Zixxtrixx
snake
The first creature, from two concept sheets by S. Hofer. One continuous subtly tapered tube swells into its face rather than wearing a separate skull, then runs through a weight-bearing middle into a thin forked tail. Twenty-eight bones carry the bend, the bulging coordinated eyes and their boundary-reaching orange stripes. This v9 main is smooth three-band cel light over calm full-colour crayon, finished by exterior ink that stays bold close and yields at gameplay distance.
The final continuous S breathes on several calm periods while a low travelling twist keeps the planted middle alive without turning idle into walking. The raised front, weight-bearing run and thin fork all retain their own rhythm. Both pupils share one restrained authored glance; each orange stripe stretches with its pupil and still reaches the bulging eye boundary at native 384×240.
The accepted locomotion reference: a broad arch travels down the grounded run, bunches the body and draws the rear forward while the signature S and planted contact survive. The fixed camera keeps the deformation and weight transfer legible; smooth toon bands travel over the tube instead of announcing its triangles.
The same caterpillar law at a hungry cadence: a taller, broader hump crosses the grounded run every 0.8 seconds, the front wave digs deeper and the nose leans into travel. Contact compensation remains exact while the projected-size ink keeps the moving silhouette coloured.
The body idles while the skull turns left, up, right, down and home with eased arrivals and soft neck follow. The two pupils aim together with restrained holds and a modest lag instead of wandering independently. Their orange vertical stripes extend and contract as one elastic marking system so each outer endpoint stays on the eye boundary through vertical and diagonal travel.
Not a rod on its fork tips: Zixxtrixx gathers into a difficult weight-bearing L while a broad section of tapered tail lies along the ground. Two slow phase-lagged waves keep curvature travelling through the complete raised body; it tries and fails to straighten, buckles progressively, flops with one declared ground bite and recovers exactly into the S.
The preserved laugh-slot performance: the front rears proud, blades flare and the skull wags while the neck and middle answer. Its complete historical 26-bone motion and root payload remain byte-for-byte unchanged; v9 merely appends the two neutral pupil channels. The new slow neck-led wobble has its own neighbouring tab.
A separate center-left-right-left-center performance. The lower neck initiates each broad change, the skull follows four authored keys later with a funny ear-to-shoulder tilt, and quiet body life continues underneath. It is more than twice the quick taunt’s duration, starts and ends at exact rest, and never becomes a fast head shake.
The complete signature S compresses almost flat in a short readable beat, releases upward as one spring, turns once, lands with a declared ground bite, absorbs the force and settles exactly. This and the three-turn jump are outputs of one deterministic fixed-point builder; only the programmed turn count differs.
Same compression, release, trajectory, airtime, wheel shape, landing and settle as the one-turn jump, now with three coherent complete rotations. Changing count does not drift the upright landing phase or create a wrap snap.
The shared whole-body spring first compacts the complete S until its raised silhouette is almost flat, holds, then releases into the approved wheel. Three turns climb to the 12 m attack apex; the body unrolls into a diagonal spear, plunges along its own axis, bites the ground by an authored 420 mm and stays planted for five seconds before extraction.
Impact folds a visible front length into an asymmetric hairpin, shoves the complete animal and sends a delayed deformation through the grounded run into the supporting tail. One broad overshoot and loose recovery return every channel exactly to rest; the longer clip lets the propagated tail response finish instead of snapping away.
Four forceful reactions remain unmistakably distinct: mirrored lateral hairpins and whole-body side shoves, a forward/down fold, and a deep top crush. Each confines the violent onset, propagates it through the long body, then recovers loosely to exact rest rather than vibrating.
The donor pairs knocked2Floor with getUp, and a creature that cannot get up is broken in a way no taunt can fix. The blow lands on key 0: the body drains onto its flank with one small rebound and HOLDS -- a held corpse no longer flashes back to the stance; that wrap was a real engine bug this pass found and fixed -- then getUp cuts in on a byte-identical knocked pose the asset compiler enforces: the head wakes first, the roll rights itself, and the S gathers back with its belly kissing the dirt the whole way up.
The S recurs instead of remaining rigid. Slow stance authority, nonuniform tumble, neck loll, lateral writhe and a stronger two-cycle accumulated bend travel through the body, alternating straighter stretches with deep C and S shapes. The propagation is faster and larger, but every term remains broad enough to read as helpless loose mass rather than twitch.
hitFloor in the donor's vocabulary: the state machine cuts falling into this at ground contact. The body flattens as it drops, slams with a declared 24 mm bite, rings out one small rebound into the same knocked pose the knockdown holds, and gets up the same way -- one shared pose, three clips, compiler-checked seams.
