Craft record: r15_p1 (materials and tools)
Each entry gives the operation, the effect I saw in the paint, and my explanation or uncertainty. Canvas: 440 mm linen, 1000 units wide, 2400 px (1 unit = 0.44 mm = 2.4 px).
Ground
- A brushed whitish top ground (#d9ccb2, 55 µm, stiff 0.4) over two knifed
layers (110 + 70 µm).
- Effect: a warm light ground with faint drifting bristle striations. Thin sky paint over it stays luminous; the striations show as a slight directional grain in the thinnest passages (low sky, mist).
- Explanation: 180 µm of knifed ground levels most of the weave, so the visible micro-texture comes from the brushed layer, as the research notes say for his grounds [KÖR p.284]. The weave was not fully gone, though (see 9).
Drawing
graphite::hand_linewith 2H along the crests, HB for firs and figure.- Effect: the 2H lines broke up on the tooth and were barely visible through the lean sky layer. HB fir axes showed clearly as grey lines in the glow until the firs covered them.
- Explanation: the harder lead lays less and sits on the crowns. KM compositing of a thin sky film over graphite lets it "shimmer through" [CATS p.132]. Body colour (the firs) hides it.
Sky
-
Broad filbert, lean (medium 0.3), coverage 4, strokes 160–340 units swept top to bottom, aimed at a gradient; two badger passes.
- Effect: a first attempt at medium 0.45 and coverage 3 left violet-blue strokes with pale ground flashing between them (thin film, gaps). Leaner, fuller paint with the second, crossing badger pass gave a continuous layer whose tonal unevenness reads as faint high cloud.
- Explanation: at medium 0.45 the film's K and S are scaled by 0.55, so one coat hides the light ground poorly and every gap or thin edge shows. The blender moves wet paint only a stroke's width. It evens stroke-scale variation but leaves variation at the passage scale (~1.6 stroke lengths), which is what survives as the cirrus.
-
Stipple over the dried sky (tip 2.6, coverage 1.3, dips every 20).
- Effect: the gradient improved at the passage scale, but at 1:1 it was a lace of lighter, worm-like clumps.
- Explanation: over dry paint each touch sits on top, so the passage is a
mosaic of discrete films. Each dip's pile is aimed at the centroid of the
~20 touches it serves, so neighbouring groups differ slightly in tone. The
default
cluster0.15 also groups touches. Where the first layer was darker than the target, the touches were lighter and stood out.
-
The same stipple into the wet sky (
cluster(0, None), dips every 8, coverage 1.4).- Effect: a fine even grain; the lace almost vanished. A few faint pale flecks remain.
- Explanation: touches into open paint lift some of it and split their film into it, so each mark is partly the mixture of the two. More frequent dips shrink each pile's patch. This is closer to what the notes describe as his stippled skies, where the thin paint pools in the ground's texture.
-
Transparent glaze of #e9a265 up to 0.4 coat over the low sky (after drying).
- Effect: the pale apricot deepened to a warm orange band above the far chain without losing the gradient. At 0.55 coat it became a hot, postcard orange.
- Explanation: a transparent layer multiplies the light reflected from the pale sky below it (KM with low S). Over a light underlayer that gives saturation without darkening much, so the depth is sensitive to amount.
Ranges and mist
-
Each range: body filbert clipped to the crest band, strokes along the crest's slope (
atanof its derivative × 0.8), colour field from crest colour down into mist colour.- Effect: crisp, found crest edges with slight bristle build-up (a hair darker at the very edge), and smooth down-slope gradients.
- Explanation: with
clip(true)each bristle's contact is multiplied by the mask, so the edge is the mask's 1.6-unit smoothstep. Strokes ending at the edge plough a little paint there, which gives the faint dark line.
-
Mist glazed (semi, 1.2–1.6 coats) over the dried ranges.
- Effect: soft and atmospheric at a distance, but at 1:1 there were regular vertical streaks ~3.6 px apart (≈ 0.67 mm), like rain.
- Explanation: the glaze levels and pools in the surface hollows. The spacing matches the 15 threads/cm warp, so the weave relief still comes through the 235 µm ground and the paint over it. A deep poured film is thicker in every interstice, so it prints the threads. I am unsure why the vertical threads dominate over the horizontal (13/cm) ones. Perhaps the warp crowns are higher, or the brushed top ground's striations happened to run mostly across them here.
-
Mist stippled onto the dried ranges instead.
- Effect: white "snow" dots on the dark range paint.
- Explanation: as in 4. Over a dark dry field each light touch is a separate opaque fleck.
-
Mist stippled into the wet range, then badger along the slope, dry, faint semi glaze (0.3 × amount).
- Effect: the mist reads as mist: soft, level wisps, with the range dissolving downward. No streaks and no dots.
- Explanation: wet-into-wet touches fuse with the range paint, and the blender drags the fused paint along the slope. The veil left to glaze is thin enough that its pooling contrast is below visibility.
-
Transparent violet glazes along the far chain (0.3) and third range (0.45) crests, fading 45 units down the flank.
- Effect: the far chain set back against the deepened glow and the ranges kept their order of darkness. The first masks, softened separately, left a pale seam along the crest; overlapped, they left a dark rim.
- Explanation: where two soft ramps don't sum to 1, the sum of glaze depths dips or bumps along the boundary. Complementary masks (s and 1 − s) fixed it.
-
A 0.6-coat transparent glaze on the second range, then a brushed glaze (
st.glaze(0.9)) instead.- Effect: the poured glaze printed weave streaks as in 8. The brushed one speckled on the weave crowns and, after two badger passes, looked paler and bluish, like snow.
