Text-to-motion generation is a rapidly growing field at the nexus of multimodal learning and computer graphics, enabling flexible and cost-effective applications in gaming, animation, robotics, and virtual reality. Existing part-aware pipelines typically split the body into segments and then recover whole-body coordination via attention-based fusion, which incurs quadratic complexity and fails to model part-level and whole-body cues simultaneously. Moreover, we observe that naively stacking spatial and temporal modules accumulates redundant features, thereby requiring deeper networks to achieve competitive fidelity. To address these issues, we propose HiSTF Mamba, a linear-time framework composed of Dual-Spatial Mamba, Bi-Temporal Mamba, and the Dynamic Spatiotemporal Fusion Module (DSFM). Dual-Spatial Mamba performs parallel part-based and whole-based modeling to simultaneously capture local joint dynamics and global coordination. Bi-Temporal Mamba employs bidirectional scanning to encode both short-term motion details and long-term dependencies. DSFM further removes redundancy and extracts complementary temporal information before fusing it with spatial features, yielding an expressive spatiotemporal representation. Extensive experiments on HumanML3D and KIT-ML demonstrate that HiSTF Mamba improves generation fidelity and semantic alignment while requiring substantially fewer parameters than existing segment-based architectures.
Figures & tables
Fig. 1: Qualitative visualization of motions generated by HiSTF Mamba.
Fig. 2: The overview of the proposed HiSTF Mamba. Given a text prompt and diffusion timestep, the model iteratively denoises from Gaussian noise to synthesize a motion sequence, as illustrated by the inference process at the bottom. The backbone consists of N Bi-Temporal Mamba that encodes bidirectional temporal features {T1,…,TN} , a Dual-Spatial Mamba that captures multi-granular spatial cues via parallel part-based and whole-based scans, and a DSFM that suppresses redundant temporal information and fuses refined temporal cues with spatial features. The right panel visualizes scan order of whole SSMs and the Part SSMs.
Methods
R Precision ↑
FID ↓
MM Dist ↓
Diversity →
MModality ↑
Params (M) ↓
Top 1
Top 2
Top 3
On the HumanML3D
0.511 ±.003
0.703 ±.003
0.797 ±.002
0.002 ±.000
2.974 ±.008
9.503 ±.065
–
–
T2M [ 5 ]
0.457 ±.002
0.639 ±.003
0.740 ±.003
1.067 ±.002
3.340 ±.008
9.188 ±.002
2.090 ±.083
31.16
MDM [ 14 ]
0.320 ±.005
0.498 ±.004
0.611 ±.007
0.544 ±.044
5.566 ±.027
9.559 ±.086
2.799 ±.072
17.88
MotionDiffuse [ 18 ]
0.491 ±.001
0.681 ±.001
0.782 ±.001
0.630 ±.001
3.113 ±.001
9.410 ±.049
1.553 ±.042
87.15
MLD [ 2 ]
0.481 ±.003
0.673 ±.003
0.772 ±.002
0.473 ±.013
3.196 ±.010
9.724 ±.082
2.413 ±.079
26.38
TABLE I: Comparison of results on the HumanML3D and KIT-ML test sets. The right arrow ( → ) indicates that values closer to real motions correspond to better performance. Each evaluation was run 20 times, and “ ± ” denotes the 95% confidence interval. Bold text highlights the best scores, while underlined entries mark the second best.
Fig. 3: Qualitative comparison of typical methods . We randomly selected three textual samples to generate motion sequences. Our results exhibit closer alignment with the text and present richer joint-level details.
N
L
R Precision (Top 3) ↑
FID ↓
MM Dist ↓
2
2
0.813±.002
0.069±.002
2.907±.005
3
2
0.813±.002
0.039±.002
2.884±.008
4
2
0.817±.002
0.074±.002
2.872±.007
2
3
0.814±.003
0.062±.003
2.893±.007
3
3
0.810±.002
0.054±.003
2.914±.008
4
3
0.816±.002
0.052±.002
2.873±.008
TABLE II: Ablation study on varying the number of Bi-Temporal Mamba blocks ( N ) and the number of layers ( L ), evaluated on HumanML3D.