The programmable attack plan commits to the grounded watchdog target 4.6 m out. The spear enters the posed model, holds with attacker and victim bit-constant instead of jittering, extracts clear, recoils through the body and settles back into the signature S. The camera retains both animals through the actual contact rather than substituting terrain.
The same planner intercepts a flying watchdog at 3.2 m. Target lead converges through the real impact key; the committed spear enters and holds against the frozen posed victim, extracts visibly, recoils and lands. A distinct apex and locked vector come from the target plan rather than a copied stunt.
The preserved limit baseline uses the shared full-body spring and coherent wheel builder for six complete turns at an exact 12,000 mm root-trajectory apex. It spends every turn before the committed diagonal dive and recovers with no wrap snap or landing-phase drift; the nine-turn probe doubles this apex without replacing it.
The deliberate system-limit probe: the shared spring and attack builder perform nine coherent turns at an exact 24,000 mm root-trajectory apex—twice the preserved six-salto baseline—then enter the posed target, hold exactly, extract, recoil and recover. Fixed-point range, duration, camera, LOD, outline, phase and target gates all pass.
Two slow shudders, and the S drains out of the body: every slope slides toward a lying pose whose centreline follows the animal's own taper -- a corpse's head rests one head-radius up, its thin tail lower -- while a computed constraint keeps the planted rear exactly where it stood and the carried front comes down to the dirt. It keels onto its flank, the tail curls once and lets go, and the last second is deliberately almost still. Stillness after motion is what reads as death.
The donor picks at random among a creature's filled deaths, so one death is a looped inevitability and two are a repertoire. This one rears up in agony first, then the collapse runs forward into a mostly-prone corpse -- the shallow tilt keeps the flank toward the lens; the first death's full keel taught us what a camera-facing back reads as -- two decaying tail slaps, and the last second deliberately still.
Six preserved generations, newest first: the v8 normal main, its style experiments and complete faceted-cel set, then the rebuilt-form, animation-proof and first-sketch records. Pick one generation tab at a time; archived clips never autoplay.
v8 normal main · 2026-08-29
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The canonical S: front reared, middle dipped, tail carried high. Five things happen at once on different periods -- the whole S gathers and releases on a 3.2 s breath, the animal bobs vertically, the body swells in girth, the tail sways, and the grounded run now SNAKES SIDEWAYS in a slow travelling S (the part that used to just stand there). The head rises with the in-breath on its own bone. The sideways wave turns about the true world vertical by quaternion conjugation, so the authored millimetres of belly sink survive it exactly.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. Locomotion is vertical and longitudinal, not lateral: an arch travels down the body, the middle rises and bunches, the rear is drawn forward. The head and neck bob with it. A small roll survives as secondary life at a fraction of the vertical authority. The camera is deliberately fixed -- a gait has to be legible before an orbit is allowed anywhere near it.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The walk's caterpillar law at a hungry cadence: the hump is taller and crosses the grounded run once per 0.8 second, the front wave digs deeper, and the nose leans into the travel. The cycle's period is the thing held constant -- the sim owns the ground covered, and the taller hump needed its own wider footprint before the exact height conversion would carry it.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The head-aim rig performed: the body idles on a quiet half-breath while the head turns left, glances up, turns right, dips down and comes home -- eased, with a tiny overshoot at each arrival, the first two neck joints following softly so the turn reads at the neck of an animal that has none. The head is the ROOT bone in this rig, so the aim lives on the dedicated skull bone: the same channel the sim will one day use to make him watch you.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. He gathers onto his planted tail, stretches up toward vertical -- ALMOST the attack's rigid spear, but effortful, wobbling, never quite straight, which is the joke -- fights for the balance, loses it, and topples flat with an authored ground bite before getting back up into the S. The root is enforced from a hand-keyed fork-height curve, so the stunt genuinely pivots on the tail, and the enforcement absorbing the wobble is what keeps the foot planted through the fight.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The donor spends eight whole slots on personality, and this is ours landing in its laugh slot: the front lobe rears up proud on lifted neck segments -- rearing by un-deepening the arch folded the hook over the head, and the overlap probe caught it -- while the blades flare wide and the head wags side to side, slow and eased, the neck following, the mid-body counter-swaying a whisper.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. Zixxtrixx does not bite. He compresses, rolls into a wheel, somersaults three times while climbing to a twelve-metre apex, hangs there a beat longer than gravity would like -- the owner asked for that hang, for effect -- then straightens into a rigid spear and drives eleven metres down one straight 30-degree line. The tip bites 420 mm into the ground and STAYS for five full seconds before the pull-out closes the loop.