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.
Figures & tables
Fig. 1: The overview of the proposed T2M Mamba. (a) T2M Mamba. Our T2M Mamba consisting of N basic blocks aims to predict clean motion sequence (b) Inference Process. Starting from Gaussian noise, the model iteratively denoises to generate a clean motion sequence M0 semantically aligned with the input text prompt.
Fig. 3: Illustration of the proposed PDCAM pipeline.
Methods
R Precision ↑
FID ↓
MM Dist ↓
Diversity →
MModality ↑
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 [ 2 ]
0.457 ±.002
0.639 ±.003
0.740 ±.003
1.067 ±.002
3.340 ±.008
9.188 ±.002
2.090 ±.083
MDM [ 4 ]
0.320 ±.005
0.498 ±.004
0.611 ±.007
0.544 ±.044
5.566 ±.027
9.559 ±.086
2.799 ±.072
MotionDiffuse [ 3 ]
0.491 ±.001
0.681 ±.001
0.782 ±.001
0.630 ±.001
3.113 ±.001
9.410 ±.049
1.553 ±.042
MLD [ 16 ]
0.481 ±.003
0.673 ±.003
0.772 ±.002
0.473 ±.013
3.196 ±.010
9.724 ±.082
2.413 ±.079
TABLE I: Comparison of results on the HumanML3D and KIT-ML test set. 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. 4: Qualitative comparison of prominent state-of-the-art methods.
Fig. 5: Visual comparison of T2M Mamba under subtle semantic variations
R-Top1 ↑
R-Top3 ↑
FID ↓
MM Dist ↓
w/o M
0.494
0.784
0.088
3.122
w/o ϕ
0.497
0.782
0.112
3.119
w/o M &ϕ
0.496
0.785
0.108
3.112
CroAtten
0.456
0.755
0.131
3.240
PDCAM(Ours)
0.506
0.793
0.068
3.034
w/o M
0.494
0.781
0.091
3.125
TABLE II: Ablation study on T2M Mamba’s core modules on HumanML3D. Upper: PDCAM variants; Lower: Periodicity-Saliency Aware Mamba ablations (w/o M: no keyframe weighting; w/o ϕ : no motion periodicity phase; w/o M &ϕ : neither; CroAtten: PDCAM replaced by standard cross-attention). Bold values highlight best performance by full models.
Layer Number N
R-Top1 ↑
R-Top3 ↑
FID ↓
MM Dist ↓
4
0.498
0.791
0.089
3.074
5
0.497
0.789
0.095
3.081
6
0.506
0.793
0.068
3.034
7
0.497
0.783
0.091
3.101
8
0.497
0.779
0.088
3.126
10
0.494
0.782
0.137
3.113
TABLE III: The effect of Layer Number on T2M Mamba Performance on HumanML3D. Bold values highlight the best performance achieved by our full models.
#Params
AITS ↓
R-Top3 ↑
FID ↓
w/o ϕ
24.6M
0.41
0.782
0.098
w/o M
24.6M
0.41
0.785
0.134
w/o M &ϕ
24.6M
0.40
0.775
0.139
w/ M &ϕ (Ours)
24.6M
0.41
0.793
0.068
TABLE IV: Ablation study on keyframe weighting ( M ) and motion periodicity phase ( ϕ ) in T2M Mamba. AITS is the inference time and bold values highlight the best performance.
Text-driven human motion generation aims to synthesize realistic motion sequences that follow textual descriptions. Despite recent advances, accurately aligning motion dynamics with textual semantics remains a fundamental challenge. In this paper, we revisit text-to-motion generation from the perspective of motion-language alignment and propose MLA-Gen, a framework that integrates global motion priors with fine-grained local conditioning. This design enables the model to capture common motion patterns, while establishing detailed alignment between texts and motions. Furthermore, we identify a previously overlooked attention sink phenomenon in human motion generation, where attention disproportionately concentrates on the start text token, limiting the utilization of informative textual cues and leading to degraded semantic grounding. To analyze this issue, we introduce SinkRatio, a metric for measuring attention concentration, and develop alignment-aware masking and control strategies to regulate attention during generation. Extensive experiments demonstrate that our approach consistently improves both motion quality and motion-language alignment over strong baselines.
Ruxi Gu, Zilei Wang, Wei Wang
Department of Automation, University of Science and Technology of China, Hefei, China · State Key Laboratory of General Artificial Intelligence, BIGAI, Beijing, China
Text-to-Motion (T2M) generation aims to synthesize realistic human motion sequences from natural language descriptions. While two-stage frameworks leveraging discrete motion representations have advanced T2M research, they often neglect cross-sequence temporal consistency, i.e., the shared temporal structures present across different instances of the same action. This leads to semantic misalignments and physically implausible motions. To address this limitation, we propose TCA-T2M, a framework for temporal consistency-aware T2M generation. Our approach introduces a temporal consistency-aware spatial VQ-VAE (TCaS-VQ-VAE) for cross-sequence temporal alignment, coupled with a masked motion transformer for text-conditioned motion generation. Additionally, a kinematic constraint block mitigates discretization artifacts to ensure physical plausibility. Experiments on HumanML3D and KIT-ML benchmarks demonstrate that TCA-T2M achieves state-of-the-art performance, highlighting the importance of temporal consistency in robust and coherent T2M generation.
Hongsong Wang, Wenjing Yan, Qiuxia Lai +1
School of Computer Science and Engineering, Southeast University, Nanjing, China. · Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University), Ministry of Education, China. · State Key Laboratory of Media Convergence and Communication, Communication University of China, Beijing, China.
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