Generating realistic co-speech gestures for virtual avatars in dyadic conversations is a difficult task, according to a new arXiv paper. The abstract notes that existing models often fail to strike a balance between kinematic stability—smooth, physically plausible motion—and rhythmic responsiveness, or the ability to align gestures with the timing and flow of speech.
The paper, titled "SynerGest: Temporal-Spatial Synergistic Gesture Generation for Dyadic Conversations," proposes a method called SynerGest to address this trade-off. The abstract suggests the approach is designed specifically for two-person conversational settings, where gesture generation must account for both speakers' dynamics.
Because the abstract is truncated, the specific techniques behind SynerGest are not described in the available text. The paper is listed on arXiv under identifier 2610.24031, and the announcement type is "new," indicating a recent submission. Further details would require reading the full manuscript.