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The S snaps tighter -- the same lever as the breath, pushed hard -- while the head whips back and aside with the first neck joints carrying a third of the whiplash. One damped overshoot, and the canonical S is back inside 0.6 s. Both ends of the clip sit on the rest pose, so the hard cut into and out of any other clip lands clean.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The donor picks the flinch by the direction of the blow, and a creature missing a direction shows NO reaction at all -- the fallback is the bare stance. So all four are filled: a lateral whiplash off the blow to either side, a forward shove with the head whipping down, and a crush from above that flattens the whole arch while the head ducks. Every neck snap rides the compensated wave lane -- composing raw neck quats pitched the whole body, and the pose probe caught the tail 233 mm under the dirt.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The donor pairs knocked2Floor with getUp, and a creature that cannot get up is broken in a way no taunt can fix. The blow lands on key 0: the body drains onto its flank with one small rebound and HOLDS -- a held corpse no longer flashes back to the stance; that wrap was a real engine bug this pass found and fixed -- then getUp cuts in on a byte-identical knocked pose the asset compiler enforces: the head wakes first, the roll rights itself, and the S gathers back with its belly kissing the dirt the whole way up.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. Helpless, and finally LOOSE: the S is no longer mandatory. Per-joint stance authority swings between 16% and 84% on one slow loop-exact cycle that travels down the body, so the animal stretches almost straight, collapses into a distorted curve, and the recognisable S only RECURS -- while the whole body turns about its own centre, head-down at the half and back up by the end. Wobble, not jitter: fewer and slower terms, never more and faster.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. hitFloor in the donor's vocabulary: the state machine cuts falling into this at ground contact. The body flattens as it drops, slams with a declared 24 mm bite, rings out one small rebound into the same knocked pose the knockdown holds, and gets up the same way -- one shared pose, three clips, compiler-checked seams.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The attack planner aims the committed spear at a target dummy 4.6 m out -- the watchdog quadruped borrowed from the console demos, reel staging only -- and the strike is a MID-AIR HIT: the spear meets the body, recoils off it, and the snake falls out of the sky, gathers, and settles back into its S. Projectile, not missile: the vector locks at the apex and whatever it meets is a collision verdict.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The flying dummy hovers at 3.2 m and the plan intercepts it there -- a shorter climb, the spear locked upward-forward, the recoil, and the long fall back to the dirt. The planner sized apex, spin and plunge from the intercept alone; the donor format's three attack slots are exactly for variants like these.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The full showcase apex earns SIX somersaults on a 44-key flight before the committed dive onto a far ground mark -- the spin law spends them all on the way up, and the diagonal spear arrives exactly oriented along its locked vector.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. Two slow shudders, and the S drains out of the body: every slope slides toward a lying pose whose centreline follows the animal's own taper -- a corpse's head rests one head-radius up, its thin tail lower -- while a computed constraint keeps the planted rear exactly where it stood and the carried front comes down to the dirt. It keels onto its flank, the tail curls once and lets go, and the last second is deliberately almost still. Stillness after motion is what reads as death.
Preserved byte-for-byte before the v9 cel-main, whole-body proportion and motion pass. The donor picks at random among a creature's filled deaths, so one death is a looped inevitability and two are a repertoire. This one rears up in agony first, then the collapse runs forward into a mostly-prone corpse -- the shallow tilt keeps the flank toward the lens; the first death's full keel taught us what a camera-facing back reads as -- two decaying tail slaps, and the last second deliberately still.
v8 experiment menu · 2026-08-28
Both concept sheets draw a bold black outline around the creature and around every colour region, and the renders never had it. This variant inks the silhouette at render time (a 2 px line that follows every pose) and draws the region boundaries into the atlas along the painter's own wobbled curves. The palette is untouched: ink is the drawing's vocabulary, not a new colour. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The same ink-line variant on the gait, so the line can be judged in motion. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The isotropic grain replaced by long directional pulls of the crayon running down the animal's length. Same pigments, different hand. Tagged experimental: the normal creature above is untouched and renders byte-identically.
Where two crayon regions meet, a hand deposits more wax: the melds carry a darker, more saturated, wobbled band instead of a computed gradient. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The paper's white tooth showing through the pigment on the high points, plus a coarser grain. Paper white is the medium's colour, not the creature's. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The same colour law with every boundary wobble amplitude tripled: the regions stop being clean curves and read drawn. Tagged experimental: the normal creature above is untouched and renders byte-identically.
Snake markings drawn into the pink dorsal band in the band's own pink pushed darker -- value variation inside a region, never a new colour. Tagged experimental: the normal creature above is untouched and renders byte-identically.
