FlashMotion: Ultra-Fast Long-Term 3D Motion Generation
Authors: Kaixing Yang, Xulong Tang, Ronghui Li, Xiangyue Zhang, Ziqiao Peng, Mingshuang Luo, Chenxi Zhao, Haoyu Wu, Ju Fan, Hongyan Liu, Puwei Wang, Zijiao Zeng, Jun He
Organizations: Renmin University of China, Beijing, China · The University of Texas at Dallas · Tsinghua University, Beijing, China · Wuhan University, Wuhan, China · Tencent, China · Nankai University, Tianjin, China · The University of Hong Kong, Hong Kong, China
3D human motion generation plays a crucial role in game development, extended reality (XR), digital humans, and embodied simulation. Beyond motion generation, practical pipelines include rigging, retargeting, and rendering, making ultra-fast long-term 3D motion generation without compromising quality valuable for real-time interactive applications. Despite promising progress, high-quality long-horizon motion generation under tight inference budgets remains challenging, owing to 2 coupled bottlenecks: (1) generative strategies that require many sampling steps for high-quality 3D motion generation and (2) model architectures with substantial per-step computational and memory costs for long sequences. We present FlashMotion, a co-designed framework for high-quality long-horizon 3D motion generation under a strict inference budget, covering text-to-motion, music-to-dance, and speech-to-gesture tasks. To enhance trajectory-level efficiency, we introduce a 3D motion-space MeanFlow formulation with kinematic and dynamic consistency constraints for few-step generation, together with an Adaptive Soft masking mechanism for temporally coherent motion editing. To improve per-step efficiency, we design a lightweight non-autoregressive BiMamba architecture with condition-guided state selection and AdaIN-based condition fusion mechanism, enhancing cross-modal alignment and generalization to longer motion sequences. Extensive experiments demonstrate that FlashMotion achieves overall state-of-the-art performance in both quality and efficiency across text-to-motion, speech-to-gesture, and music-to-dance generation. The project page is available at https://sun-happy-ykx.github.io/FlashMotion-Project/
Figures & tables
Fig. 1: FlashMotion generates physically plausible, semantically expressive motion at leading throughput with a compact ∼ 60M-parameter model. Top: quality–throughput comparisons with representative baselines; bottom: conditioned motion examples. Panels (a)–(c) use BABEL, BEAT2, and FineDance. FPS is measured for 10,000-frame outputs.
Fig. 2: Overview of FlashMotion . (a) Inference flow with clean-motion prediction and MeanFlow updates. (b) Training model inputs and output: noisy motion, conditions, and timesteps are mapped to the clean-motion target. (c) Shared BiMamba/CoBiMamba block. Dashed purple paths denote optional conditioning, omitted in BiMamba and used in CoBiMamba for condition-guided state selection.
TABLE I: Quantitative comparison on the BABEL [ 63 ] dataset for long-sequence text-to-motion generation. Bold and underlined values denote the best and second-best generated results, excluding ground truth.
Fig. 3: Text-to-motion comparison under a shared sequence of 3 prompts.
Figure 5
Method
MQ ↑
MS ↑
MC ↑
T2M
DART
3.45±0.90
3.12±0.98
3.65±0.82
FloodDiffusion
3.62±0.89
3.35±0.99
3.85±0.86
FlowMDM
3.62±0.86
3.36±0.95
3.50±0.87
FlashMotion
3.78±0.86
3.60±0.93
3.86±0.80
S2G
CaMN
3.33±0.89
3.02±1.00
3.77±0.81
EMAGE
3.42±0.92
2.87±0.98
3.98±0.84
TABLE IV: User study ratings (mean ± SD; 1–5, higher is better).
Loss setting
R@1 ↑
R@2 ↑
R@3 ↑
MM ↓
FID ↓
Δ Div ↓
Δ PJ ↓
ΔAUJ↓
w/o Dynamic
27.393
41.426
50.905
4.324
0.928
0.360
0.00097
0.01241
w/o Kinematic
28.613
42.530
51.813
4.386
0.983
0.090
0.00201
0.06503
MeanFlow only
NaN
NaN
NaN
NaN
NaN
NaN
NaN
NaN
w/o MeanFlow
27.460
40.683
49.149
4.379
1.074
0.248
0.00995
0.16019
w/ Uncond
25.456
38.893
47.860
4.516
2.232
0.311
0.00241
0.06472
Full Model
29.134
42.829
51.683
4.340
0.907
0.210
0.00027
0.01295
TABLE VII: Training objective ablation on BABEL (4-step).
