Video & Animation

Produce and revise faceless long-form YouTube videos or assemble supplied footage in ChatGPT Work, from vidIQ topic research and approved scripts through ElevenLabs narration, Canva and stock visuals, measured captions, MP4 rendering, QA and review. Use for video production, resuming an approved video, scene timing, captioning and final assembly. Exclude standalone thumbnails and unrelated voice agents.

口播短视频"看懂再剪"工作台。把一条口播视频丢进来，自动理解（本地转写、分句、抽帧、镜头切换、人脸安全区、静音与语速），出一份剪辑方案与故事板，选好动效后逐帧渲染成片并生成抖音封面。口播原声与口型 100% 保留，视频不出本机。触发词：自动剪辑、视频剪辑、口播剪辑、看懂视频、视频理解、转写、字幕、自动配乐、视频动效、动效模板、故事板、分镜、封面、抖音封面、剪一条视频、给视频加动效、auto-video-editor、video editor。

Validate, export, and inspect multiple Adobe Premiere Pro deliverables with explicit pending and unverified states. Use for YouTube, social, web, mobile, archive, audio-only, versioned export, delivery QA, caption validation, direct export, AME submission, or repeatable batch output through the Premiere Pro MCP server.

Produce vertical, square, and platform-specific versions of an Adobe Premiere Pro sequence while preserving the master edit. Use for 9:16, 1:1, and social cutdowns, Auto Reframe workflows, safe-zone checks, title repositioning, or multi-platform exports through the Premiere Pro MCP server.

Import, verify, structurally validate, and export timed captions or subtitles in Adobe Premiere Pro. Use for SRT workflows, supplied timed transcripts, caption QA, accessibility checks, or sidecar subtitle delivery through the Premiere Pro MCP server.

Compile a product or technology architecture into a one-page animated A4 blueprint (live HTML, poster PNG, HyperFrames MP4) from a small YAML spec. Use for animated architecture explainers, "A4 architecture page with motion", HyperFrames architecture videos, or turning a system analysis into a motion one-pager.

Makes a product launch, feature demo or ad video from a website URL. Asks a short brief (what to show, for whom, the end action), crawls the home page and up to 20 inner pages headlessly, ranks the site's use cases in its own words, records real click-throughs of each (clicks, typing, ⌘K, scrolling, desktop and phone size) with a real cursor, and cuts them from a shot library (cursor POV, keycaps, tilted plane, macro rack focus, isometric stack, push-through, split, orbit, whip, grid, type) in a measured style (Linear default, Apple, Stripe, Arc, Raycast). Scores its own music and UI sounds (synthesised, nothing to licence), cuts on the beat, -14 LUFS. 16:9, 9:16 filled by the product, 1:1. Use when the user pastes a website and says "make a launch video for <url>", "product video", "demo video of my site", "turn my landing page into a video", "make an ad from my website", "feature video for <url>". Not for talking-head footage (cut, style-edit) or measuring a creator (creator-teardown).

Put background music under a FINISHED video that has speech (an mp4 with voice): several cropped stretches of the mood-music library in videos/ASSETS/music, one per section of the story, crossfaded, ducked under the voice, with real musical drops landing exactly on the key lines (a number, an accusation, the call to action). Script-driven: you write one small JSON file, the scripts do the audio work and verify it. Use for "put music in this video", "add a music bed / background music", "make the video more dynamic with music", "use the music library". Keeps an intro's own audio untouched when asked. Not for wiring music into a HyperFrames composition during a build (/beat-bed), and not for mixing audio that is already in a composition (/hyperframes-audio).

Edit footage (a walk-through, a timelapse, a phone video, someone talking on camera) into a 9:16 HyperFrames video through seven traced states — script, audio-analysis, video-analysis, video-edition, sound-effects, music, deliver — where every state records its inputs, params, exact commands, outputs, checks and errors, so a broken step can be found and reproduced at a glance (state.json, TRACE.md, log.txt, status.mjs, trace.html). Two ways in: PLANNED (a topic, text and reference images/clips -> a Spanish teleprompter script -> the user records themselves on camera -> the takes are aligned to the script and trimmed, retakes and stumbles cut out) or FOOTAGE-FIRST (an unplanned video -> a script that follows it beat by beat with a word budget per stretch -> the user records a voiceover -> each paragraph is placed over its stretch and the footage's own sound is ducked under it); footage with no voice skips the script. Cuts pieces at different speeds, maps verified events (things to count) onto the cut, lays ...

Create polished short brand films with art direction, continuous 3D motion, kinetic typography, sound design, and verified MP4 delivery. Use for premium brand animations, logo-led concept ads, or artistic product promos; adapt the visual metaphor to the brand rather than copying the example.

Use when making anything that moves: motion graphics, animated or explainer videos, product demos and launch films, social clips and reels, logo stings, kinetic type, showreels, B-roll over footage, Lottie files, or UI and website animation, and whenever Remotion, HyperFrames, GSAP, Motion (Framer Motion), Lottie or ffmpeg rendering is involved.

Turn a document, script or notes into a short animated explainer video (MP4), written as HTML and rendered locally with HyperFrames. Use when the user asks for a motion video, an animated explainer or a "Claude Motion"-style animation. Free starter with two styles.

