Everything tried for showing colour in Nijimi, what happened to it, and what it led to.
Where this landed. Of everything tried here, the card of five colours is what shipped. The mixing well, the radial dial and the single arc were each built or drawn, and each was dropped.
One dark pigment in three dilutions plus one accent, with ‹ › to page between eight sets.
It matched the field, which carries exactly two pigments. But three of five slots were the same colour three times, and eight sets behind an arrow meant most of the palette was invisible.
The sets collapsed into nine parent pigments. A 2×2 well on the left: drag right for more of the second pigment, drag down for more water.
An XY pad shaped like a dish. You set a value, you never did anything, and the result was one flat colour because a coordinate cannot be streaky.
Dabs dropped into a 20×20 field; dragging smeared them through each other, streaky until you had worked it.
Worked at 300px with a mouse pointer. At 91pt a fingertip covers a third of the well, so you stirred blind. Every drag also reloaded the brush, so the colour kept moving under you.
A second 48×48 simulation using the paper’s own shaders, with settings for a dish: no sizing, no tooth, slow evaporation. It opened to 264pt so aiming was possible.
Pixelated at that field size, unclear what was happening, and the colours that came out were not good. Three attempts at a mixing area, all worse than what they replaced.
Top row the current set; bottom row a chip per set, split between its dark and its accent, to switch families.
Half the palette was navigation. You spent a tap to see colours rather than to use one.
Every dark and every accent in the direction on screen at once, 17 swatches at 24pt or 21 at 17pt. Tapping the dark you were on walked it down its dilutions.
It worked. But it is two rows, and two rows of small swatches is a lot of palette for an app that is meant to be about the paper.
A direction holds seventeen to twenty-one colours. One row holds about ten slots at a size worth tapping. So a slot has to carry more than one colour, or the row has to open. Each of these is drawn at true phone scale (393pt wide, 345pt of usable row), because the last idea to fail did so on scale, not on the idea itself.
Every colour in the direction, in one row, one tap each.
393 pt
Each slot is cut on the diagonal: the family’s dark on one triangle, its accent on the other. Ten slots carry twenty colours. Nothing is hidden and nothing is navigation: every triangle is a colour you can paint with.
The cost: A triangle is a smaller target than a square, and the two halves of one slot belong to the same family, which the shape has to make obvious or it reads as decoration.
Where you tap on the swatch is how strong the paint is.
393 pt
Each slot is that pigment down its three dilutions, drawn as the gradient it actually is. Tap high for concentrated, low for a pale wash. One tap picks the colour and the strength together, and the slot shows you the whole range before you commit.
The cost: Accents need somewhere to go, either alternating slots or the diagonal from 半. And a 41pt slot split three ways gives each dilution about 13pt of height.
One row of families. Hold one and it opens.
393 pt
The row shows eight or ten families at a comfortable size. Press and hold any of them and its full set (three dilutions, the accent and water) fans out above your thumb, at a size you can actually hit. Release on the one you want.
The cost: A hidden gesture. Everything past the first tap has to be learned, and the fan covers the paper while it is open.
No row at all. The palette comes to your hand.
393 pt
Hold anywhere on the paper and the colours arc around your thumb, where it already is. Let go on one. The bottom of the screen goes back to being paper, which on a phone is most of what there is.
The cost: The most invisible and the most to learn. It also competes with the brush itself, because a hold has to mean "palette" without ever meaning "start a wet blot".
半 split swatch is the one that answers the complaint directly: one row, every colour visible, nothing that costs a tap without giving paint. It is also the least novel, which after five attempts counts in its favour.
濃 depth swatch is the more interesting idea, because it makes dilution a place rather than a repeated tap, and dilution matters a great deal in this medium. The two combine: a diagonal split where each half is itself a gradient. That is a lot of information in 33pt, and worth trying before assuming it is too much.
扇 and 弧 both give the paper back its screen, and both spend that on a gesture nobody will find on their own. Worth keeping for when the palette is settled and the app is being made quieter, rather than now, while the thing being judged is the colour itself.