Photo to 3MF: What File Formats and Lighting Actually Work for Gridfinity Trays

The Most-Asked Question on Day One: "What Photo Should I Use?"

Every week, we get the same message from new GridPilot users: "Do you have sample pics I can use?" or "What kind of photo actually works?" It's a fair question — most AI tools are picky about input, and nobody wants to take three runs at a tray before getting a good result.

This post answers it directly. Here's what file formats GridPilot accepts, what lighting and angles produce the best Gridfinity layouts, and the specific mistakes that cause the AI to misread your scene. If you bookmark one page about photo prep for GridPilot, this is it.

Supported File Formats

GridPilot accepts the formats every modern phone and camera produces:

• JPG / JPEG — the standard for almost every Android phone and most cameras. Works perfectly. No conversion needed.
• PNG — common from screenshots and laptop cameras. Works perfectly, including transparent backgrounds.
• HEIC / HEIF — the default format for iPhones since iOS 11. GridPilot converts these server-side, so you can upload them directly without changing settings on your phone.
• WebP — supported, mostly useful if you grabbed an image from a web page.

The maximum file size is 25 MB, which is roughly a 50-megapixel photo at full quality. Almost no phone produces a file that large, so you're unlikely to hit the limit.

Formats that don't work and need converting first: TIFF, RAW (CR2, NEF, ARW, etc.), PSD, GIF, BMP. If you only have a RAW file, export a JPG from your photo app first.

The Best Photo for a Gridfinity Tray

The single biggest factor in layout quality is the photo itself. The AI is good — but it's only as good as the input. Here's what produces a clean tray on the first try.

Camera Angle: Top-Down, Centered

Hold your phone directly above the items, lens parallel to the surface. The closer you can get to a true 90-degree top-down shot, the more accurately each pocket maps to the actual item shape. A tilted angle stretches items in perspective, and the AI compensates, but you lose precision.

If you struggle to hold the phone level, use a chair, ladder, or shelf to brace your arm. Most phones also have a built-in level indicator in the camera app — turn it on and align the bubbles before you tap the shutter.

Lighting: Even, Diffuse, No Hard Shadows

Hard side-shadows are the number one cause of bad detections. The AI sees a shadow as part of the item and draws the pocket too large. To avoid this:

• Best: overcast daylight near a window, or under a softbox if you have one
• Good: a bright overhead light directly above your layout (kitchen ceiling light works fine)
• Avoid: single side-lamp, late-afternoon window light coming from one direction, the flashlight on the back of your phone

If you can see distinct shadow stripes next to your tools in the preview on your camera screen, move the light or wait for better conditions. Even, low-contrast lighting always beats dramatic lighting for this job.

Background: Contrast With Your Items

The AI separates items from the background by detecting edges. Pick a surface that contrasts with what you're laying down:

• Dark tools (wrenches, sockets, pliers) → light surface (white paper, light cardboard, white workbench mat)
• Light items (white plastic parts, light-colored components) → dark surface (black mat, dark wood, black cardstock)
• Mixed colors → medium-gray neutral background works best

Avoid backgrounds with patterns, wood grain, or text. They confuse edge detection. A piece of plain matte poster board from any craft store costs a couple of dollars and pays back instantly in better results.

Item Spacing: Leave Room to Breathe

Items touching each other often get merged into one pocket by the AI. The fix is simple: leave 5–10mm of space between items when you arrange them. This is also how they'll sit in the final tray — pockets need a thin wall between them anyway.

If you want two items in the same pocket (say, two small bits that share a slot), put them touching on purpose. The AI will treat them as one shape, which is what you want.

Resolution: Higher Helps, But You Don't Need a DSLR

A modern phone camera (12 megapixel or better) is more than enough. We've seen great results from photos as low as 1500x2000 pixels. Don't bother with a "high quality" mode that produces 100MB files — the AI downsamples anyway.

Specific Photo Mistakes to Avoid

If you've tried a photo and the result looks off, check this list:

• Tilted angle — items appear stretched in the preview
• Single side-light shadow — pockets are oversized on one side
• Items touching — neighboring items get merged into one big pocket
• Low contrast background — some items missed entirely
• Reflective surface — chrome on glass produces ghost detections
• Cluttered background — outlets, drawer pulls, or clutter detected as items

Any one of these is fixable in 30 seconds. Move the light, swap the background, retake the photo.

Don't Have a Photo Ready? Use One of These

If you want to test GridPilot before laying anything out, the project page has a few sample photos you can run through the pipeline — a socket set, a screwdriver lineup, a bag of small electronics parts. They're a good way to see what the output looks like before committing to your own layout.

The Photo Prep Checklist

Before you hit upload, verify:

• Top-down angle, phone level
• Even lighting, no hard side-shadows
• Background contrasts with items
• 5–10mm gap between items
• Background is plain (no patterns or text)
• File is JPG, PNG, HEIC, or WebP

Hit all six and the first generation almost always lands.

Try It With Your Layout

The fastest way to get comfortable with what works is to try one photo, see the result, and adjust. Start a new project on GridPilot → and upload your photo. If something looks off, swap one variable (lighting, background, or angle) and try again.

Want a deeper walkthrough of the full upload-to-3MF flow? See the 60-second walkthrough →