FIELD GUIDE · 8 MIN READ

How to photograph your cards for Gemmr

A data-backed walkthrough of every variable that affects your scan quality. By the end you'll know exactly what separates a “rejected” scan from a “very high confidence” one.

Gemmr's grading model is only as good as what it can see. We've been watching the beta data closely, and one finding stands out hard enough to lead with:

64%
of every scan rejection in our beta is caused by glare. Nine times more common than any other rejection reason.

If you do nothing else from this guide, fix the glare on your scans. Everything below explains how — plus the smaller things that nudge a scan from “good” to “very high confidence.”

Kill the glare.

Glare is what happens when a light source bounces off a card's surface straight into your camera lens. Holos and full-art cards are the worst offenders because their reflective foiling is exactly designed to throw light back at the viewer. That's great in your binder. It's terrible for a grading model trying to assess surface and centering.

The setup that works

MATTE DARK SURFACECARDLIGHT~45°LIGHT~45°PHONE
Lighting setup, viewed from above. Two diffused sources at ~45°, on opposite sides of the card. Phone directly overhead, lens parallel to the card surface.

Two diffused light sources, both at roughly 45° angles to the card, on opposite sides. “Diffused” means the light passes through something soft before hitting the card — a piece of white paper, a thin curtain, a ring light's built-in diffuser, even a shopping bag in a pinch. Direct, hard light is what creates glare; diffused light fills the surface evenly.

What to avoid

  • Single overhead light. The most common cause of glare we see. The light bounces straight off the holo center back into the lens.
  • Direct sunlight from a window. Way too harsh. Move to indirect light or wait until the sun isn't hitting the surface you're shooting on.
  • Phone flash. Always off. It creates a hotspot in the dead center of every card.
  • Glossy desktops. The desk reflects light back up through the card edges. Even a dark glossy desk does this.

Pick the right surface.

What you place the card on matters almost as much as your lighting. The wrong surface can cause centering misreads, fake “edges” that Gemmr mistakes for damage, and color shifts in the holo.

USEBLACK FELTUSEDARK GRAY MATAVOIDWHITE POSTERAVOIDWOOD GRAIN
Surface choice has a real effect on scan quality. Dark and matte work; light and textured don't.

Use

  • Black felt (Amazon, craft store, cheap)
  • Dark gray microfiber (a phone cleaning cloth works for one card)
  • Non-reflective dark mousepad
  • Matte black or charcoal poster board

Avoid

  • White poster board. Causes overexposure and washes out card edges. The model can struggle to find where the card actually ends.
  • Wood grain or busy patterns. Texture from the surface bleeds into surface-condition analysis.
  • Anything glossy or reflective. Defeats your lighting work.
  • Bright colored fabric. Tints your holo's reflected light.

Frame the card properly.

After glare, the next biggest preflight rejection category in our data is partial cards (14% of rejections) and wrong angle (7%). Both are fixable in five seconds.

✓ CORRECT90°✗ TILTEDkeystonedistortionEven 10-15° of tilt visibly skews centering measurements.
Phone lens directly above and parallel to the card. Tilt creates keystone distortion that throws off centering analysis.

Angle

Phone lens directly above the card, parallel to it. Imagine a line straight down from your lens to the center of the card — it should be exactly perpendicular to the table. Even a small tilt (10-15°) introduces keystone distortion that throws off centering measurements significantly.

CARD✓ CORRECTCARD (cropped)✗ TOO TIGHT✗ TOO LOOSE
Frame the card so it fills the image with a small surface margin visible around all four sides.

Framing

The card should fill the frame edge-to-edge with a small margin of surface visible around all four sides. Not zoomed in so tight that edges are cropped. Not zoomed out so the card is a postage stamp in the middle of your desk.

Why the margin matters: Gemmr needs to see where the card ends to assess centering. If the card touches the frame edge, the model can't tell whether the border continues or got cropped.

Where users currently are.

Here's the confidence distribution of completed scans on Gemmr so far. Most users cluster in the “good” and “medium” ranges, with a smaller group reaching “high” or “very high.” The advice above is what separates those tiers.

0.90 – 1.00
3.7% (4)
0.80 – 0.89
2.8% (3)
0.70 – 0.79
42.2% (46)
0.60 – 0.69
45.9% (50)
0.50 – 0.59
5.5% (6)
Confidence distribution across completed scans on Gemmr. Most users currently land in “good” or “medium” — pushing into “high” is what better technique unlocks.

