A LEGO portrait maker that checks the face will read

The photograph almost certainly has no directional light in it. A brick face is built entirely out of the boundary between a lit plane and a shaded one — a cheekbone exists because the plane above it is lighter — and a flash fired straight down the lens, or an overcast sky, removes those boundaries before the file is even written. The tonal reading catches this: it takes the 5th and 95th percentile of perceptual lightness across the studs and compares the gap to the neutral ramp's own step size of 0.164. Under two steps the page refuses. Re-shooting beside a window with the subject turned 45 degrees away from it fixes more portraits than any setting here will.

There is no single one, and the honest answer is that the family is nine shades wide because real skin spans that range under one light. The minifigure colors are Light Nougat, Nougat and Medium Nougat, and they are not interchangeable with the tans and browns around them so much as ranked with them by lightness. Availability makes the choice harder rather than easier: Nougat has 3,084 cataloged elements against Reddish Brown's 180,748, so the shade named after skin is roughly 59 times harder to buy in bulk than the brown a meter of hair will need. Build the ramp from what you can order and let the lightness do the work.

Further than any room in a house, which is the fact that most advice about brick portraits gets backwards. A stud is 8 mm, and 8 mm subtends one arcminute — the ordinary limit of sharp vision — at 27.5 meters. At three meters, a normal distance for looking at something on a wall, one stud still covers 9.2 arcminutes, so it is nine times larger than the smallest thing your eye can pick out. You never stop seeing the grid indoors. What does fade with distance is hue: color vision resolves detail roughly three times more coarsely than lightness does, so a single stud of the wrong shade blurs into its neighbors around nine meters out while a single stud of the wrong brightness is still visible at twenty. That asymmetry is why this page organizes its palette as a lightness ramp and tests the photograph on lightness alone.

Because on a face they win cells they should not. Nearest-color matching has no idea that a region is a cheek, so a warm highlight sitting between two available tones can resolve onto Coral or Bright Light Orange when Nougat was the intended reading — and a single saturated stud on skin is read by a viewer as a mistake rather than as a highlight, in a way that the same stud on a roof or a hedge never is. Restricting the list to fifteen removes the possibility rather than relying on luck. The full palette stays one click away so you can see the difference on your own photograph, and for a landscape or a logo it is the right choice, which is what the mosaic page defaults to.

Usually not, and the exception is Atkinson. Error diffusion scatters isolated studs of a neighboring color through smooth areas, and skin is the largest smooth area in the picture — so on a forehead it reads as acne rather than as gradation, and it multiplies the number of colors you have to buy and keep sorted. Atkinson withholds a quarter of the rounding error instead of passing all of it along, so the mid-tones, which is where nearly all of a face lives, pick up far fewer stray studs. If a photograph genuinely needs help spanning the ramp, the tone stretch is the better instrument, because it changes the input rather than adding noise to the output.

Yes, but only because of the symbols, and this palette proves the point better than most. Medium Nougat and Flat Silver both sit at a perceptual lightness of 0.625 — a warm mid-brown and a mid-gray that a monochrome printer renders as the identical tone. Dark Red and Pearl Titanium are 0.001 apart on the same scale. The tone column in the color table shows those numbers, and the mark in the first column is what distinguishes the rows on paper, in poor light, and for the roughly one person in twelve who cannot separate two adjacent shades by eye anyway.

4:5 for a head with shoulders under it, square for a face that fills the frame, and the tool reports the percentage of the photograph's area you lose either way. The crop is taken from the center, so a subject standing off to one side will be cut into — trim the photograph yourself first if the composition matters. Keeping the photograph's own shape is the third option and is right when the framing is already what you want on the wall, but it means the stud count each way is whatever the camera happened to produce rather than a number you chose.