New primitives (tapered capsule with rounded caps, superellipsoid
rounded box) and a full body redesign: three-segment torso (chest plate,
dark abdomen, pelvis plate) with glowing core and belt light, helmet
with visor/jaw/ears/antenna, shoulder pads, tapered capsule limbs with
elbow/knee joints, hand plates with wrist lights, boots with toe lights.
Two-tone armour tinted by the palette, emissive details in colorB.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Sitting at a desk the hips are rarely visible, which made the avatar
never appear. The pelvis is now synthesized below the shoulders when
hidden (torso/head/arms as a bust; legs return when actually visible).
The output also reports diagnostics to the panel: pose tracker failed,
or avatar active with no body detected for 5s.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Raise your arm and the robot raises its arm. ModelSim gains primitive
meshes (cube/sphere) and renderAvatar(): parts placed by limb matrices
(basis aligned to each bone segment) with the same lighting as the GLB
renderer. interactive.js keeps the mirrored raw landmarks (with depth,
smoothed) and builds the robot every frame: torso sized from the
shoulders, head with accent visor and antenna, articulated arms/legs
with joint spheres, glowing hands, feet. Falls back to the motion mist
when no body is tracked.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
MediaPipe PoseLandmarker (BlazePose lite) integrated fully offline: the
WASM runtime and the bundled .task model are served to the renderer over
a new djvres:// protocol (fetch does not work on file://), with
'wasm-unsafe-eval' added to the output CSP. pose.js exposes PoseTracker;
interactive.js detects at ~30 Hz, maps landmarks into mirrored
cover-fitted screen space and smooths them with velocities.
Upgrades when a body is tracked (silent fallback to motion otherwise):
- all physical modes: moving head/hands/feet stamp precise motion blobs
- Sciame: the swarm follows your head
- Interattivo Modello: hand swipes spin the robot, it leans toward your
head, raising a hand fast makes it jump
- Silhouette: neon skeleton (bones + joints + glowing head) over the
electric outline
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New interactive mode 'model': the webcam motion drives the GLB model as
a physical puppet — a hand swipe spins it with inertia, it leans toward
the motion centroid, strong agitation makes it hop with squash & stretch
and a touch-glow rim. ModelSim.render accepts an optional pose (yaw,
leanX, hopY, squash, rim) and slows its auto-spin when driven live.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
interactive.js becomes a multi-mode engine sharing camera, motion grid,
gradient/centroid tracking and the mouse fallback. New renderers: point/line
sprites (mirror/swarm/cloth/pops) and a ping-pong trail FBO (painting).
fluid.js gains a 'touch' emitter mode fed by motion splats from the camera.
Mirror mode reforms the custom SVG/image if loaded, else a DJ LUZA wordmark.
UNPACK_ALIGNMENT set to 1 for non-multiple-of-4 single-channel uploads.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New 'Interattivo Palle' family (interactive.js): frame-differencing motion map
from the webcam, ball physics (push-away forces, collisions, beat kicks),
palette-tinted rendering with motion mist. Camera is requested lazily and
released when switching preset; mouse/touch fallback when unavailable.
Excluded from Auto VJ so the camera never turns on by itself.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
