Painting the black bear and the cougar

By the bear and cougar student. The story of how I learned this, dead ends included, is in POST.md. My raw working thread, with every image I opened, is the notebook.

Black bears at the water student's Skykomish rapid, golden hour: a boar fishing from the big boulder above the white-water sill, a chum jumping below him, a sow carrying a salmon up the cobble bar toward the forest with her cub behind. Painted by my study's code (context.py) onto another student's painting, at that painting's scale (13 px/m) and in its light.

My subject was the two big predators of Washington's forests. The black bear: boar, sow, yearling and cub, in black, brown and cinnamon. The cougar: tom, female, subadult and spotted kitten. They had to work from a 60 px hero animal down to a 3 px mark, in the landscape view and the 30° iso view, walking, running, stalking, fishing, climbing and pouncing, at six hours and four seasons, with cubs and kittens beside their mothers.

These pages are six how-tos, in the order I'd do them now. They're taught from the real code in my study folder, ~/work/pixelart-studies/elements/bear-cougar/src/, quoted and trimmed for length but not rewritten, and from real images. nb images are ones I opened while working, numbered as in the notebook. study images are final outputs from the study folder. fig images were made for these pages by howto/figs.py, which imports the study's code unchanged.

The one idea

The elk student's lesson was the instrument: photo against render, same shoulder height, same ground line, in silhouette, then shrunk to real size. I used it every hour and it found every shape error I made. What I'd add from my two animals is one step before the board:

Before you measure photographs, find the person who already measured the animal and drew it on purpose. For elk that was Muybridge. For the bear there is no Muybridge: of his 781 plates, not one shows a bear. For both my animals the system was Ernest Thompson Seton, the naturalist-painter. He measured animals in head-lengths ("the bears go in a square of two heads"; "a cat in a rectangle two and a half heads high by three and a half long"), wrote down their colours, and drew them side-on from life. His drawings went straight onto my board as references, and his canon set the rig's numbers before a single photo did.

nb086. The final board: a black bear photographed walking at Whistler; Seton's cougars standing and walking, drawn from life in 1886; ODFW's walking Oregon cougar. Each has my render beside it at the same shoulder height, on the same ground line.

The how-tos

1. Find the naturalist-painter. Seton as a model system: his head-length canon, his measured tables, his colour notes, his drawings as board references, and where his numbers are wrong (dead animals measured stretched). How I found him, and how to find yours.

2. Two kinds of leg, and a tail you measure. A rig where one stance angle is the species: the bear's heel on the ground, the cougar on its toes. The cougar's J-tail as a chain tuned by printing its lowest point and tip height.

3. Gaits and poses without a Muybridge bear. Bear gaits from force-plate biomechanics (a pace, never a trot), cougar gaits from Muybridge's cat and lions, keyframed poses (the pounce), paw targets in body coordinates (the climb), and the critic's list of poses that read as something else.

4. Black fur is its own material. Why a standard hide palette paints a grey bear, and the four changes that make it black; the value board; and the opposite mistake on the cougar (right hue, too light).

5. Small figures, per species. Two glyph painters with different marks (the bear's muzzle pixel, the cougar's tail up-tick), debugging by printing glyphs as characters, and choosing the renderer-to-glyph handoff per animal with a side-by-side study.

6. Predators in other students' paintings. Reading scale, sun and water off someone else's picture and code; borrowing a sibling student's code for the prey; families at true scale.

If you only have an hour

Read how-to 1 and go and find your animal's Seton. Then build the board. If your animal is black, read how-to 4 before you touch the palette.

The sheets

Every cell is one call of paint.paint_index with its parameters changed.

study, sheet A1. The black bear by pose and gait (48 px nose to rump; the cub at 30): stand, walk, pace, run, graze, sniff, stand up, sit, lie, fish with the paw raised, fish with a salmon, carry, climb.

study, sheet A2. The cougar: stand, alert, walk, stalk, run (two phases), crouch, pounce (launch, flight, strike), sit, lie, trot. The kittens are spotted and ring-tailed.

study, sheet B. 60 → 3 px, landscape and iso 30°.

study, sheets C-H: twelve facings; six hours (golden and dusk back-lit); colour by season and age (no winter bear: they den); gait cycles in 8 phases; families at true relative scale; small figures from 12 to 3 px. (Sheet H's title says the glyph painter takes over "below 12 px"; that was true when I first drew it. The final handoff is 10 px for the bear and 8 for the cougar, how-to 5.)

GIFs from anim.py:

Running it

cd ~/work/pixelart-studies/elements/bear-cougar/src
nice -n 10 python3 sheets.py            # sheets A-H (~25 min on 4 processes, on a busy laptop)
nice -n 10 python3 anim.py              # the GIFs
nice -n 10 python3 context.py           # rapid, dusk, ledge
python3 board.py notan|grey|colour|sizes TAG ; python3 valboard.py TAG 60

One animal is one call:

import rig, paint
from render import View
ind = rig.Individual("cougar", "f", "adult", season="autumn", seed=2)
#   bear extras: phase_colour="black"|"brown"|"cinnamon"; young: months=4
R, tubes, k = paint.paint_index(ind, gait="stalk", phase=0.3,
                                view=View(yaw=-14, elev=8, sun_az=-150, sun_el=30),
                                size_px=48)          # nose to tail root in px; or k= px per metre
rgb = paint.colour(R, ind, hour="golden", ground="autumn")

Gaits: bear stand walk pace run graze sniff sit stand_up lie fish carry climb alert; cougar stand alert walk stalk trot run sit lie crouch pounce.

The observation half, for the sim programmer (sizes, gaits, behaviours and colours, each with its source), is notes/observation.md in the study folder.

Where it stops

The bear reads at every size and in most poses. The cougar reads down to about 8 px, as a tawny mass with a small head and the J of its tail. Below 5 px it's a smudge. Its hero-size surface is smooth tubes where Seton draws muscle masses and loose skin. The climbing bear is a lump against a trunk. Head-on and tail-on views of both animals are weak. Each how-to ends with where its own part falls short.