Research map / 3D vision / 3D generation
MIBURI: Towards Expressive Interactive Gesture Synthesis
M. Hamza Mughal, Rishabh Dabral, Vera Demberg, Christian Theobalt
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 |
|---|---|---|---|
| use Moshi internal speech-text tokens instead of wav2vec features (ablation) | MIBURI(+Face) · BEAT2 multi-speaker | FGD: 0.665 -> 0.480 | Improved |
| split into temporal and kinematic transformers instead of a single transformer (ablation) | MIBURI · BEAT2 multi-speaker | FGD: 1.256 -> 0.480; step time 0.096 -> 0.035 | Improved |
| add contrastive InfoNCE loss on predicted latents (ablation) | MIBURI · BEAT2 multi-speaker | FGD: 0.499 -> 0.480 | Improved |
| add MSE loss on estimated latents (ablation) | MIBURI · BEAT2 multi-speaker | FGD: 0.499 -> 0.577 | Hurt |
| increase RVQ residual levels from K=1 to K=8 (ablation) | gesture codec · motion reconstruction | MPJPE (m): 0.043 -> 0.016 | Improved |
| increase RVQ residual levels from K=1 to K=8 (ablation) | gesture codec · motion reconstruction | FGD: 0.55 -> 0.059 | Improved |
| use larger MIBURI-L model (ablation) | MIBURI(+Face) · BEAT2 multi-speaker | FGD: 0.480 -> 0.582 | No clear effect |
| train causal versions of GestureLSM and MambaTalk | GestureLSM · BEAT2 multi-speaker | FGD: 0.537 -> 2.792 | Hurt |
| train on more speakers (23 vs single speaker) | MIBURI · BEAT2 | FGD: 0.753 -> 0.480 (single-speaker to multi-speaker, +Face) | Improved |
| add facial expression generation | MIBURI · BEAT2 multi-speaker | FGD: 0.585 -> 0.480 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- EMAGE* — EMAGE: Towards Unified Holistic Co-Speech Gesture Generation via Expressive Masked Audio Gesture Modeling (CVPR 2024) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Co$^{\mathbf{3}}$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion (ICLR 2025 · similarity 0.90)
- ConvoFusion: Multi-Modal Conversational Diffusion for Co-Speech Gesture Synthesis (CVPR 2024 · similarity 0.89)
- Chain of Generation: Multi-Modal Gesture Synthesis via Cascaded Conditional Control (AAAI 2024 · similarity 0.88)
- The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion (CVPR 2025 · similarity 0.88)
- Emotional Speech-driven 3D Body Animation via Disentangled Latent Diffusion (CVPR 2024 · similarity 0.87)
- EMAGE: Towards Unified Holistic Co-Speech Gesture Generation via Expressive Masked Audio Gesture Modeling (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.
