Seventeen animated caption looks for talking heads, podcasts, ads, lyric and short-form videos (karaoke fill, active-word pop and colour, sliding pill, word tiles, frosted glass, neon, serif italic, white card, bounce and tilted word cards and more), each with presets in exact numbers: font (OFL), size per format, words per screen, active-word timing 1-2 frames before the voice, 9:16 safe box, contrast, padding and easing. The rules come from the Netflix and BBC subtitle guides, the DCMP Captioning Key, ITU sync tolerances, eye-tracking research and short-form editors. Use when someone asks for TikTok, Reels or Shorts captions, word-by-word or highlighted subtitles, karaoke captions, or 'captions like Hormozi / MrBeast / podcasts'. NOT for transcription itself (use WhisperX or ElevenLabs speech-to-text first), translation, or plain SRT delivery without styling.

Make a 3D explainer video of something you cannot film (a cell, an atom, a virus, a protein, an engine, a reactor, a black hole, a maths idea) entirely in code: a sourced fact sheet checked by a fresh critic subagent before any 3D work, the subject built in three.js from procedural geometry (and real data such as PDB coordinates) one part per beat with a label and a one-line fact, a camera that tours the build and zooms through scales to an x-ray cutaway of the whole, a live scale bar, deterministic 1080p rendering with headless Chrome (scripts/capture.ts) or HyperFrames, and QA scripts that check label overlap, phone legibility, reading time and that every word on screen is in the approved fact sheet. Includes a working 21 s animal-cell film to copy. Use when someone asks for a 3D explainer, a science explainer video, "explain how X works in 3D", a cell / atom / virus / molecule / engine / reactor / black hole animation, "build it one part at a time", a cutaway or exploded view explained, a Kurzgesagt-style 3D explainer, a "zoom into the invisible" film, or a Fable 5.2 code explainer. Not for 2D maths and equation animations (manim-skill), short 2D motion formats (motion-formats), general HTML-to-video work (hyperframes), writing the script itself (video-script) or sound design and music (video-sound).

Turn a browser game or three.js / WebGL scene an agent built into a cinematic trailer MP4: a virtual clock injected before the page's scripts makes every frame exact at 30 or 60 fps, a director JSON scripts each shot (key and mouse inputs, game setup, trailer camera paths, slow motion, HTML titles and end cards), headless Chrome captures 1920x1080 frames, ffmpeg cuts them on the beat, a music bed (code-written or from varg) is normalised to -14 LUFS, and a checker measures frozen frames, loudness, debug text, third-party names and claims. Includes a tested example game. Use when someone asks for a game trailer, to record my three.js game, a gameplay video, a trailer for my browser game, a Fable 5.2 game video or a fake AAA gameplay trailer. Not for HTML motion graphics (hyperframes), 3D previz for AI video (threejs-animatic), AI-generated footage (varg-ai) or editing existing footage (short-form-ffmpeg-cookbook).

Make a short animated film in which everything is code the agent writes, including the renderer: a path tracer in one WebGL2 shader (no Blender, no three.js, no meshes, no textures), characters and props as signed distance fields, a meadow of analytic curved grass blades, moon, firefly and window lights with soft shadows, depth of field and motion blur, a variance-guided denoiser, a score synthesized in numpy from the same timeline, captions and an encode. Includes the complete tested film "The Smallest Star" (70 s, 11 shots, two characters) as a starter, a live page that runs the same renderer in the browser, and the speed-ups and traps found while making it. Use when someone asks for a film, short or music piece "made entirely in code", "write your own renderer / ray tracer / path tracer and make a film with it", a storybook or clay-diorama 3D short without Blender, a "no assets, no video model" showcase of what a coding model can do, or a remake of The Smallest Star with another story. Not for a Blender film (blender-code-film), a three.js explainer (code-3d-explainer), recording a browser game (code-game-trailer), 2D motion graphics (kinetic-type-art, hyperframes) or AI-generated footage (varg-ai).

Prototype a video in code with an agent. Decides if a shot needs a 3D prototype at all and which tool fits (Three.js, Blender, a phone act-out, storyboards, Remotion, Cinema 4D, Unreal), stages geometry, people, camera and light in real metres, copies a reference video shot by shot with a solved 3D camera, stages a shot from a painting or still, invents one when there is no reference, checks everything with numbers and frames, and writes the handoff to Seedance or other video models. Same scene file renders in Three.js and headless Blender. Use for "previz", "blockout", "3D animatic", "three.js or blender", "prototype this video", "copy this shot", "recreate this video in 3D", "stage this painting", "camera blocking with Claude or Codex", "code to video", "rank takes", "LLM judge", or when generated shots keep drifting. Not for a full multi-shot three.js film with a review UI (threejs-animatic, varg-previz), prompting a shot with no prototype (varg-video-prompting) or the full animation pass (video-animate).

