Research map / Video / Video generation
GestureLSM: Latent Shortcut based Co-Speech Gesture Generation with Spatial-Temporal Modeling
Pinxin Liu, Luchuan Song, Junhua Huang, Haiyang Liu, Chenliang Xu
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 spatial attention across body regions to temporal attention (ablation) | GestureLSM · BEAT2 | FGD: 8.232 -> 4.088 | Improved |
| add temporal attention to spatial attention (ablation) | GestureLSM · BEAT2 | FGD: 22.412 -> 4.088 | Improved |
| add spatial and temporal positional encodings (ablation) | GestureLSM · BEAT2 | FGD: 4.523 -> 4.088 | Improved |
| use beta(2,1.2) timestep sampling instead of uniform (ablation) | GestureLSM · BEAT2 | FGD: 5.051 -> 4.088 | Improved |
| use beta sampling instead of Logit-Normal, Mode or CosMap (ablation) | GestureLSM · BEAT2 | FGD: 4.447 -> 4.088 | Improved |
| use latent shortcut instead of standard rectified flow (ablation) | GestureLSM · BEAT2 | FGD: 4.724 -> 4.088 | Improved |
| use latent shortcut flow matching instead of DDPM diffusion (ablation) | GestureLSM · BEAT2 | AIST (s): 2.942 -> 0.039 | Improved |
| add flattened spatial-temporal attention layer on top of separate attentions (ablation) | GestureLSM · BEAT2 | FGD: 4.088 -> 4.132 | No clear effect |
| replace spatial-temporal attention with Bi-Mamba (ablation) | GestureLSM · BEAT2 | FGD: 4.088 -> 4.323 | Hurt |
| increase sampling steps from 8 to 20 (ablation) | GestureLSM · BEAT2 | FGD: 4.088 -> 4.040 | No clear effect |
| reduce sampling steps from 8 to 2 (ablation) | GestureLSM · BEAT2 | FGD: 4.088 -> 4.988 | Hurt |
| use cross-attention fusion instead of concatenation or addition (ablation) | GestureLSM · BEAT2 | FGD: 6.012 -> 4.088 | Improved |
Later papers that compared with this method
5 later accepted papers report GestureLSM as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- HolisticSemGes: Semantic Grounding of Holistic Co-Speech Gesture Generation with Contrastive Flow-Matching (ECCV 2026 · better on 6 of 6 shared results)
- ReactDance: Hierarchical Representation for High-Fidelity and Coherent Long-Form Reactive Dance Generation (ICLR 2026 · better on 4 of 4 shared results)
- LiveGesture: Streamable Co-Speech Gesture Generation Model (CVPR 2026 · better on 1 of 3 shared results)
- Past as State, Present as Attention: Persistent-State Blockwise Flow Matching for Long-Horizon Co-Speech Motion Generation (NeurIPS 2026 · better on 1 of 1 shared result)
- StreamTalk: Streaming Co-Speech Gesture Generation with Key-Pose Anchoring (ECCV 2026 · better on 1 of 1 shared result)
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.
- MambaTalk — MambaTalk: Efficient Holistic Gesture Synthesis with Selective State Space Models (NeurIPS 2024) · better on 1 of 1 shared result
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)
- DiffSHEG: A Diffusion-Based Approach for Real-Time Speech-driven Holistic 3D Expression and Gesture Generation (CVPR 2024 · similarity 0.89)
- EMAGE: Towards Unified Holistic Co-Speech Gesture Generation via Expressive Masked Audio Gesture Modeling (CVPR 2024 · similarity 0.89)
- Co-Speech Gesture Video Generation via Motion-Decoupled Diffusion Model (CVPR 2024 · similarity 0.89)
- DIDiffGes: Decoupled Semi-Implicit Diffusion Models for Real-time Gesture Generation from Speech (AAAI 2025 · similarity 0.88)
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.
