Video & Animation

Create, revise, and verify precise architecture diagrams for agentic AI, RAG, software systems, robotics, transformers, CNNs, GNNs, audio/speech, video, multimodal systems, and neural networks. Use for architecture figures, research-paper diagrams, model schematics, TikZ/LaTeX figures, editable SVGs, tensor and data-flow diagrams, and code-to-architecture documentation. Produce editable source plus a rendered figure; check semantics, routing, text, and provenance.

Builds robot perception, planning and verification with Gemini Robotics-ER 2 (gemini-robotics-er-2-preview, gemini-robotics-er-2-streaming-preview) through the Google Gen AI SDK 2.x Interactions API and Live API. Use when code calls Gemini Robotics / ER models, or for pointing, bounding boxes, trajectories, grasp points, 2D-to-3D lifting, agentic vision (code execution), task orchestration with robot function calling, video moment finding / progress classification / success detection, real-time Live API robot agents, or migrating generateContent / ER 1.6 code.

Edit video projects, motion graphics, text, captions, effects, audio and exports in AK Screen 3.0 or later through its MCP tools. Use for AK Screen editing requests when the AK Screen MCP server is available.

Renders MuJoCo simulations with Three.js (in-browser WASM, a streamed Python MuJoCo process, or logs). Use for Z-up scene setup, converting MuJoCo geoms and meshes to Three.js meshes, per-frame body pose sync (wxyz quaternions), capturing camera snapshots and canvas video for Gemini Robotics-ER, mapping ER [y, x] 0-1000 points to 3D by raycasting, projecting 3D back to image coordinates, and debugging sideways, twisted, mis-sized or blank renders.

Grow a small business on Threads (Meta). Audits a Threads account from its posts and numbers, finds the one thing holding it back, tears down best and worst posts, writes ready-to-post captions with one job each, plans 30 days, sets a reply routine, watches competitors, and briefs creators for paid posts. Use when someone asks how to get customers, reach, replies or sales from Threads, what to post, why posts flop, how to reply, how to work with Threads creators, or for a Threads content plan or audit — for any small or home business in any country.

Generate images, infographics, speech, podcasts, music, and sound effects with the framechord CLI, using optional reusable character identities and local YAML and Markdown briefs.

Generate images, infographics, speech, podcasts, music, and sound effects with the framechord CLI, using optional reusable character identities and local YAML and Markdown briefs.

Records visual proof while testing UI behavior — the agent tests the app hands-on via computer use while a screen recording with structured test/assertion annotations captures the session — then posts the video and a results summary to the PR and tracker issue. Use whenever a change needs verifiable evidence that it works, instead of prose claims — including headless environments (scripted screenshots and probes) and non-UI changes (measured numbers, output pairs).

Design banners for social media, ads, website heroes, creative assets, and print. Multiple art direction options with optional generated or supplied visuals. Actions: design, create, generate banner. Platforms: Facebook, Twitter/X, LinkedIn, YouTube, Instagram, Google Display, website hero, print. Styles: minimalist, gradient, bold typography, photo-based, illustrated, geometric, retro, glassmorphism, 3D, neon, duotone, editorial, collage.

Install ToonKit Shorts (Python deps, ffmpeg, free Piper voices, whisper timing model, YtKit) and, only if the user wants uploads, walk them through the free Google Cloud OAuth setup for their own YouTube channel. Use when the user installs the plugin, asks to set it up, or uploads fail with auth errors.

Add b-roll motion graphics to a talking-head video. Transcribes the recording with word timings, decides line by line when the speaker is shown, when a graphic is laid over them and when a full-frame graphic covers them, draws each graphic from what is actually said (numbers, comparisons, steps, key words) in JavaScript on a canvas, and renders the finished MP4 with the original sound, plus the b-roll alone with alpha for an editor. Use when someone has a recording of a person talking to camera (a talking head, a spoken explainer, a course lesson, an ad read, a podcast clip) and wants it more engaging to watch, wants b-roll, or wants motion graphics added to it.

Turns a passage from a long recording (interview, podcast, Zoom/Riverside call, talk, webinar) in which someone explains an idea into a split-screen explainer video – the speaker on top, captions, and hand-made motion graphics that visualise the explanation underneath – in 4:5, 9:16 or 16:9, rendered locally with HyperFrames. Use this skill whenever someone wants "the part where I explain X" from a recording as a video with animations, graphics, illustrations or a split screen, wants to turn an interview answer, podcast segment or talk excerpt into a LinkedIn/Instagram/YouTube video "with motion graphics", or asks for "a video like the S4/change one" – even if they do not say "split screen" or "HyperFrames". Not for pure captioning, short motion-only clips or videos without a speaker.

>- Shared voice, human texture, and community-behavior rules for Skool posts. Use before writing any Skool welcome, YouTube drop, retention unlock, network thread, MVP recognition, or identity announcement.

Change the show's look and feel via toonkit/look.yaml - host names, personalities, Piper voices, caption colors/size/position, palette, font, title card and thumbnail style, episode length, topic niche and upload defaults - or switch presets (mort-and-blitz, clean-tech, retro). Use when the user wants to rebrand, rename hosts, change voices/colors/fonts, or pick a different style.

