Scaling a Reference Photo Without Distorting It
The grid method covers how squaring-up works in general. This one is about a messier, more common situation: a reference photo whose dimensions don't line up neatly with anything, and a physical support (a board, a pre-cut sheet) that doesn't necessarily match the proportionally correct size either.
The starting problem: awkward ratios
A typical phone or camera photo doesn't arrive in round millimetres — it's a pixel grid, often at a 4:3 or 3:2 ratio, that has no obvious relationship to whatever paper or board you're planning to draw on. Take a 1200 by 900 pixel reference photo (a standard 4:3 ratio) and a plan to transfer it at 12 grid columns onto a 210mm-wide support — the width of an A4 sheet. Run those numbers through the proportion & grid-transfer calculator and it returns a scale factor of 0.18, a proportionally correct target height of 162mm, a source cell size of 100 pixels, a target cell size of 17.5mm, and 9 rows. Notice the source width was in pixels and the target width was in millimetres — and the calculator didn't need to know a resolution (dots per inch) to get a correct answer, because a scale factor is a ratio, and a ratio between two lengths in the same source unit is unaffected by what that unit actually is. This is worth knowing on its own: you never need to look up a photo's DPI to grid-transfer it accurately, only its raw pixel dimensions as reported by whatever camera or phone took it.
When the support doesn't match
The proportionally correct height for that transfer is 162mm. Say you've already got a pre-cut board on hand at 158mm tall instead — close, but not exact. Entering 158 as the actual target height flags a distortion warning, because 158mm sits about 2.5% below the 162mm the geometry calls for, past the calculator's 2% tolerance. That's not a huge squash in absolute terms, but on a face or any subject with recognisable proportions, a consistent few-percent vertical compression is often just barely visible — eyes read as slightly too close together relative to the width of the face, verticals look a hair stockier than they should. Move to a 160mm board instead and the same check comes back clean: a 1.2% difference sits inside the tolerance, small enough that it's genuinely hard to spot without a side-by-side reference to compare against, which is exactly why a small tolerance band exists in the first place rather than demanding a mathematically perfect match to the millimetre.
What to actually do about a mismatch
Catching the warning is only half the problem; the more interesting question is what to do next, and there are three honest options, not one correct answer. The first is to change the support to fit the geometry — trim the 158mm board down further, or choose a different sheet size that's proportionally correct to begin with, which is the cleanest fix whenever it's practical. The second is to keep the support's size but stop insisting the whole reference has to fill it: draw at the correct 162mm height and leave a small margin or mat border on the short board, rather than forcing the whole image to squash awkwardly into a space that's the wrong shape for it. The third is to deliberately crop the reference to match the support's actual proportions, choosing which edge to lose — trimming a strip off the top of a portrait rather than compressing the whole head, for instance, is usually far less noticeable than distortion, precisely because a crop removes information rather than warping what remains. What doesn't belong on this list is quietly stretching the transferred grid to fill the mismatched support anyway; that's exactly the failure mode the tolerance check exists to catch before it's baked into a finished drawing. Of the three, matting a correctly-sized image onto an oversized or awkwardly-shaped support is usually the least destructive choice when it's available, since nothing about the reference itself has to be altered — the trade-off is purely in the finished piece's presentation, not in its underlying accuracy, which is usually the easier trade-off to accept.
Cropping before or after gridding
If you already know you'll need to crop, it's worth deciding on the crop first and drawing your reference grid over the cropped version, rather than gridding the full original and cropping the drawing afterward. Gridding after cropping means every square in your grid maps to content you're actually planning to use, so the column and row counts, and the resulting cell sizes, are chosen for the composition you'll end up with — not diluted by rows or columns of the original photo that were only ever going to be trimmed away, and it avoids the mildly demoralising experience of carefully transferring a beautifully gridded square that turns out to sit entirely outside your final crop.
A second scenario: enlarging rather than shrinking
The mechanics run identically in the other direction. Take a small 400 by 300 pixel thumbnail sketch, scaled up onto a 600mm-wide final support with a 10-column grid: the calculator returns a scale factor of 1.5, a target height of 450mm, a source cell of 40 pixels, a target cell of 60mm, and 7.5 rows. Only the direction of the multiplication changed — the scale factor is now above 1 instead of below it — not the underlying logic. Whichever direction you're scaling, the discipline is the same: compute the proportionally correct target dimensions first, compare them honestly against whatever you're actually planning to draw on, and make a deliberate choice (resize the support, mat it, or crop the source) rather than letting a small mismatch quietly become a stretched drawing.
Uneven detail calls for an uneven grid
A single grid density across the whole reference is the standard approach, but it isn't the only one, and awkward references are exactly where it's worth knowing the alternative. A portrait, for instance, typically needs far more positional precision across the face than across the shoulders or background — a 12-column grid sized for the whole photo might give the face itself only two or three columns to work with, not enough to place features accurately. A common workaround is a hybrid grid: draw the standard grid across the full reference for overall placement, then draw a second, finer sub-grid just over the face (or whatever region carries the most detail), transferring that region at the higher density while the rest of the image uses the coarser one. The scale factor from the original transfer still applies to the sub-grid's cell size, so the two grids stay proportionally consistent with each other even though their densities differ.
Marking a grid on a photo without damaging it
For a printed reference, the practical choice is usually a photocopy or a cheap reprint rather than the original, so the grid lines can be drawn directly and boldly without worrying about marking anything irreplaceable. For a digital photo viewed on a screen or tablet, a grid overlay drawn in image-editing software (or even simple guide lines added in a free viewer) avoids the physical-marking question entirely, and has the added benefit of being trivial to erase and redraw at a different density if the first attempt turns out too coarse or too fine for the subject. Either way, the final grid on the drawing surface itself is usually best kept light — a construction-line grade well up the H end of the scale, per how to choose the right graphite grade — since it exists purely to guide placement and should disappear easily under whatever rendering goes on top of it.
Choosing which dimension to anchor
The calculator takes a target width and derives the proportionally correct height from it, which matches how most people plan a drawing — by the width of paper they have. But the constraint sometimes runs the other way: a frame with a fixed height, or a support that's tall but narrow. In that case, swap which measurement you feed in as the "width" input — the underlying arithmetic is symmetrical, since a scale factor computed from height-to-height is exactly as valid as one computed from width-to-width, as long as you're consistent, throughout the whole transfer, about which pair of numbers you're actually comparing. What matters is anchoring the scale factor to whichever dimension is actually fixed by your materials, then treating the other dimension as the one to check for a mismatch, not the reverse.
A habit worth keeping
Running this check takes under a minute and it's worth doing before a single grid line gets drawn, not after the transfer is half finished. A reference photo's pixel dimensions and a physical support's millimetre dimensions rarely agree by accident, and treating that mismatch as a planning question — not something to spot only once it's visibly wrong on the page — is most of what separates a confident, deliberate enlargement from one that's subtly, invisibly warped until a viewer notices something is slightly off without ever quite being able to say exactly what it is.