To use the grid method with a photo, crop the reference to the canvas’s exact aspect ratio, place the same number of rows and columns on both surfaces, transfer only contour crossings and major landmarks, and subdivide a cell only when it contains an ambiguous turn. The canvas cells may be larger than the photo cells; their proportions may not change. A perfect square copied into a rectangle is still a distortion system.
That ratio rule matters more than drawing a very fine grid. In our test, a 2:3 portrait mapped cleanly to a 16×24-inch canvas. Sending the same 4-column by 6-row map to a 20×24-inch canvas made every cell 25 percent wider than the proportional target, so the face, neck, and shoulders widened together. More lines could not repair the wrong geometry.
What the grid does—and what it does not do
Getty’s Art & Architecture Thesaurus calls this transfer technique “squaring”: a drawing is moved to another surface at a different scale, often a larger canvas or wall. The Metropolitan Museum of Art found the same basic logic beneath Jacques Louis David’s The Death of Socrates: corresponding grids helped move preparatory figure and drapery studies onto the final canvas.
The grid supplies addresses. A point can sit one-third of the way across column B and halfway down row 3 on both surfaces. It does not decide whether that point matters or whether the contour describes form. Those remain visual judgments.
Our cover uses one deterministic vector portrait rather than a person or physical artwork. Its landmarks stay fixed; only transfer geometry and grid density change:
- a 2:3 reference divided into four columns and six rows;
- the same 4×6 map enlarged proportionally to 16×24 inches;
- the same map forced onto 20×24 inches;
- a coarse map with only the face cells subdivided.
Gate 1: match the crop before drawing the grid
Reduce the photo’s width and height to a ratio, then compare that ratio with the canvas. A 4×6 photo reduces to 2:3. A 16×24 canvas also reduces to 2:3, so one can enlarge onto the other without stretching. A 20×24 canvas reduces to 5:6. It needs a new crop or intentional added space; it cannot receive the untouched 2:3 image by proportional scaling.
Adobe’s current Crop tool documentation lists fixed ratios and an original-ratio option. Work on a duplicate. If the canvas is chosen, crop the reference to its ratio before placing the grid. If the crop is fixed, choose a proportional canvas or plan the extra border.
Do not trust “looks close.” Divide canvas width by canvas height and photo width by photo height. The two results should match. This check catches the slow, global drift that is hardest to notice while working one cell at a time.
Gate 2: use the same map, not the same physical cell size
Choose a row-and-column count on the reference. Copy that count to the canvas. For our 4×6 map, each cell on a 16×24 canvas is 4×4 inches. On a small 4×6-inch print, each reference cell is 1×1 inch. The cells differ in physical size but share the same shape and address.
Number the columns and letter the rows so “the chin crosses C4” identifies one location without recounting from an edge.
Photoshop can display a grid with chosen units and subdivisions. Adobe describes it as a nonprinting overlay, useful for a digital working reference. A ruler and duplicate print work too.
Gate 3: transfer landmarks before contours
Mark a small set of high-consequence points first: top and bottom of the subject, left and right extent, eye line, chin, shoulder corners, and the largest negative-space corners. Locate each point as a fraction of its cell rather than rounding it to the nearest grid line.
Connect those points with the simplest useful contour. Note where an edge enters and exits each cell, but keep checking the whole silhouette. The National Gallery of Art describes Chuck Close maintaining proportional scale between a selected photograph and a much larger canvas, then transferring the image unit by unit. The map should not make the painter blind to the composition as a whole.
Every three or four cells, step back and ask:
- Does the full silhouette have the same lean and width?
- Do the eye line, chin, and shoulders keep their vertical order?
- Do the largest negative spaces have the same shape?
- Does a long curve continue smoothly across cell boundaries?
If a curve kinks at a grid line, compare the neighboring cells together. The line is a coordinate aid, not an edge in the subject.
Gate 4: subdivide only the ambiguous cells
A dense grid can feel safer because it creates more reference points. It also creates more lines to label, copy, erase, and accidentally treat as part of the image. Roger Fayet’s 2026 Getty Research Journal study records Ferdinand Hodler warning that an excessive number of small squares can confuse the view. Hodler used grids for transfer and scaling, but did not treat maximum density as the goal.
Our test reached the same practical decision without claiming a universal ideal count. The 4×6 grid was enough for the outside silhouette and shoulder slope. It was too coarse around the eye, nostril, and mouth corner, so we split only those face cells in half. This kept the body readable as a single shape while adding addresses where a small placement error would change the expression.
Start with the coarsest grid that gives unambiguous contour crossings. Subdivide one cell when two nearby landmarks share it, a curve reverses direction inside it, or the subject’s identity depends on a small interval. Do not subdivide background cells merely to make the page look systematic.
A reproducible 20-minute transfer test
Minutes 0–4: duplicate and crop the source to the final canvas ratio. Write both reduced ratios beside the images. Stop if they differ.
Minutes 4–7: choose a coarse grid, copy the same row-and-column count to the canvas, and label the edges. Measure from one origin rather than hopping from line to line; repeated small measuring errors accumulate.
Minutes 7–12: place eight to twelve landmarks. Include subject extents and negative-space corners, not only facial features or object details.
Minutes 12–17: connect the large contour and essential internal axes. Ignore eyelashes, hair strands, fabric pattern, and texture.
Minutes 17–20: step back, compare width-to-height relationships, and subdivide only the cells that still contain an uncertain turn. Photograph the transfer before erasing the grid so later drift can be diagnosed.
For transparent watercolor or thin glazing, remove or isolate construction lines before paint because they may remain visible. For an opaque block-in, keep them light enough that you do not spend the first layer fighting the map. The Met’s technical images show that historical transfer lines can survive beneath paint; that is useful evidence about process, not a promise that every unwanted line will disappear.
Failure modes and the right correction
The likeness is uniformly too wide or too tall. Recheck aspect ratio and cell shape. Do not redraw every feature inside a stretched map.
One side drifts while the other looks correct. Verify column widths from the original edge and remeasure major landmarks. A single misplaced grid line can shift every later address.
The contour breaks at cell boundaries. Review two or three cells as one shape, then redraw the long gesture across the artificial divisions.
The drawing is accurate but lifeless. Stop transferring texture. Keep the placement map, then make value, edge, and paint-mark decisions at the whole-image level.
The grid takes longer than the painting. Use fewer base cells and local subdivisions. The method should protect expensive relationships, not record every pixel.
Once the drawing passes the ratio, landmark, silhouette, and continuity checks, put the grid away. If you want to compare a painterly interpretation before committing physical materials, keep the same crop in the FreePainter photo-to-painting workspace. Judge whether the major shapes and focal relationships survive; the grid has finished its job when it no longer needs to be seen.
References
The museum sources establish the definition and documented use of squaring in real artworks and studio practice. Adobe’s current documentation establishes the crop and digital-grid controls. The 2:3-to-5:6 mismatch, 25-percent horizontal stretch, selective-subdivision plate, and four-gate workflow are original FreePainter Studio Research teaching artifacts, not measurements of a particular app or reconstruction of a cited painting.