STEER: steerable 3D head avatar motion prior for two-person conversation, code released
rsasaki0109 · x · 2026-09-17
STEER (SIGGRAPH Asia 2026) released inference code. It is a causal flow-matching motion prior generating a target person's 3D facial motion (61-D FLAME state: expression, jaw, neck, eye rotation, eyelids) in dyadic conversation, conditioned on the partner's motion/audio and the target's own audio, with explicit semantic control over gaze, head rhythm, and emotion. The repo includes the prior + autoencoder pipeline and FLAME visualization, outputting side-by-side comparison videos.
More from Multimodal
- Pruna launches P-Video-2-Pro on MiniMax H3: 5s clip in ~2s, from $0.02/s — umesh_ai · 2026-09-18
- Parallel's COLONY: prompt-to-3D pipeline ships 2,000+ player-crafted assets in 30 days — templecrash · 2026-09-18
- a16z partner Justine Moore to join fal's GMC 2026 in San Francisco — jfischoff · 2026-09-18
- Gemini 4 builds Airbus H145 helicopter 3D model in ~10 minutes, skeptics unmoved — teortaxesTex · 2026-09-18
- Runway's Model Router auto-picks the best model per generation by cost, quality or latency — tlakomy · 2026-09-18
- AI photo-edit prompt template for celebration shots that keeps your real face and body — aitrendz_xyz · 2026-09-18