Turn raw product footage into a TikTok ad with FFmpeg. Recuts creator, UGC or product footage into a 9:16 ecommerce ad for TikTok, TikTok Shop, Reels and Meta: the best moment moved to second 0, a text hook inside the safe zone, dead air cut, captions, a CTA end card, loudness at -14 LUFS, and one MP4 per hook for testing (plus a 4:5 feed cut and a cover frame). Every command was run on test media. Free and local; on the varg ffmpeg tool (credits) when the agent has no shell. Use when someone has footage of a product (a creator video, an unboxing, a demo, a winning ad) and wants it edited into an ad, wants new hooks on an existing ad, or wants it TikTok-ready. NOT for generating new footage (use product-photo-to-video or ai-ugc-ad-video), animated word-by-word captions (use a Remotion or HyperFrames caption skill), or generic crops and subtitles (use short-form-ffmpeg-cookbook).

Find real, human-made references before anything is made, for AI image, video and music and for video made with code. Turns a request into one query pack per job (light first, then framing, people, location, texture, type, motion, palette, music and a "never" board), searches each source in its own words, opens everything it keeps, proves who made it, measures what code can measure, checks light with an LLM judge, and hands a person three candidates per job. Use when someone says "find references", "reference board", "what should this look like", "lighting reference", "framing reference", "texture reference", "font reference", "motion reference", "casting reference", "music reference", "moodboard references", "make it not look like AI", "avoid slop", "rank these references", or starts a video with no look yet. Not for writing the look from chosen references (video-moodboard, varg-video-prompting), building the cast (character-passport) or downloading one known link (media-download).

Guide to OpenAI's GPT Image 2.6 on varg (gpt_image_2_6) and the entry point to the GPT Image 2.6 skill set: Flare vs Sunburst, quality vs reasoning_effort, sizes and aspect presets, what fal rejects (xhigh and max effort, SVG, over 16 GIF frames), measured prices, failure modes, the exact varg MCP and REST calls, and prompt recipes for the text-heavy layouts people shared on X: infographic, product launch poster, photo to editorial travel print, photo to Baroque gallery poster, landmark postcard, character turnaround sheet, storyboard sheet and low-res to clean upscale, each credited to its author. Use when someone asks for GPT Image 2.6, gpt-image-2.6, Flare or Sunburst, wants an infographic, poster, art print, character sheet or storyboard with readable text, or asks how to prompt, size or price this model. Not for looping GIFs (gpt-image-gif), sticker packs (gpt-image-sticker-pack), game sprite sheets (gpt-image-sprite-sheet), circle-and-arrow edits (gpt-image-markup-edit) or video (varg-ai).

How to make AI video skills for Claude Code and AI agents: one search over varg how to (varg.ai/how-to, 2,000+ checked AI video skills, step-by-step how-to tutorials, FFmpeg recipes, video formats, prompts and styles) turns a video request into the best skill with its install command, the 2-4 options to show the person (with example videos) and the steps. Use when someone asks "how do I make a ... video", "make me a UGC ad / lyric video / explainer / talking avatar / product video / faceless short", "make a whole film / launch video from scratch" (production order across skills), "add captions", "edit this video", "which skill should I use for ...", "find a skill for ...", or the same in any language, and no installed skill already covers that format; use it before writing a video pipeline from scratch or searching registries. Also use it to list a video skill you built on varg how to. Not for running a generation itself (use varg-ai), and not for non-media coding tasks.

Seventeen art-directed kinetic typography templates with exact presets: typeface strobe (one word through nine faces), one-word-per-beat slams, word corridor tunnel, perspective photo wall, type specimen grid, justified manifesto, Didone city title, scanned-sheet glitch wordmark, hyphen bridge wordmark, CRT decrypt, particle burst, anime command screen, photocopy explainer, echo stack over a halftone portrait, glyph-tracking terminal camera, bitten word and liquid art reveal. Presets cover beat maths (frames per beat), letter-spacing, free font pairings for Didone, condensed, grotesk and mono looks, riso and paper palettes in hex, 3D camera values and seeded glitch timing, from Art of the Title, Practical Typography, Fonts In Use and motion designers on X. Use when someone wants a kinetic type intro, a manifesto or lyric-style type piece, a wordmark animation or a typographic poster in motion. NOT for subtitles (caption-styles) or broadcast lower thirds (title-design).

Turn three or four key frames into a 30-second action montage with Kling 4 (kling_v4): a breakdance battle, BMX or skate tricks, a scooter or bike chase through traffic, parkour or a sports highlight, with a crash or punchline at the end. A one-line prompt over strong stills, "no slow motion", the trick names, the end state as the last still, and how to pick the usable seconds. Includes a tested varg call with the reference route pinned. Use when someone wants an AI action montage, a dance or trick sequence, an extreme sports clip, a chase, a "fail" ending, or fast-cut energy in one Kling 4 generation. For a full story use kling-4-keyframe-film; for prompt rules use kling-4-prompting.

