SKETCH EXPRESSION STUDIO

rig: Dean · hardhat v1 LOD — ← avatar lab
STATE
DEMO D SKETCHPAD — the agent draws while it teaches
INPUT & SEED
0
Deterministic replay: same seed + same demo ⇒ identical performance. Mic grants an analysis-only tap — audio never leaves the page. $0/min: no model calls anywhere on this surface.
FACE — manual controls
SKETCH STYLE
HEAD & BODY
📷 CAST A NEW FACE (upload)
Upload a straight-on portrait (face + shoulders, ≥1024×1024, square is best). B&W ink: clean line-art like Dean/Miles. Color: color art — the palette is sampled from your image during casting. The image is prepared here ($0, stays in your browser), previewed behind the current rig (onion), and saved as a named cast; the casting pipeline then builds the animated rig from it.
ABOUT THIS SURFACE
Semantic stroke-graph rig (206 landmarks / 68 paths / 40 morphs) traced from the WCIS Dean portrait; animated by the v3 performance runtime (word-anchored plans, 11-viseme coarticulation, critically-damped springs, expression budget, procedural life) — LLM never in the frame loop. Demos are deterministic fixtures built from text by a grapheme→viseme synthesizer; they exercise the real event architecture. Docs: reports/SKETCH-STUDIO-20260715.md · skill: avatar-live-chat.