R-Top3 ↑
FID ↓
MM Dist ↓
w/o DSFM
0.817±.002
0.096±.003
2.889±.010
w/o Part-b
0.816±.002
0.066±.003
2.869±.008
w/o Whole-b
0.812±.002
0.065±.003
2.901±.009
w/o Part-b & Whole-b
0.815±.002
0.070±.003
2.874±.008
HiSTF Mamba (Ours)
0.813±.002
0.039±.002
2.884±.008
TABLE III: Ablation study on key components, evaluated on HumanML3D. We use Part-b and Whole-b to denote the part-based learning and whole-based learning in Dual-Spatial Mamba , respectively.
Text-to-motion generation is driven by learning motion representations for semantic alignment with language. Existing methods rely on either continuous or discrete motion representations. However, continuous representations entangle semantics with dynamics, while discrete representations lose fine-grained motion details. In this context, we propose FlowCoMotion, a novel motion generation framework that unifies both treatments from a modeling perspective. Specifically, FlowCoMotion employs token-latent coupling to capture both semantic content and high-fidelity motion details. In the latent branch, we apply multi-view distillation to regularize the continuous latent space, while in the token branch we use discrete temporal resolution quantization to extract high-level semantic cues. The motion latent is then obtained by combining the representations from the two branches through a token-latent coupling network. Subsequently, a velocity field is predicted based on the textual conditions. An ODE solver integrates this velocity field from a simple prior, thereby guiding the sample to the potential state of the target motion. Extensive experiments show that FlowCoMotion achieves competitive performance on text-to-motion benchmarks, including HumanML3D and SnapMoGen.
Dawei Guan, Di Yang, Chengjie Jin +1
School of Artificial Intelligence and Data Science, University of Science and Technology of China · Suzhou Institute for Advanced Research, University of Science and Technology of China
Text-to-motion generation, which converts motion language descriptions into coherent 3D human motion sequences, has attracted increasing attention in fields, such as avatar animation and humanoid robotic interaction. Though existing models have achieved significant fidelity, they still suffer from two core limitations: (i) They treat motion periodicity and keyframe saliency as independent factors, overlooking their coupling and causing generation drift in long sequences. (ii) They are fragile to semantically equivalent paraphrases, where minor synonym substitutions distort textual embeddings, propagating through the decoder and producing unstable or erroneous motions. In this work, we propose T2M Mamba to address these limitations by (i) proposing Periodicity-Saliency Aware Mamba, which utilizes novel algorithms for keyframe weight estimation via enhanced Density Peaks Clustering and motion periodicity estimation via FFT-accelerated autocorrelation to capture coupled dynamics with minimal computational overhead, and (ii) constructing a Periodic Differential Cross-modal Alignment Module (PDCAM) to enhance robust alignment of textual and motion embeddings. Extensive experiments on HumanML3D and KIT-ML datasets have been conducted, confirming the effectiveness of our approach, achieving an FID of 0.068 and consistent gains on all other metrics.
Xingzu Zhan, Chen Xie, Honghang Chen +2
Shenzhen University, Shenzhen, China · Jinan University, Guangzhou, China
Text-to-motion generation aims to generate 3D human motions that are tightly aligned with the input text while remaining physically plausible and rich in fine-grained detail. Although recent approaches can produce complex and natural movements, they usually operate at only one temporal scale, which limits both semantic alignment and temporal coherence. Inspired by the fact that complex motions are conceptualized hierarchically rather than at a single temporal scale in the human cognitive system, we propose MotionHiFlow, a hierarchical flow matching framework to generate motion progressively by constructing flow path from low to high temporal scales. The flows at lower scales capture high-level semantics and coarse motion structures, while flows at higher scales refine temporal details. To link the flows across scales, we introduce a novel cross-scale transition process, ensuring continuity and preserving noise consistency. Furthermore, by integrating a Text-Motion Diffusion Transformer and a topology-aware Motion VAE, MotionHiFlow explicitly models structural dependencies among joints via joint-aware positional encoding and skeletal topology, enabling precise semantic alignment alongside fine-grained motion details. Extensive experiments on HumanML3D and KIT-ML benchmarks demonstrate state-of-the-art performance, with ablation studies confirming the effectiveness of the hierarchical design and key components. Code is available at https://github.com/ai-lh/MotionHiFlow.
Heng Li, Xiaotong Lin, Ling-An Zeng +3
Sun Yat-sen University · Shandong University · Guangdong Province Key Laboratory of Information Security Technology, China +1