Walkthrough
How GradeX works
From opening the camera to listing a graded card on the marketplace, the whole flow takes about three minutes — most of which is the user reviewing what the AI found. Here's exactly what happens at each step.
Step 1
Capture front and back
Open the scanner. The viewfinder shows a card-shaped guide, a rule-of-thirds grid, and a real-time level indicator (driven by your phone's orientation sensor) that turns green when you're within 2° of perfectly level. Frame the card with a couple of inches of background visible, tap the shutter, and repeat for the back.
Step 2
Capture corners and edges
The flow walks you through close-ups of all four corners and four edges. Auto-tiles cropped from the rectified front are also generated server-side so the AI sees pixel-native evidence even if your close-ups are imperfect.
Step 3
Set the centering
Before submitting, an 8-line wizard appears for the front and back. Four red lines snap to the card's outer edges; four green lines mark the inner design border. A leveling slider compensates for any residual tilt. The AI uses your numbers exactly — centering is no longer a guess.
Step 4
Mark anything obvious (optional)
If you've already spotted a scratch, fingerprint, or chip, you can draw a box on it before analysis. The AI factors your annotations in and the calculator can't score below them.
Step 5
Wait for the analysis
The grading job runs as a background task. Even if your phone screen turns off, the job keeps going server-side and you'll be notified when it's done — typically under a minute. The status indicator shows: Compressing → Sending to AI → Identifying → Centering → Corners → Edges → Surface → Calculating.
Step 6
Review the result
You'll see predicted grades for PSA, BGS (with subgrades), CGC, and TAG, plus a defect map on the front and back. Tap any defect to highlight it on the card; tap Reject to flag any AI mistake — your rejections are sent back into the next reanalysis as explicit "do not include" instructions.
Step 7
Save, sell, or trade
Save the card to your vault to track purchase price, target sell price, and price history. Verified users can list directly from the vault to the marketplace — sale, trade, or both — with the 2% fee transparently broken out at listing time.
What if something goes wrong?
The pipeline is wrapped in step-tagged error handling so any failure surfaces with a clear name (rectify / compress / submit-job / server-grade) plus the underlying cause. Image rectification falls back to the original photo if the homography warp can't lock onto the card. Defect specialist failures are non-fatal — the primary grading pass still returns. If a job times out (rare; the 5-minute server budget is generous), the result page surfaces a retry CTA.