Research map / Image generation / Autoregressive and latent generation
Style Transfer for Co-Speech Gesture Animation: A Multi-Speaker Conditional-Mixture Approach
Chaitanya Ahuja, Dong Won Lee, Yukiko I. Nakano, Louis-Philippe Morency
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add mixture of sub-generators (M>1) (ablation) | StAGE, 8 speakers · PATS style preservation (8 speakers, mean) | Mode classification F1: 0.32 -> 0.33 | Improved |
| add mixture of sub-generators (M>1) (ablation) | StAGE, 4 speakers · PATS style preservation (4 speakers, mean) | Mode classification F1: 0.34 -> 0.35 | Improved |
| add mixture of sub-generators (M>1) (ablation) | StAGE, 2 speakers · PATS style preservation (2 speakers, mean) | PCK: 0.36 -> 0.34 | No clear effect |
| add mixture of sub-generators (ablation) | StAGE, 8 speakers · PATS style transfer (8 speakers) | Inception score: 3.89 -> 4.48 | Improved |
| add mixture of sub-generators (ablation) | StAGE, 2 speakers · PATS style transfer (2 speakers) | Inception score: 2.17 -> 2.61 | Improved |
| add mixture of sub-generators to single speaker model (M=1 to M=8) (ablation) | CMix-GAN · PATS single speaker | Mode classification F1: 26.6 -> 28.0 | Improved |
| increase number of sub-generators from 8 to 12 (ablation) | CMix-GAN · PATS single speaker | Mode classification F1: 28.0 -> 27.8 | No clear effect |
| add multiple speakers in training (single-speaker S2G to multi-speaker Mix-StAGE, 8 speakers) | S2G · PATS style preservation (8 speakers, mean) | PCK: 0.36 -> 0.40 | Improved |
| add style embedding and consistency loss with multi-speaker training (8 speakers) (ablation) | CMix-GAN · PATS style preservation (8 speakers, mean) | PCK: 0.37 -> 0.40 | Improved |
| use Mix-StAGE instead of MUNIT | MUNIT · PATS style transfer (8 speakers) | Inception score: 2.06 -> 4.48 | Improved |
| increase number of training speakers from 2 to 8 | Mix-StAGE · PATS style transfer | Inception score: 2.61 -> 4.48 | Improved |
| increase number of training speakers from 2 to 8 for style preservation | Mix-StAGE, individual speakers · PATS style preservation | per-speaker PCK/F1 | Mixed |
Closest papers by meaning
- ConvoFusion: Multi-Modal Conversational Diffusion for Co-Speech Gesture Synthesis (CVPR 2024 · similarity 0.90)
- Co$^{\mathbf{3}}$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion (ICLR 2025 · similarity 0.89)
- Weakly-Supervised Emotion Transition Learning for Diverse 3D Co-speech Gesture Generation (CVPR 2024 · similarity 0.88)
- DiffSHEG: A Diffusion-Based Approach for Real-Time Speech-driven Holistic 3D Expression and Gesture Generation (CVPR 2024 · similarity 0.88)
- Emotional Speech-driven 3D Body Animation via Disentangled Latent Diffusion (CVPR 2024 · similarity 0.87)
- Co-Speech Gesture Video Generation via Motion-Decoupled Diffusion Model (CVPR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
