like the story studio.