Generate Vidu Q4 videos from a first-frame image or reference images and optional reference audio through RunAPI. Use the RunAPI CLI for one-off generation and an SDK for application integration.

>- Routes a Skool writing request to the right post skill. Use when the user asks for a Skool post, a community post, or a scroll post and the type may be a welcome, YouTube video, bonus unlock, network thread, MVP list, or announcement.

Use when the user invokes /music-mv with a character image, song, LRC lyrics, source time range and a hand-painted 2D music-video style, including 小镇姑娘.

Wan3.0 (wan3.0-video) 短剧/剧情视频提示词专家。把剧本、分镜、场景描述改写成 wan3.0 可用的提示词，覆盖文生/首帧/首尾帧/全能参考(图·视频·音色)/视频编辑V2V/白膜转成片/文件网页解析全形态；落地七大公式（基础·进阶·图生·声音·参考生·多镜头·视频编辑）与 @图片1 指代口径，内置角色一致性、口型台词、人声六要素、情绪表演、多镜头时间戳、跨镜头空间站位、运镜设计、音效BGM 六维检查，以及时间线连续与段数卡尺、素材职责声明、复杂动作五要素、排除层3–8项取舍、参考片反推与 BadCase 诊断补丁。Use when 用户要写或优化 wan3.0/万相3.0/wan3.0-video 的短剧/剧情/影视类视频提示词，或提到 剧本转提示词、拆分镜、多镜头、一镜到底、角色一致性、口型对不上、台词截断、情绪特写、打戏、站位漂移、转场、改台词保音色、白膜转成片、剧集续拍、reference_voice 音色、参考片反推、照着这条视频拍一条、时间戳不连续、素材职责、短剧 badcase、wan3.0 短剧 case。

Edit a raw talking-head recording into a finished, ready-to-post 9:16 short (Instagram Reel, TikTok, YouTube Short) entirely in code. It covers syncing camera and clean audio, transcribing with word timestamps, picking the last good take of every line and cutting retakes, pauses and false starts with a J-cut voice track, and an eye-contact check. Each run gets one of three layouts (full cam, split with the head popping out of a camera card, full-frame "luxury tech" HyperFrames graphic with a moving world camera and motion blur). It plans the edit as a storyboard and beat table and waits for the user's approval before rendering anything. It burns in two caption styles (small pill captions, and big hero words in the speaker's clothing colour for the hook and key lines), adds a cover title and a CTA comment card, synthesizes SFX, a riser and a music bed in numpy, mixes to -14 LUFS, and writes a report. Use it whenever someone hands you raw footage (a .MP4/.MOV of themselves talking) and wants a short, reel, T...

Read a project's codebase, work out what the product is and which capabilities matter most, write a story script, then drive the real running app through LIVE flows (typing, answers arriving, tabs and navigation, agents being built) and cut a 25-40 second beat-synced launch video in the product's own theme, ending on its logo and category line, with post copy. Use when someone says "/flaunt", "flaunt this", "make a launch/demo/ship video of this app", "announce this release on video", or wants a marketing video made from their actual product. Works from a local dev server; can focus on what changed since a tag with --since.

Pick and apply one of 18 premium visual styles (fonts, palette, motion grammar, SFX kit) for the user's 9:16 HyperFrames videos — AI model evaluation (评测), tutorials (教程) and science explainers (科普). Use when making, designing or restyling a video or composition for this user, when choosing a 视频风格/画面风格/样式, or when they mention the style kit, 风格样张, or a style by name (数据新闻, 评测仪表, 暗场发布会, 终端控制台, 水墨留白…).

Make an animated explainer or product demo video for a software product (SaaS, web or mobile app) with Remotion — interviews the user, writes a scene-by-scene script for approval, then builds motion-graphics scenes that recreate the product's real screens, with original music and sound effects, a horizontal version and a vertical short cut. Use when the user asks for an explainer video, product demo video, promo/launch video, landing-page video, "vídeo explicativo", or a Reels/Shorts/TikTok video about their software.

Let Claude see the physical world through an iPhone (Continuity Camera) or the Mac camera, token-cheaply. Snapshots, before/after sheets, a watcher that stays silent until something changes, timed captures (5/10/15/30/45/60 min), 3D-print monitoring that finds the print and shoots when the print head is clear, build-step documentation, demo video clips, and timelapse videos. Use INSTEAD of asking the user for photos whenever a task depends on something physical - a print, a prototype, a fit test, a mechanism, a test rig, specimens, parts on the desk - and whenever they want progress pictures, assembly documentation, a demo video or a timelapse. If it isn't set up yet, run `setup` and walk the user through what it prints.

Edita un video crudo vertical de Gus Pascaner (talking head para redes, marketing médico) con corte ajustado sin pausas ni tomas falladas, motion graphics modernos con alfa, subtítulos dinámicos, b-roll semántico y música de Envato, diseño sonoro automático con su librería de SFX, reencuadres y zooms según el discurso, y una revisión de dirección creativa; entrega un XML para Premiere con todos los assets enlazados (sin render de vista previa) e informa cuánto tardó. Usar cuando el usuario escribe /edicion con un video o carpeta, o pide "editá este video como los anteriores".