← Back to Blog

    How a Photograph of a Painted Card Becomes a Finished Deck

    Behind every printed Orim card is a pipeline that finds the card's edges from a photograph, then fits its name onto the card automatically — at any size, from any scan.

    [IMAGE PLACEHOLDER: hero — a raw card photograph next to its finished, name-stamped counterpart]

    Every card in an Orim deck starts as a photograph of the original painted artwork. Getting from that photograph to a finished, printable card face is its own small piece of engineering — one built to survive photographs that are never perfectly consistent.

    Finding the card, not assuming where it is

    A naive version of this problem just crops a fixed rectangle out of every photo. Orim's pipeline doesn't — because a stack of physical card scans is never perfectly aligned, and a fixed crop on a slightly shifted photo cuts into the art.

    Instead, the tool scans each photograph for the card's actual edges: it looks for a sustained band of bright pixels, the card's white border against whatever surface it was photographed on, and crops to what it finds rather than what it assumes.

    [IMAGE PLACEHOLDER: a raw scan with the detected border highlighted]

    Fitting a name onto the card, automatically

    Once the card is cropped and its corners rounded, its English name gets stamped onto a detected zone at the bottom of the card. But names aren't all the same length — some are one short word, others need more room — so the type size isn't fixed. The tool tries progressively smaller sizes until the name fits its space cleanly, and if a name is still too long even at a small, legible size, it breaks across two lines at the most natural word boundary automatically.

    [IMAGE PLACEHOLDER: two finished cards side by side — one short name, one that wrapped to two lines]

    When a card also carries three keywords underneath its name, that same bottom zone splits itself in two — most of the space for the name, a smaller strip below it for the keywords — sized proportionally, so the layout holds together whether the source scan is small or huge.

    [IMAGE PLACEHOLDER: a finished card showing both the name and its keyword line, zones visible]

    A small detail that closes a loop

    The keyword text is printed in plain black, with no extra letter-spacing — a deliberate choice. Orim's own AR scanner reads printed keywords back off physical cards later, and plain black text with no manipulated spacing is what that scanner reads most reliably. The way a card gets printed was tuned against the exact camera that will eventually photograph it again.

    The quiet part of a deck

    None of this shows up on the finished card — no one holding the deck sees the edge-detection or the auto-sizing math. What they see is 90 cards that all feel consistent, whether the name is one word or four, whether the original scan was perfectly lit or not.


    Shot list:

    1. Hero: raw photo next to finished stamped card
    2. A raw scan with detected border highlighted
    3. Two finished cards — one single-line name, one two-line wrap
    4. A finished card showing name + keyword zones together