Tag
This paper investigates using a Transformer-based generative model to learn emotional body motions from motion-capture data of Japanese actors, generating motions conditioned on discrete emotion labels. Evaluations show the generated motions improve emotion recognition when used for data augmentation and enable smooth transitions between emotion intensities.
PhysiFormer uses coordinate-space diffusion to generate physically-plausible 3D object motions without explicit inductive biases, enabling efficient multi-object reasoning and generalization to complex materials and geometries.
This paper identifies an embodiment gap in humanoid co-speech motion generation caused by human-centric pipelines, and proposes PhysDrift, an embodiment-aware framework that directly predicts executable humanoid joint trajectories from speech, improving speech-motion alignment and physical plausibility.
Proposes MotionVLA, a vision-language-action model for humanoid motion generation using a dual-stream frequency tokenizer that separately encodes pose and physical dynamics, achieving better diversity and consistency.
VideoMDM trains 3D human motion priors from 2D poses using a diffusion framework with 2D reprojection loss and 3D motion regularizers, achieving near-3D supervised performance without requiring 3D ground truth.
GRAIL generates diverse humanoid manipulation and locomotion data using 3D assets and video foundation models, enabling effective sim-to-real transfer for humanoid robot control with high real-world success rates.