Caliper-free 3D printer calibration from a flatbed scanner: skew, shrinkage, pressure advance, and extrusion multiplier. Print a coupon, scan it on an ordinary office scanner, and get ready-to-paste firmware or slicer corrections. No calipers, no measuring, no eyeballing test prints, no typing numbers into a calculator.
Runs entirely in your browser. Nothing to install, nothing uploaded to a server, on desktop or phone.
Tip
A regular office flatbed scanner is all you need. The same kind you'd use to copy a document, including the scanner built into an all-in-one printer. No camera photos, no special hardware, and no high end machine required. Scan at 600 DPI, which any normal home or office scanner can do.
Warning
XY calibration is solid. XZ and YZ are experimental: the standing-plate scans work, but the correction math for those planes hasn't seen the same real-world validation as XY yet. Sanity-check the results before trusting them on your printer.
- Once per scanner: scan any plastic card (a credit, debit or loyalty card) so ScanNTune learns your scanner's true scale.
- Print the plate(s): one plate per plane you want to check. XY prints flat; XZ and YZ print standing on-edge. Print only the planes you care about.
- Scan each twice: lay a plate on the scanner and scan it flat, then give it a quarter turn and scan it again. Repeat for any other plates.
- Drop them all in: open every scan in ScanNTune at once. It sorts them by plate automatically and gives you the firmware or slicer snippet for X/Y/Z scale and skew.
That's it. The whole thing takes a couple of minutes once the plates are printed.
ScanNTune also calibrates pressure advance from a scan, instead of you squinting at a tower or a row of lines and picking the one that "looks best".
- Set up a printer profile in the app (or import your PrusaSlicer or OrcaSlicer config) and download the generated G-code.
- Print the coupon: a solid base in one filament, then a pause for a filament swap, then 16 test lines in a contrasting color. Any two filaments work as long as they differ in brightness. Each line prints at a different pressure advance value and contains slow, fast, slow speed changes, so a wrong PA value bulges or starves the line at the transitions.
- Scan it once. ScanNTune measures each line's width along its length and scores how much it deviates at the speed transitions. The line that stays most even wins, refined to a continuous value between the steps.
The result is ready to paste: Klipper SET_PRESSURE_ADVANCE, Marlin M900, or RepRapFirmware M572.
On Klipper there's an optional follow-up coupon that sweeps smooth_time the same way.
ScanNTune also calibrates the extrusion multiplier (PrusaSlicer) / flow ratio (OrcaSlicer) from a scan, instead of you measuring a thin wall with calipers or judging a top surface by feel.
- Generate and print the coupon from your printer profile: a single-color part with rows of parallel single-bead lines at precisely known spacings.
- Scan it once, face down. ScanNTune measures the air gap between neighboring lines to sub-pixel precision; since the line spacing is known exactly, the deposited bead width falls out of a single subtraction, averaged over more than a hundred gaps.
- Enter your current slicer flow and get the corrected value back in the same format, plus an
M221command for prints that are already sliced.
Line centres don't move when beads print fatter or thinner, so the measurement is immune to printer axis stretch and material shrinkage. Filament that won't come off the plate (TPU, PETG) can be printed at the bed's front edge and scanned together with the build plate.
I built this because I got tired of dimensional calibration. The usual routine is a printed coupon and a matching calculator, like Vector 3D's "Califlower": print it, measure it corner to corner with calipers, measure the diagonals for skew, type all of that into the calculator, and paste the result into your firmware. The measuring is the annoying part: several caliper readings to take and keep track of, a diagonal for skew that's awkward to measure squarely, and then all of it typed into the calculator without a mistake. So I let a scanner do the reading instead.
Here's the same printer measured both ways: ScanNTune's result (left) and Califlower's coupon hand-measured into its calculator (right).
The two come out almost exactly the same, differing by only 0.05% in X, 0.08% in Y, and 0.03° in skew. They should match, because both are measuring the same printer. The only difference is that ScanNTune reads it from a single scan instead of by hand with a caliper.
- Extrusion width doesn't affect it. The coupon is a lattice of rings, and a ring's centre doesn't move when the walls print fatter or thinner. Over- or under-extrusion can't shift the scale or skew.
- Two scans remove the scanner's distortion. A flatbed scanner has its own slight stretch and skew. Scanning the coupon flat, then again quarter-turned, and averaging the two cancels the scanner's error and leaves the printer's. The leftover half-difference even tells you how far off your scanner is.
- It calibrates scale off a plastic card. A scanner's stated DPI is rarely exact, so ScanNTune measures a standard plastic card instead (all cards are ISO/IEC 7810 ID-1, 85.60 by 53.98 mm) and reads the true pixels-per-millimetre from its edges.
The computer vision runs client-side in a Web Worker with OpenCV.js, so a full scan is analysed on your own machine without the page ever freezing.
The app is a plain Vue 3 + TypeScript + Vite project under
web/. You'll need Node.js 22 or newer.
cd web
npm install
npm run dev # dev server at http://localhost:5173/
npm run build # production build to web/dist
npm test # Vitest unit + fixture-backed engine tests
npm run e2e # Playwright end-to-end over real scansWant a different coupon size or grid? Edit calibration_coupon.scad in OpenSCAD
and export your own STL.
MIT © 2026 Jakob Eichberger