Research map / Vision-language and multimodal / Human and embodied modeling
Text-driven human motion generation
Generates human motion and interactions from text using language-aligned motion models.
36 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2023–2026) · an approach in Human and embodied modeling, keyword cluster “pre · generation · unified”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
10 papers in 2023–24 and 26 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 1.9% to 3.9%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2023 · MotionGPT: Human Motion as a Foreign Language (NeurIPS 2023 · 55 citations)
- 2024 · ConvoFusion: Multi-Modal Conversational Diffusion for Co-Speech Gesture Synthesis (CVPR 2024 · 27 citations)
- 2025 · UniMuMo: Unified Text, Music, and Motion Generation (AAAI 2025 · 8 citations)
- 2026 · M3ashy: Multi-Modal Material Synthesis via Hyperdiffusion (AAAI 2026 · 1 citation)
Earliest papers here
- MotionGPT: Human Motion as a Foreign Language (NeurIPS 2023 · 55 citations)
- ConvoFusion: Multi-Modal Conversational Diffusion for Co-Speech Gesture Synthesis (CVPR 2024 · 27 citations)
- MoST: Multi-Modality Scene Tokenization for Motion Prediction (CVPR 2024 · 11 citations)
What has been tried
24 papers here report 239 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 3%, hurt 26%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add word-excitation guidance | BEAT · Semantic gesture recall | 0.34 -> 0.40 | Helped | ConvoFusion: Multi-Modal Conversational Diffusion for Co-Speech Gesture Synthesis (CVPR 2024 · 27 citations) |
| Use SAM ViT-H image features instead of other tested image encoders | WOMD validation · minADE | 0.5670 -> 0.5483 | Helped | MoST: Multi-Modality Scene Tokenization for Motion Prediction (CVPR 2024 · 11 citations) |
| Use pre-trained ViT weights rather than training from scratch | HumanML3D, text-to-motion · R@1 | 8.46 -> 10.80 | Helped | Exploring Vision Transformers for 3D Human Motion-Language Models with Motion Patches (CVPR 2024 · 9 citations) |
| use skeleton-aware pre-training | Mixamo test motions · MSE | 7.798 -> 0.284 | Helped | Semantics-aware Motion Retargeting with Vision-Language Models (CVPR 2024 · 9 citations) |
| Use an independent motion VQ-VAE without initializing the motion embedder from MusicGen, instead of shared music codebooks | MusicCaps · FAD_VGG | 5.93 -> 6.75 | Hurt | UniMuMo: Unified Text, Music, and Motion Generation (AAAI 2025 · 8 citations) |
| add the face generator and temporal facial feature decoder | BEAT test ablation · FGD | 88.50 -> 60.44 | Helped | Chain of Generation: Multi-Modal Gesture Synthesis via Cascaded Conditional Control (AAAI 2024 · 8 citations) |
| Add textual adaptation with P* to visual adaptation | PerMo, Single Input · FID | 7.45 -> 5.06 | Helped | PersonaBooth: Personalized Text-to-Motion Generation (CVPR 2025 · 7 citations) |
| Add language pre-training | BEATv2 speaker-2 · FGD ×10−1 | 7.470 -> 5.301 | Helped | The Language of Motion: Unifying Verbal and Non-verbal Language of 3D Human Motion (CVPR 2025 · 7 citations) |
| replace clustered intention points with route-sampled intention points | Private validation · Plan ADE | 0.4788 -> {0.4141} | Helped | Int2Planner: An Intention-based Multi-modal Motion Planner for Integrated Prediction and Planning (AAAI 2025 · 5 citations) |
| Add the Prior Motion Distillation Decoder | KIT-ML · FID | 4.258 -> 1.060 | Helped | Everything2Motion: Synchronizing Diverse Inputs via a Unified Framework for Human Motion Synthesis (AAAI 2024 · 5 citations) |
| use contact vertices as actions instead of no action | Text-to-Interaction · CMD | 0.424 -> 0.151 | Helped | InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction (NeurIPS 2024 · 5 citations) |
| Replace the HOI-decomposed VQ-VAE with separate hand and object VQ-VAEs | HOI tokenizer reconstruction · FID | 0.43 -> 0.87 | Hurt | HOIGPT: Learning Long-Sequence Hand-Object Interaction with Language Models (CVPR 2025 · 3 citations) |
What reviewers will expect
Datasets and baselines named in the 18 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- HumanML3D · 7 of 18 papers
- nuPlan · 2 of 18 papers
- ARCTIC · 2 of 18 papers
Baselines
- MotionGPT · 6 of 18 papers
- T2M-GPT · 3 of 18 papers
- TM2T · 3 of 18 papers
- MotionDiffuse · 2 of 18 papers
- MoMask · 2 of 18 papers
- BLIP-Diffusion · 2 of 18 papers
- KOSMOS-G · 2 of 18 papers
- Emu2-Gen · 2 of 18 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 36 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 4 papers
- ShanghaiTech University · 2 papers
- Fudan University · 2 papers
- Chinese Academy of Sciences · 2 papers
- Peking University · 2 papers
Use it
Download BibTeX of 21 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Vision-language pretraining (182 papers)
- Medical vision-language models (130 papers)
- Layout-aware document understanding (73 papers)
Elsewhere in the topic
- Scene graph generation (126 papers)
- Language-grounded scene text and structure (124 papers)
- fMRI visual decoding (106 papers)
- Human-like visual cognition in VLMs (97 papers)
- Multimodal and affective interaction modeling (87 papers)
- Multimodal retrieval-augmented modeling (85 papers)
- Language-grounded visual concept learning (83 papers)
- Composed and text-based image retrieval (78 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).