Especialista en captions/subtítulos animados para video con código (Remotion e Hyperframes): estilo Hormozi, karaoke palabra por palabra, typewriter, minimal, cinemático y más, con diferentes tipografías. Usa esta skill cuando el usuario pida añadir, crear o mejorar captions/subtítulos de un video, elegir fuentes para ellos, sincronizar texto con voz, o replicar un estilo de caption que vio en redes.

Dựng video ngắn 9:16 thật "fancy" bằng HyperFrames (HeyGen, HTML + GSAP → MP4) từ một CHỦ ĐỀ hoặc SCRIPT — beat sheet, chọn block catalog theo từng beat, chuyển cảnh có ngữ pháp, camera luôn chuyển động, safe zone TikTok/Reels, karaoke caption, QC bằng snapshot rồi render. Use when the user asks to "làm video hyperframes", "dựng video fancy", "edit video bằng hyperframes", "video kiểu halftone/giấy in", or invokes /hyperframes-fancy.

End-to-end production pipeline for making HyperFrames videos (write HTML, render MP4). Use for any request to plan, author, check, preview or render a HyperFrames video — promo, product launch, explainer, title card, captioned clip, motion graphic. Orchestrates brief -> scaffold -> offline asset vendoring -> composition -> check -> preview approval -> render -> ffprobe verification, and encodes the gotchas that break naive agent builds (blocked CDN needs a locally vendored GSAP, FFmpeg often missing from PATH, the <audio>-id silence bug that renders the whole video mute, the preview-before-render gate). A thin orchestrator that sits on top of the hyperframes-* skills rather than duplicating them.

Generate sound effects from text descriptions using ElevenLabs. Use when creating sound effects, generating audio textures, producing ambient sounds, cinematic impacts, UI sounds, or any audio that isn't speech. Supports looping, duration control, and prompt influence tuning.

Make animated explainer videos, GIFs, single-image flow diagrams and PDF carousels from a topic - narrated, captioned or silent. Choose landscape, square, LinkedIn 4:5 or vertical 9:16, a clean or hand-drawn sketch style, and a light, warm or dark theme. Use when the user wants a video or animated diagram that explains a technical idea, an arrow-flow image or GIF, a LinkedIn carousel, or to turn an article, repo or concept into a short visual explainer, with or without a voiceover.

把任意主题制作成 3 分钟以内的手绘卡通风格科普教学短片（JS + Canvas 逐帧渲染成 1080p MP4，无配音）。当用户要求'做个教学视频/科普短片/手绘动画视频/讲解视频'，或给出一个主题、文章链接希望生成讲解视频时使用。关键词：教学视频、科普视频、手绘动画、科普短片、Canvas 动画、讲解视频

Use when the user wants a short explainer animation or video (math, algorithms, AI agents, system flows, any concept) generated purely from code with no external image or audio assets, optionally with synthesized background music and TTS narration (配音/解说/背景音乐), or exported to MP4. Not for live-action footage, real image/audio assets, or editing long-form video.

根据商品图文和参考视频，端到端完成商品视频复刻或创意改编：核对素材、抽帧理解、选择一至三种方案、长视频分段或剧情压缩、调用视频模型、拼接与音轨处理、验收、版本归档及可选标注平台发布。适用于普通口播、商品展示、情景剧及混合视频；支持严格复刻、参考生成、时间轴重建、局部替换、首尾帧和跨品类剧情改编。默认提供 Codex 视觉理解与小云雀 Seedance CLI 适配，也允许替换理解 API、视频生成服务和发布平台。

>- Make narrated motion-graphics explainer videos entirely in code with Remotion (React + TypeScript): vertical 1080x1920 TikTok/Reels/Shorts explainers or 16:9 educational films on any topic (science, history, how things work, a product). Covers the whole proven pipeline: fact-checked research, script, voice-over first, word-synced chapters built and polished by agents, kinetic captions, sound effects, music, loudness master and QA, with a starter template and a catalogue of rules learned the hard way. Use this skill whenever the user wants an explainer video, an animated educational or science video, motion graphics, a Kurzgesagt-style animation, Remotion compositions or animations, a short that explains something, or wants to turn a script, article or topic into an animated video, even if they never say "Remotion" or "motion graphics".

Write the 20–45 second voice-over script of a mascot explainer video and its beats.json — which pose, card, title or background lands on which phrase. Use when starting an explainer or app-promo video in the explainer-video-pipeline format, when its rhythm feels flat, or when the script changes and the beats must follow.

Produce a vertical "faceless" explainer or app-promo video in which a mascot sticker changes pose on every phrase, captions light up word by word, and UI cards pop in sync with a voice-over — rendered from HTML with HyperFrames, no video-generation model. Use when the user wants a TikTok/Reels/Shorts video for a product, app or idea in this mascot-and-voice style, wants to set up a project for it, check its prerequisites, or rebuild one after changing the script, voice or beats.

Produce the voice-over of a mascot explainer video and its word-level timings — from the user's own recording, a free local voice clone in VoiceStudio, ElevenLabs, or a macOS placeholder — then check it says the script. Use when an explainer-video-pipeline project needs its narration, when the user wants to clone their voice, change provider, fix a mispronounced name, or speed the delivery up.