- Explanation, uncertain: the soft filbert's thin medium-rich film touched mostly the crowns (light pressure clears only the tops), leaving the hollows as they were. The badger then redistributed a film too thin to darken much. I did not find why it came out lighter than the underlayer; possibly the aimed masstone plus residual scatter of lead white in the pile lightened it. I repainted the range darker (#525369, less mist) instead.
-
Ridge made to dive into the mist (crest y raised by a smoothstep toward its end) rather than faded by an x-ramp on the mask.
- Effect: a natural spur sinking into fog. The x-ramp version ended in a hard vertical cut with a blender swirl and a light fleck at the tip.
- Explanation: a side-faded mask still clips every stroke's crest edge where the mask is non-zero, so the silhouette ends abruptly. The blender, following the steep slope at the tip, stirred wet paint in a curl.
Foreground, rocks, grass
-
Body filbert along the ground's fall; unclipped first.
- Effect: strokes overshot the crest and one left a sliver like a log standing up from the knoll.
- Explanation: unclipped strokes reach past the region's edge by design
(
hug). Clipping plus a sable pass along the rim gave a found edge.
-
Heather:
hatch()touches at −1.4 rad, lighter and warmer, coverage 0.9, into the wet heath.- Effect: a lighter olive-rust band under the crest that models the ground turning up to the sky; with an OKLab lightness shift of 0.035 it became too light and grassy, so I settled on 0.022.
- Explanation: small touches into wet paint mix partly with the dark heath. Their net lift is less than the shift, but it is spread over a band, so it reads strongly.
-
Rocks from a
Form(rough ridged blocks and ellipsoids), light(-0.35, -1)from behind, ambient 0.25; body filbert from the lit value.- Effect: with a lit colour up to #79736c they read as pale pillows and a mattress-like plinth, pasted on. Darker (to #5d5852) and angular, they read as stone, but still sat on the heath.
- Explanation:
shade.valueincludes ambient × sky, so upward-facing planes of every rock get most of the light, whatever their mass. The silhouette is complete, so nothing sinks them.
-
Sinking: per column, the rock mask's top and bottom found by scanning (
f.per_column), the heath brushed back up to a wavering mid line, then grass blades along it.- Effect: the rocks sit in the ground; their bottoms are lost in heath.
- Explanation: this is geometry, not physics. The painter decides where the ground covers the stone.
-
Grass as pointed-sable blades (
point0.9, 1.1 wide, pressure 0.45 → 0.05, release 0.7).- Effect: fine tapered blades that break the crest against the mist; some catch a warmer ochre.
- Explanation: at light pressure only the cone tip touches, and the long release draws the mark to a point.
Firs
-
Hatch fill clipped to a tiered silhouette mask plus thin branch lines.
- Effect: stacked flat slabs with hard horizontal edges (the mask's own outline).
- Explanation:
clip(true)reproduces the mask edge exactly; the fill strokes can't make a softer edge than the mask.
-
Wide hatched core with thin (2.0) pointed branch strokes.
- Effect: a solid black cone with spidery arms, like a dead tree.
- Explanation: at pressure 0.3–0.6 a 2-unit pointed sable makes a mark well under a unit wide; the branches had no mass.
-
Thin spine, then each branch as one stroke of a pointed round 2.4–4.6 wide (
point0.6), pressure 0.55–0.9 → 0.08, release 0.7, sagging then lifting; twigs of a 1.2 round hanging from the outer two-thirds.- Effect: spruce branches: heavy at the trunk, tapering to a point, drooping with a lifted tip, fringed below. Overlapping branches build a ragged, asymmetric silhouette with sky between the tiers.
- Explanation:
mark_widthfalls with pressure, and the cone narrows as the brush lifts, so a single stroke is a wedge. Per-tier and per-side random lengths, with 12 % short tiers and 22 % missing branches, break the symmetry.
-
Trunk: a round scaled to the tree (1 + 0.045 × width), pressure 0.8 → 0.02,
point1.- Effect: with a fixed 2.6 width the tops were thick bare poles. Scaled and fully pointed, the trunk vanishes into the leader.
Figure and moon
-
Figure from polygon masks, painted with the sable (
detail, coverage 4,threshold0.05).- Effect: clean dark silhouette. A flat cap over the head read as a chimney; a bare head over rounded shoulders reads as a man.
- Explanation: at 140 px tall the silhouette is everything. The rim light (#6c6670, pressure 0.35 → 0.1) is almost invisible at this size.
-
Crescent: disc minus a disc offset 0.42 r away from the sun, sable touches of #f3ecd2, over a 0.18-coat pale halo glaze.
- Effect: a clean crescent with a soft glow; the thin lit limb faces the set sun.
Finish
-
Cracks::ageddefaults.- Effect: a strong dark network with islands of ~1 cm dominating the image.
- Explanation: hierarchy 1 and patchy 1 let the first cracks open widest. Dirt 0.4 and grime fill them dark over light passages.
-
Cracks at
width_um4,dirt0.12,veil0.15,hierarchy0.5,patchy0.6, depth 8 µm, cupping 6 µm; varnish 0.22 ± 0.06.- Effect: faint hairlines visible only at 1:1 in the lights; the varnish barely warms the picture.
Computation
- Timing and reproducibility.
- Timing: the whole render takes ~220 s. The sky (a broad pass plus two swept badger passes over 80 % of the canvas) takes 70 s. Each range takes ~18 s, the ground 37 s, the firs, figure, grass and moon under 1 s each.
- Iteration: resuming from a checkpoint cut an iteration to 10–75 s. A crop with its own checkpoints was useful for the mist and sky tests.
- Reproducibility: a from-scratch render was byte-identical to the render assembled from resumed checkpoints.