Confidence isn't the same as your predicted grade — it's how sure the model is about its prediction. Higher confidence scans give you a tighter range and a more reliable verdict. Roughly 88% of users are at “good” or “medium” right now. With the adjustments in this guide, “high” and “very high” are reachable.

Special case: holos, full-arts, and modern Pokémon.

The majority of cards we see are Pokémon, and a lot of them are holos or full-arts. These cards make glare control harder because the foil is doing exactly what foil is supposed to do: throw light in every direction.

  • Use even-more-diffused light. Two layers of diffusion (paper through a softbox, or paper plus a ring-light diffuser) instead of one.
  • Tilt the card very slightly — like 2-3° — to redirect the worst hotspot away from your lens. Re-level for the actual shot once you've found the angle that minimizes reflection.
  • Shoot in a darker room. Counterintuitive, but with less ambient light, your diffused sources have more control over what reaches the card. Bright rooms create reflections you can't see until you check the photo.

If your card is already in a slab.

Slab scanning is its own beast — the plastic adds an entirely new reflective surface between your camera and the card. Gemmr automatically detects slabs, reads the label for service and grade, and crops the card out so the prediction is independent of what the grader said. But to get clean data, you still want to:

  • Remove any case sleeves or top loaders before shooting. The extra layer adds reflections that Gemmr can't see through.
  • Light the slab the same way as a raw card — two diffused sources, 45°, dark surface. Plastic glare is just like holo glare.
  • Shoot front and back. The cert strip on PSA slabs lives on the back. Gemmr uses it for cert verification.

From our data:on the small sample of slab scans where we've been able to compare our prediction to the grader's actual verdict, PSA-slabbed cards have come back at an average difference of 0.00 grade points so far. The sample is small (n=2), but it's an early sign Gemmr's independent assessment is calibrated where it should be.

Phone settings worth changing.

iPhone

  • Turn off HDR auto — it can blow out highlights on holos. Settings → Camera → toggle HDR off (or to manual).
  • Turn off flash. Always.
  • Tap the card on your screen to lock focus and exposure. Then tap-and-hold to lock AE/AF.
  • Use the main camera (1×), not the ultrawide. The ultrawide introduces edge distortion.

Android

  • Disable scene optimizer / AI enhancement. These tools over-process card surfaces.
  • Disable flash.
  • Use Pro mode if available, with auto white balance and ISO 100-200.
  • Use the main lens (1×), not the wide.

What to expect after you submit.

Once you upload, the pipeline runs in four stages: a preflight check, slab detection (if applicable), card identification, and the grade. The full process averages around 26 seconds end-to-end. If you see your scan stuck on a stage for more than 90 seconds, something's wrong — head to the dashboard, it's probably already marked failed and your Gems refunded.

Quick FAQ.

Why did my scan say “poor lighting” when my room is bright?

Bright doesn't mean even. A bright room with a single overhead light is harder to shoot in than a darker room with two diffused lamps. The model is looking for even light across the card surface, not raw brightness.

Do I really need to shoot the back?

Yes, almost always. 97% of scans on Gemmr include a back image, and for good reason — corner and edge wear show up differently on each side, and surface damage is often only visible from one direction. A front-only scan can still complete, but the confidence will be lower because the model is making decisions on half the data.

What if I'm shooting outdoors?

Find shade. Direct outdoor sun is the worst lighting you can use — it's an undiffused point source so bright it overwhelms any setup. Open shade (a porch, the shadow side of a building) on an overcast day is actually some of the best lighting there is. The clouds are doing the diffusion for you.

Will a DSLR or mirrorless beat my phone?

Marginally, if you know what you're doing. A modern phone with good lighting beats a fancy camera with bad lighting every time. If you have a real camera, use it with a macro or 50mm lens and a tripod — but optimize lighting first.

Can I scan cards through a top loader or penny sleeve?

You can, but you shouldn't. The plastic adds reflections and a subtle blur that both hurt confidence. Take the card out, scan it, put it back. Total extra time: about 8 seconds.

My scan got rejected. Did I lose Gems?

No. Any scan rejected at preflight (the first stage) refunds your Gems automatically. The same is true if the scan fails partway through for any reason that wasn't a deliberate choice on our end. You only ever pay for completed scans.

Ready to scan?

Set up two soft lights, grab a black cloth, and shoot directly above. That's 90% of it.

Sign up free →

Have questions or a tip we should add? Drop into Discord and let us know.