Hand-colouring never lands exactly inside the line: each pigment region is shifted by its own wandering offset against an inked boundary, leaving slivers of bare paper where the crayon fell short and spill where it overshot. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The one temporal experiment: a chunky deterministic per-4-frame displacement of the creature's pixels, the shimmer of hand-drawn animation. A pure function of the frame index -- replay-exact, never a clock. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The same boil on the walk, where the shimmer reads against travel. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The Gouraud lighting term snapped to two flat bands (hue preserved; thresholds are named constants). A hard terminator on a curved tube -- judge whether it reads as style or mistake at 240p. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The softer cel: three bands, so the tube keeps a mid-tone between key and shadow. Tagged experimental: the normal creature above is untouched and renders byte-identically.
Flat cel bands bounded by the black ink line: toon shading and line art evolved together, and this pairing is the sheets' own look -- bold outline, flat crayon fill. Every other variant stands alone; this is the single authorised mix. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The cel-plus-contour pairing on the walk. Tagged experimental: the normal creature above is untouched and renders byte-identically.
The normal smooth-lighting pipeline, unchanged, with the contour atlas supplying interior region ink and a deliberately heavy five-pixel screen-space silhouette drawn inward from the rendered creature. Because the outside line is a presentation pass rather than atlas paint, it stays thick through distance and mips. Tagged experimental: normal remains the default and renders byte-identically.
The independently selectable faceted-cel presentation: the shared creature renderer holds each triangle at one of three named light levels, uses the same interior contour page, and adds the same screen-consistent five-pixel silhouette. Geometry, animation, skinning, visibility, camera and texture sampling are shared with normal; only the final light and ink strategy changes. Flat faces intentionally expose the authored topology.
The second honest cel read, kept separate rather than silently replacing the faceted one: coherent shared-vertex normals produce one smoothly interpolated light scalar, then each fragment chooses one of the same three authored levels. The broad bands follow the surface instead of revealing each triangle. Interior RGB565 ink and the inward five-pixel mask silhouette remain identical.
Long wandering graphite searches and sparse broken cross-contours ghost across the colour, like construction lines the finished drawing never fully erased. Graphite is a medium colour rather than a new creature pigment; line weight, alpha and cadence are named authoring controls chosen at 384x240. Tagged experimental: the shipping page is untouched.
v8 faceted cel collection · 2026-08-28
The complete faceted-cel3 and five-pixel-ink animation set
One separately selectable presentation strategy, applied after the normal creature and initial experiment menu were already live: all 17 canonical clips pass through the same face-light comparator, authored three-level table, contour RGB565 page and inward five-pixel visible-creature mask. The geometry, animation, skinning, cameras and texture sampling are the canonical set. Clips use controls instead of autoplay so opening this one collection does not start seventeen videos at once.
v7 rebuilt form · 2026-08-28
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
Kept because the owner asked for it by name. This is the animal whose head five straight passes could not fix: the front third of the S DESCENDED into the skull, so the head hung from the bottom of a downward hook no matter how the skull itself was sized, angled or blended. Every pass enlarged or re-angled the head to compensate, which only made the wrong mounting more visible. It is also ~2x the drawn thickness, because a pose-confounded measurement read the tube as a wire and it was thickened in the wrong direction.
v2 animation proof · 2026-08-27
The canonical S, breathing on four incommensurate periods. The motion was right and was frozen -- what dates this pass is the surface it moved: a flat-shaded 4,370-triangle tube, an indexed page sampled nearest so the crayon reads as pixels, and a dorsal pink narrowed to a thin stripe to compensate for a camera pitch that had already been fixed.
The caterpillar gait and the head-bob, both frozen and carried forward unchanged. Same surface as the idle above, and the same flat face: at the eye stations the skull was 1.97 times as wide as it was tall, which is a disc, not a head.
Three overhead somersaults into a rigid spear. The trajectory, the 12-metre apex, the aligned plunge and the impact shake are all frozen. Kept because it is the pass that proved the animation system works while the head was at its worst.
The corkscrew after it was slowed down -- three fast incommensurate terms per joint had read as vibration rather than as a body yielding. Fewer and slower, not more and faster.
v1 first sketch · 2026-08-26
Kept deliberately. Ten independent rigid cylinders at six sides and about 450 vertices, a horizontal ground-snake silhouette, and colours taken as the MEDIAN of each crayon region -- which is the pigment lightened by however much paper showed through, and was then hand-saturated to compensate. The orange was deleted outright to fit a 256-colour GIF palette. This is what the creature looked like when an export format was allowed to decide its shape.
The tail rose to vertical and came back down behind the animal, because the per-joint rotation was rear-weighted and the accumulated turn was therefore a fifth of what it looked like on paper. The joints also switched between two rotation axes on a magnitude threshold, which popped as they crossed. Both are fixed; this is the record.