128 ORIGINAL PREVIEW MEASUREMENTS · OCTOBER 26, 2026
Roblox Shirt Template Checkerboard Offset: What Changes in a 64-Pixel Preview?
Moving the background alone can change a fine line's measured contrast. We shifted a 4-pixel checkerboard through every x/y offset from 0 to 7 while keeping each saved Shirt PNG fixed. The one-pixel source stroke produced 0–42 strong-contrast pixels across 64 offsets; the two-pixel stroke produced 70–112. The artwork did not move. These are pixel measurements, not a human visibility or Roblox avatar test.
“Strong contrast” here means Euclidean RGB distance of at least 128 from the blank checkerboard at the same pixel. That cutoff is our analysis choice, not a Roblox requirement.
An eight-pixel shift cycle, measured
We made two new 150 × 150 transparent source PNGs, each with a centered red 45° line. One source used a one-pixel stroke; the other used two pixels. Each went through this site's Shirt editor at 45% Image size and was downloaded as a 585 × 559 PNG. We cropped the same 128 × 128 FRONT panel from each saved file, following the panel dimensions in Roblox Creator Hub's classic clothing guide. Chrome Canvas smoothed the crop to 64 × 64 over white and charcoal squares, each four preview pixels wide.
The squares repeat after eight pixels, so x offsets 0–7 crossed with y offsets 0–7 cover all 64 distinct placements of this fixed grid. For each placement we compared the finished preview with an identical blank checkerboard pixel by pixel. The table below records all eight x offsets with y held at zero, plus the full 64-placement range.
| Checker x offset (y = 0) | 1 px source: strong pixels | 2 px source: strong pixels |
|---|---|---|
| 0 | 42 | 107 |
| 1 | 31 | 103 |
| 2 | 21 | 87 |
| 3 | 11 | 70 |
| 4 | 0 | 75 |
| 5 | 11 | 79 |
| 6 | 21 | 95 |
| 7 | 31 | 112 |
| All 64 x/y offsets | 0–42; average 21 | 70–112; average 91 |
For the one-pixel line, the first one-pixel x shift changed the strong count from 42 to 31: a drop of 11, or about 26% of the initial count. At x = 4 it reached zero. Eight of the 64 offset pairs gave zero under the strong cutoff. The two-pixel line never reached zero; its smallest count was 70. The averages above are arithmetic means over the 64 offsets, not observations about viewers.
At the weaker distance cutoff of 64, every offset still had 42 changed pixels for the one-pixel source. Thus “zero strong pixels” does not mean the design vanished or the Shirt PNG lost its line. It means the composited colors missed our stricter threshold in that alignment. For the two-pixel source, counts at the weaker cutoff varied only from 125 to 128.
Four exact-size examples
These are the measured minimum and maximum phases for each line. They are displayed at their native 64 × 64 pixels. Background placement is the only difference within each pair; the crop, artwork, colors, preview size, and smoothing stay fixed.
offset (4, 0)
offset (0, 0)
offset (3, 0)
offset (7, 0)
A practical check for your own shirt
- Export the design once. Keep the saved Shirt PNG as your fixed source for each comparison.
- Use the intended region. Crop the FRONT or another panel you are evaluating, and use the same crop every time.
- Reduce to the actual preview size. A large zoomed canvas can conceal problems that appear in a small listing or mockup.
- Shift the background. Move a patterned swatch horizontally and vertically; changing its colors alone does not cover every alignment.
- Revise, then retest. If a feature depends on one favorable alignment, try a wider stroke or a different color, export again, and compare the same preview setup. Test the file on a Studio Block Avatar before making claims about garment appearance.
Scope: This test covers one diagonal shape, two source widths, this editor's Shirt export, a four-pixel checkerboard, Chrome Canvas smoothing, and a 64-pixel preview. It does not measure human recognition, every pattern, or Roblox's 3D clothing renderer.
Sources and reproducibility
- Roblox Creator Hub: Classic clothing — official panel sizes and Studio test method.
- Avatar Shirt Maker editor — the actual import and Download PNG workflow tested here.
- Earlier two-phase comparison — the question that prompted a complete offset sweep.
The original browser test script, 128 measurements, and four exact-size PNG examples are retained in the project audit. The controlled experiment, calculations, table, images, and repeatable five-step procedure were made for this article.