Reproduce “CFGComposer: Compositional Audio-Text-to-Motion Generation via Copula Score Correction” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Pengyang Shi, Jinlu Zhang, Chunze Lin, Yizhou Wang, Li Yi

CFGComposer composes speech-to-motion and text-to-motion diffusion models to generate gestures that follow speech rhythm and a gesture description. It combines audio and text experts with a jointly conditioned model trained on paired audio, text, and motion; its prediction estimates the missing interaction. On BEAT, it reaches SRGR 0.9916 at λ=0.7, within the reported dataset and evaluation protocol.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.