Papparti — talking avatar, one character in two rendering tiers

The rig is drawn as layered vector geometry. Animation runs client-side in the phone — no GPU, no server round-trip for the face. Mouth is driven from the TTS viseme stream; the audio comes from the self-hosted voice stack.
Character: one — Indian professional, 30s, business attire Tier A: detailed mesh, shading, eye + brow motion Tier B: flat vector, minimal geometry, same face

High-configuration device

Tier A WebGL2 / canvas-2D · 50-60 fps · ≈₹12,000+ phones · 6-9 MB face asset

Idle

eyes open · mouth closed
head bob ±1.2° · blink every 4-6s
asset 6-9 MB · 50-60 fps

Talking

बुकिंग कन्फ़र्म हो गई — प्लंबर एक घंटे में पहुँचेगा।
mouth open, viseme A · brows raised
jaw dropped 8px · head tilt -1.6°
lip-sync from TTS visemes

Low-configuration device

Tier B canvas-2D / SVG · 30 fps · ₹8,000-13,000 phones · ≈0.4-0.8 MB face asset

Idle

eyes open · mouth closed
head bob none · blink every 6-8s
asset 0.4-0.8 MB · 30 fps

Talking

बुकिंग कन्फ़र्म हो गई — प्लंबर एक घंटे में पहुँचेगा।
mouth open, viseme A · brows raised
jaw flat, no shading · head fixed
lip-sync from TTS visemes