Turn one selfie into a satirical "I MADE ME WOKE" profile set with GPT Image 2.6, the image model in ChatGPT (aka Wokify): 11 woke versions of you (character sheet, woke friends photo, LinkedIn-style profile, DEI video call, protest TV interview, action figure, holographic trading card, woke cat, oat latte, "In this house" home, WOKEOPOLY), four as looping GIFs, in a 4:5 collage sized for X plus an MP4, and an optional HyperFrames presenter-mode video. One command on varg, about $0.78; the prompts also work pasted into ChatGPT. Use when someone wants to make themselves woke, wokify a photo or profile picture, a woke AI filter, the "I made me woke" trend, a ChatGPT image prompt pack for a selfie, an AI action figure or trading card of themselves, or a collage of AI versions of one person. Not for photos of others without consent, public figures, or mocking LGBTQ+ people themselves (the joke is on performative habits and brands).

Eight map and data animations with exact presets: globe-to-country dive, satellite flyover between two places, vector route map, watercolor travel route, self-drawing bar and line chart, count-up statistic, versus timeline (two numbers diverging over the years) and an audio visualizer (silk ribbons, wireframe terrain, ASCII stream). Presets give camera paths on the van Wijk zoom-pan curve, flight durations, line widths, chart rules (bars from zero, one accent series, direct labels), number formats and count-up easing, and FFT sizes, log bands and attack/release for audio, from Heer & Robertson's animation study, MapLibre and Mapbox, Datawrapper, John Burn-Murdoch (FT), the W3C Web Audio spec and MilkDrop. Map credits and tile licences are part of every preset. Use when a video needs an animated travel map, a route between cities, a flyover, an animated chart or statistic, or a music visualizer. NOT for interactive web maps or dashboards.

Instead of describing where a change goes, you draw it. GPT Image 2.6 takes a second, transparent PNG with only your marks on it, and the prompt says what each mark means. OpenAI: "You can now send a marked-up image to the API to show exactly where you want an edit" ([@OpenAIDevs](https://x.com/Open

There are two scripts in `scripts/` next to this file. Each prints only the paths of the files it saved on stdout. Messages and errors go to stderr. Neither shows a progress bar, so the output stays short. Run long downloads with a generous timeout or in the background.

Seven complete motion formats, each with a beat structure in seconds, a shot list and presets: a found-footage analog-horror trailer of a website, an authentic VHS bouncing-logo screensaver, a pixel-art runner on a CRT, a wordmark with a wave running through it, a one-shot music video that follows a DAW timeline, a launch trailer for an import feature, and a kinetic lyric video. Values come from VHS signal models (ntsc-rs, Finn McKenty), pixel-art conventions (Pedro Medeiros, Slynyrd), School of Motion, Netflix timing rules, screenlife rules and launch-film designers on X. Use when someone asks for one of these formats, a VHS or analog-horror look, a pixel art animation, a lyric video or a beat-synced launch video. NOT for generating the song or footage (varg-ai, varg-music-video).

The hook is the first 5-15 seconds of a product reel: real-looking footage of the product doing the one thing that makes people stop. In the dropshipping reels this skill was built from (@tradelle, Sept-Oct 2026), the hook carried the video: a snake biting a leg gaiter, a worker getting sand and gri

Every model launch fills X with the same format: one prompt, two models, two results side by side, and the bill under each. On Claude Fable 5.2's launch day (2026-11-09) it was the fifth biggest format, with 824k views across 9 posts. Cost and minutes have become part of the caption:

Input: a video (any size), optionally an SRT file and a music track. Output: a 1080×1920 9:16 MP4 ready for Reels, TikTok and Shorts. Probe first, run one recipe at a time, check the result with ffprobe and a frame grab before the next step.

Eleven conversion overlays for short-form ads, UGC and creator videos, with presets per platform: CTA pill, follow tap prompt, story sticker tag, LIVE badge with viewer count, notification drop, review proof card, story caption bar, photo wall loop, follower counter and goal bar, comment callout (reply-to-comment) and subscribe-and-bell nudge. Presets use one universal 9:16 safe box built from TikTok, Meta Reels, YouTube Shorts and Snap specs, CTA timing from creator analysts (ask after value, never in the intro), native number formats, contrast and motion tokens, plus the FTC rules: reviews, comments and counts come from real data only and paid content is disclosed. Use when an ad or creator video needs a call to action, a subscribe or follow ask, a review or comment on screen, a LIVE badge or a counter. NOT for fake reviews, fake follower counts or copying a platform's logo.