gooseworks-ai/render-airdrop-carousel
Assemble a viral iOS "AirDrop" notification-carousel video ad (≈6–8s, 9:16) from a brand line plus 6–16 real product photos — a native AirDrop share-sheet card ("Brand would like to share a ___ · Decline / Accept") springs up and its preview window CYCLES through the products, landing on a range/lineup payoff with an Accept tap; a chime + soft per-swap ticks track the swaps. DETERMINISTIC assembly — an HTML card (real DOM text) rendered to PNG via headless Chrome, chroma-keyed, its magenta window refilled per-product in PIL, then animated + audio-synthed with FFmpeg. FREE (no paid model calls); the recipe supplies the brand line, product images, and payoff and gates the only optional paid step (a hero shot when the brand has NO usable photo → create-image-fal). Use for the airdrop-notification-carousel format.
npx skills add https://github.com/gooseworks-ai/goose-skills --skill render-airdrop-carousel
The free, deterministic renderer for the airdrop-notification-carousel video ad
format — the viral iOS "AirDrop" trend where a native share-sheet card
("Brand would like to share a *candle* · Decline / Accept") springs up on the
phone and its preview window flips through a carousel of real product photos,
landing on a range/lineup payoff with an Accept tap.
This is a DETERMINISTIC composite — no generative video, no AI-rendered product.
A real-DOM AirDrop card is rendered once to PNG (headless Chrome), chroma-keyed, and
its preview window is refilled per-product in PIL, then animated + audio-synthed with
FFmpeg. The whole point of the format is crisp system-UI text and real product
photography, both of which a video model would smear. This capability makes **no paid
calls**; the recipe gates the only optional paid step — generating a hero product shot
when the brand has NO usable photo at all (→ create-image-fal).
Default output ≈ 6–8s, 1080×1920, h264 + aac. Duration is DERIVED, not trimmed —
first_hold + (N-1)·per + final_hold. Add images or raise per to lengthen.
scripts/build_card.py — brand params → chrome.html + chrome-pressed.html: theAirDrop card on a green page (#00e000) with a magenta preview window
(#ff00ff) and a solid brand band (wordmark SVG or text + tagline). Real DOM text —
AirDrop, Decline, Accept, and the brand line are DOM/SVG, never AI-rendered.
scripts/one_shot.py — glue: build_card → headless-Chrome (Playwright) fullPagescreenshot of both card states → compose_carousel. One --config, one MP4.
scripts/compose_carousel.py — the render engine: green-key the card → detect themagenta window → fill it per-product (cover-crop) → blurred per-product backdrop +
push-in → card spring-up (iOS ease-out-back) + carousel + Accept tap → synth audio
(whoosh on entry, chime on land, a tick per swap, a pop on the tap) → encode h264+aac.
scripts/config.example.json — the shape of the brand config the recipe binds(brand-neutral worked defaults; replace every /abs/path/... placeholder).
#00e000 green = page background, keyed to the card's alpha. #ff00ff magenta
= preview window, replaced per product. **Neither color may appear in the card art or
any product photo** — a product image containing near-pure green or magenta gets
keyed/misread. Swap the image or tighten the window_mask thresholds if it bleeds.
AirDropped items — no AI-generated products in the carousel.
they stay pixel-crisp. Cover-crop the window; white-bg PDP shots reading as "product
in a white tile" is on-brand and fine.
final_image) shouldread as "the whole line" (lineup / family / all-shades).
1M+ sold tagline must be the brand's OWN stated figure,not filled from a brand-kit proof section.
per; never cut the audio.numpy + Pillow, and ffmpeg/ffprobe on PATH.install chromium). If Playwright is unavailable, run build_card.py`, screenshot
chrome.html / chrome-pressed.html (fullPage) → chrome-green.png /
chrome-green-pressed.png via the chrome-devtools MCP, then call compose_carousel.py
directly.
watch (QC the final master). The recipe gates create-image-fal (optional hero shotwhen the brand has NO usable product photo) — the only paid, proxy-routed step.
Take gooseworks-ai/render-airdrop-carousel from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.
The instructions reference pip.
Without those the skill loads but fails at the first command.