The free core is a full plugin in its own right: it hooks WooCommerce order completion, renders the reveal on the confirmation page, and holds back the confirmation email until every ticket has been played — otherwise the inbox spoils the result before the customer even lands on the page.

On top sit eight premium game modules, each an HTML5 canvas engine licensed separately at £9.99 a month. Every module has its own admin settings page with a live device-emulating preview, global defaults with per-product overrides and a 'Use Global' fallback, and theme files — one JSON that re-skins a game from Galaxy Grab to Candy Catcher, importable at either level. Artwork is uploadable down to the trackside advertising boards inside the race track.

The commercial side is mine too. A licence server validates subscriptions and serves every game engine remotely — nothing ships in the zip, so a stripped licence check leaves a plugin with no games in it. It builds white-label cores on demand for agencies, drives in-dashboard auto-updates, and stages every release to the demo sites first: customers only see a version once it's been played through and published from the vault.

There is no national dataset of UK lay-bys or lorry parking — I checked data.gov.uk, the DfT lorry-parking surveys and the devolved studies, and every one is aggregate only. So the dataset is built from OpenStreetMap: Geofabrik extracts for Great Britain and Northern Ireland processed locally in Python, filtered to features with an explicit HGV, lay-by or rest-area signal, deduplicated, then enriched with nearby roads, rail, land use and administrative geography. 4,581 stops ship in the app, alongside 1,847 low bridges and weight limits, 2,514 forecourt shops a lorry can pull in to, and every service area with its showers, fuel brand and overnight price scraped from the operators' own pages.

Honesty is the product. Nothing in OSM says a site permits overnight parking, so every location carries a classification that says exactly how much is known, and 'nobody has said' is shown as exactly that. Drivers fill the gaps: one-tap I've Been reports (parked OK, no room, moved on, not free), spaces left, photos of the way in, artic-OK and canopy-height reports on forecourts, and spots the map never had — all shared through a PHP/MySQL service with a trust dial from desk research to well used. Position is used once to sort by distance and never leaves the phone.

The planning kit is the Pro tier at £1.99 a month: type two towns and every stop within two or five miles of the corridor is listed in the order you reach it with an arrival time, the driving-hours ring fades out anything you can't legally get to, offline map regions, fuel forecourts filtered to the brand your card runs on, and cab mode that dims the whole screen for a lay-by at 11pm. The map, the sites and reporting stay free for everyone — the shared data is the product, and nobody should be priced out of contributing to it.

Every card ever printed ships inside the app as a binary, so identification is instant and works with no signal at a fair or in the car. Reading is ML Kit OCR, but OCR of a card is a mess: attack names read like card names, collector numbers arrive as 'GOBno35/ 197', promos print a different set code than the data stores. The parser picks the name by layout — the tallest line above the collector number — matches word by word, sanity-checks the number against the printed set size, and needs two agreeing reads before it auto-files. When text fails on foil glare or a full-art name, a difference hash of the illustration finds the card by its picture instead.

Selling is the half that makes it worth having. Link an eBay account through my own OAuth server — the app never holds the token — tick fifty cards and it writes every listing: title, set, number, rarity, finish, condition, photo and each item specific eBay demands for trading cards, priced off today's market with a floor you set. Sold cards leave the binder on their own. For the marketplaces with no usable API it writes the file their own tooling takes: Cardmarket's bulk-import CSV, TCGplayer, Whatnot. Nothing ever lists without the seller seeing the price first.

Around that: a grade checker that straightens the card with a homography and measures centring as geometry — reported as a ceiling, with corners and edges as whitening and an explicit refusal to guess at surface; pack sessions that count every pull against what the box cost; sets as checklists; price history the app records itself; and Google sign-in backup to my own server, which holds exactly three things — who you are, your binder, and the last few versions of it. Pro sells monthly, yearly or once for good, and scanning is never gated.

Every candidate window is snapped to sentence boundaries and scored on five axes — hook, emotion, payoff, clarity and whether a stranger could follow it cold — then overlapping cuts are suppressed so you get distinct moments, not ten versions of one. Claude re-ranks the shortlist and writes the titles and captions; without a key the local scorer carries everything. Rendering is ffmpeg with ASS captions burned in one word per dialogue event, five caption presets, three framing modes including a speaker-follow crop, and ten-minute 16:9 compilations with cold open, chapter cards and a thumbnail, rendered in about two minutes on NVENC.

The Whop side is what makes it a business tool. Whop Content Rewards pays per thousand views for clipping footage a brand has licensed, under rules the brand sets and a submission window of thirty minutes after a clip goes live. ClippD connects to your Whop account, searches the marketplace from inside the app, and a marketplace autopilot scans every six hours: it refreshes each campaign's remaining budget and burn rate, pauses the ones about to run dry, adds the best new listings and starts cutting their footage unattended. Each brief's rules — must-say keywords, banned topics, length limits, logo overlay, required tags and mentions, posts per account — travel with every clip, so nothing gets rejected for a rule somebody forgot.

Posting is scheduled to each account's own slots with jitter, caps and minimum gaps. The moment a campaign clip goes live, a headless browser drives Whop's submission form inside the window, reads the verdict back, and a sync mirrors Whop's review status, view counts and payouts onto the post — so the Growth page shows what each clip actually earned, not an estimate. A rights gate blocks any clip without a recorded basis for reuse and appends attribution at commit. ClippD runs locally on your own machine — your accounts, your browser profile, your footage never leave it. I licence copies; get in touch below.