Figure 8
Fig. 7: With lightweight post-processing, FlashMotion supports various downstream applications, including (a) robot motion generation, (b) XR character animation, and (c) mobile motion interaction. For clarity, we showcase text-to-motion as a representative task.
Music-to-dance generation aims to translate auditory signals into expressive human motion, with broad applications in virtual reality, choreography, and digital entertainment. Despite promising progress, the limited generation efficiency of existing methods leaves insufficient computational headroom for high-fidelity 3D rendering, thereby constraining the expressiveness of 3D characters during real-world applications. Thus, we propose FlowerDance, which not only generates refined motion with physical plausibility and artistic expressiveness, but also achieves significant generation efficiency on inference speed and memory utilization. Specifically, FlowerDance combines MeanFlow with Physical Consistency Constraints, which enables high-quality motion generation with only a few sampling steps. Moreover, FlowerDance leverages a simple but efficient model architecture with BiMamba-based backbone and Channel-Level Cross-Modal Fusion, which generates dance with efficient non-autoregressive manner. Meanwhile, FlowerDance supports motion editing, enabling users to interactively refine dance sequences. Extensive experiments on AIST++ and FineDance show that FlowerDance achieves state-of-the-art results in both motion quality and generation efficiency. Code will be released upon acceptance.
Kaixing Yang, Xulong Tang, Ziqiao Peng +4
Renmin University of China · Malou Tech Inc · Wuhan University +1
Generating realistic 3D human motions in real-time within interactive applications is key for animation, simulation, and humanoid robotics. While recent offline motion generation approaches offer precise control via text and kinematic constraints, they lack the inference speed required for interactive settings. Conversely, existing online methods enable real-time synthesis but often sacrifice controllability or struggle with complex text semantics and long-horizon goals due to limited context windows. In this work, we introduce ARDY, a streaming generation framework that bridges this gap by enabling high-fidelity motion generation controllable via online text prompts and flexible kinematic constraints. ARDY employs a hybrid representation that combines explicit root features with a latent body embedding, balancing precise trajectory control with efficient generative learning. We propose a two-stage autoregressive transformer denoiser that features variable history context and supports conditioning on flexible, long-horizon kinematic constraints. By training on a large-scale motion capture dataset and being directly conditioned on text labels and kinematic constraints sampled from ground truth poses, ARDY natively learns controllable generation that supports online prompting and flexible long-horizon goals. Extensive evaluations on the HumanML3D benchmark and the large-scale, high-fidelity Bones Rigplay dataset demonstrate ARDY's high motion quality and constraint adherence, validating the efficacy of our key architectural decisions. Finally, we demonstrate the method's practical versatility through an interactive demo featuring dynamic text control, diverse keyframe pose constraints, path following, and interactive locomotion control via mouse and keyboard. Supplementary video results, code, and model releases can be found at https://research.nvidia.com/labs/sil/projects/ardy/.
Kaifeng Zhao, Mathis Petrovich, Haotian Zhang +3
NVIDIA, Switzerland · ETH Zürich, Switzerland · NVIDIA, USA
We present World2Motion, a framework that generates scene-aware 3D human motion and corresponding video from a single image and a text prompt. While existing 3D motion generators learn from motion datasets, their generalization is constrained by limited coverage of environments. In contrast, video world models such as Cosmos 3 offer broader environmental priors but are not designed for full-body motion generation; recovering motion from their generated videos requires costly two-stage inference. To address these, we turn Cosmos 3 into a single-stage 3D motion generator. This adaptation has two challenges: the scarcity of paired video--motion data and temporal instability in the generated motion. First, we construct a training dataset combining synthetic video--motion pairs with real videos paired with estimated 3D motion. Second, we propose a shift-decoupled noise schedule that assigns different noise levels to video and motion through shared denoising progress. This design accommodates the different denoising requirements of the two modalities, reducing motion jitter. Experiments on a multi-source interaction benchmark show that World2Motion has better motion--text alignment and scene interaction compared with the evaluated 3D motion generators. It also matches the interaction success rate of the two-stage baseline while achieving approximately 3.3× faster inference. Our project page is available at https://fyantu.github.io/World2Motion/.
Fangyuan Tu, Xiangyue Zhang, Yiyi Cai +8
Japan Advanced Institute of Science and Technology · The University of Tokyo · Institute of Science Tokyo +1