1. Home
  2. Our work
  3. Card scanner
Case study ยท Mobile app

Card scanner & score calculator

A private tool for the host of a live trading-card game show. Point the phone at the table: the camera reads the cards, adds up the hand and shows instantly whether it qualifies, while the show keeps moving.

Mobile web app Android On-device OCR Web Worker
The problem

Mental math, live on camera.

Our client is a live-stream seller who runs a trading-card game show. In every round, cards go down on the table and the host has to read each card's stats, add up the hand and call whether it qualifies.

Doing that by eye while talking to viewers is slow, and a wrong total in front of a live audience is awkward. The brief: a phone tool that reads the cards and does the sum, fast enough to use mid-show.

  • ๐Ÿ‘€
    Small print, bright lightsCard stats are tiny and hard to read quickly on a table under studio lights.
  • โž•
    Adding up under pressureHP values plus boosters, every round, while keeping the conversation going.
  • โฑ๏ธ
    No time to double-checkThe audience wants the result now, and it has to be right.
What we built

Point, scan, call the result.

A mobile web app, also packaged for Android, used privately by the host during the show.

Camera reads the table

The phone camera captures the cards laid out on the table, straight from the browser or the Android app.

On-device OCR

Text recognition runs on the phone itself, in a Web Worker so the screen stays smooth. About one second per scan on a phone or an M1 Mac.

Card stats detected

The app picks out each card's stats from the recognised text, such as HP and boosters.

Instant hand total

Stats are added up automatically, so the host reads one number instead of five.

Qualifies or not, at a glance

A big green or red result shows immediately whether the hand meets the rule.

History of recent hands

The last hands stay on screen, handy when a viewer asks about an earlier round.

A closer look

One scan, one answer.

The phone shows the camera, the total and the verdict. The detail view shows what was read from each card.

Illustration with demo data

How it works

From table to verdict in about a second.

Everything happens on the device. No photos are uploaded, so it works the same on any connection.

Capture

The host lays the cards down and taps scan. The app grabs a sharp frame from the live camera.

Read

A Web Worker runs the OCR model on the device, so the interface never freezes while it reads.

Detect

The recognised text is turned into card stats: HP for each card, plus any boosters.

Decide

The total is checked against the rule and shown big and colour-coded, then saved to the history.

Stack
Mobile web appHTML, CSS and JavaScript, built for a phone screen first.
Camera APILive camera preview and frame capture in the browser.
PaddleOCR + ONNX RuntimeOn-device text recognition running in WebAssembly.
Web WorkerKeeps recognition off the main thread so the UI stays responsive.
Android app (Kotlin)The same scanner packaged for Android phones.
Rules in codeStat parsing and qualifying rules that are easy to adjust.

Built as a private tool for the host; it is not a public app.

Custom tools

Need a custom tool for your business?

Tell us what slows you down. We build small, practical tools that fit the way you already work.