Street Scan

Cinevva's analyzer read this game's design doc and its source, then wrote the brief below. "Make your own" hands it to the game creator as your opening message, so you start from a real spec rather than a one-line idea.

Recreate a mobile-first AR-style car collecting game titled "Street Scan". The primary visible experience is a full-screen phone-camera view: use an autoplaying, muted, inline rear-facing camera video when permission is available, filling the screen with object-fit: cover. If camera access is unavailable, show a dark blue-gray simulated camera background with a subtle radial gradient. Add a translucent dark vertical shade over the view and a centered pale cyan scanning reticle, about 180 by 130 pixels, with rounded corners, a bright outline, soft cyan glow, and a short center tick extending above it.

Build the HUD as fixed-position glassmorphism panels over the camera. At the top, place a dark navy translucent rounded card reading "Street Scan" in bold white text, with the subtitle "Find cars. Frame them. Catch them." Under/alongside it place two compact stat cards showing a large cyan number and uppercase label: "CAUGHT" and "STREAK", both initially 0. At the bottom center place a prominent pill-shaped cyan-to-blue gradient button labeled "📸 SCAN CAR" in dark bold lettering, with a pressed scale-down effect. Above it show the instruction: "Point your phone at a car and tap SCAN CAR." At lower right show another dark rounded card with "Collection: —" initially; replace the dash with collected car names separated by bullet dots. Use white, pale blue, cyan, and dark navy colors, thin translucent borders, soft shadows, responsive sizing for narrow mobile screens, and a clean Inter/system sans-serif visual style.

The interaction is deliberately simple: request the device camera on startup using getUserMedia with an ideal environment-facing camera. Pressing SCAN CAR also starts/request camera if it has not started. Each press scans the next uncaught car in a fixed collection order; it does not need actual computer vision or distance validation. The three targets, in order, are "Skyline Hatch", "Crimson Sport", then "Mint Sedan". On a successful scan, increment CAUGHT and STREAK, append the car name to the collection, hide/remove that target from the 3D street, and show a bright green toast near the top saying "MATCHED! [car name] added" for about 1.8 seconds. After all three are collected, show "Collection complete!" (and if the user scans again, show the same completion message).

Also create the underlying Three.js street scene, even though the camera/video overlay is the dominant visible layer: use a pale blue sky background and matching fog, hemisphere and directional sunlight, shadows, and a simple low-poly city street. Make a long dark charcoal road roughly 12 units wide, bordered by light gray sidewalks and a large green ground plane. Put short pale yellow dashed center road markings down the road. Along both sides place repeated simple block buildings: tan blocks on the left and cool light blue-gray blocks on the right. Add dark slim streetlight poles on the left with small warm yellow glowing lamp spheres. Make the road run along the z axis.

Place a glowing cyan player scanner marker on the road near z=15: a cyan emissive capsule-like body hovering above the ground with a thin bright cyan torus/ring around its base. Use a third-person camera/controls setup targeting this player, but keep the HTML canvas hidden behind the phone-camera UI as in the original presentation. Add three static target car assets/models to the street: a blue hatchback named Skyline Hatch around (-3.4, 0, -8), a red sports sedan named Crimson Sport around (3.5, 0, 1), and a mint-green sedan named Mint Sedan around (-3.5, 0, 10), with rotations approximately facing across the road. Cars should cast/receive shadows and carry caught/visible state. The scan collection and HUD state should be the key game rules; there are no enemies, health, timer, driving controls, weapons, or collision challenges.