Posecode

Based on the available evidence, Posecode looks early but credible rather than broadly adopted. GitHub stars and recent X reposts show attention, but that is only proof of interest, not proof that it is already reliable in production. Stronger evidence comes from the official repo plus the creator’s implementation notes around Posecode 0.2, including turn timing, left/right foot locks, and embed preparation. That suggests a real tool being iterated in public. The best analogy is an LLM-writable motion DSL plus a browser renderer, not a general AI video generator and not a full clinical rehab platform.

In practical terms, the supported claim is consistent across sources: Posecode lets a model or developer describe movement in a small language, then parse it, enforce range-of-motion constraints, and render an animated 3D figure in the browser with Three.js. That makes sense for rehab instruction demos, posture education, yoga/training visualization, or chatbot/app interfaces that need a moving body guide instead of plain text. One community post reports converting PoseCode into a Second Life animation and getting it to work, which is limited but useful hands-on evidence.