Movens-Zero: Scaling Humanoid Motion Foundation Models with Large-Scale Human Videos
Authors: Wendong Bu, Tinghong Ye, Qizhou Wang, Jiacheng Li, Zhiqi Ge, Yuze Lin, Zhongqi Yue, Yi Su, Jingyu Wang, Kai Shen, Jingdong Liu, Chengyuan Deng, Xiaoxuan Chen, Xin Tan, Jiaming Ji, Wenqiao Zhang, Chenghong Jiang, Tuoyu Li, Siliang Tang, Jun Xiao, Tat-Seng Chua, Yueting Zhuang, Xingxing Wang, Juncheng Li
Organizations: Zhejiang University · Joint Laboratory of Embodied Intelligence, ZJU & Unitree Robotics · Bit Oasis · Unitree Robotics · HiThink Research · Microsoft Research · East China Normal University · Peking University · National University of Singapore
Scaling humanoid motion foundation models is constrained by the scarcity of diverse, robot-executable data and by the difficulty of jointly learning from heterogeneous multimodal inputs while supporting real-time closed-loop control. We present Movens-5M, constructed from raw Internet videos through clip generation, quality filtering, semantic annotation, motion recovery, and physical refinement with humanoid retargeting and execution verification. It contains 15 million human-motion sequences and 5 million physically executable humanoid-motion sequences spanning over 10,000 hours, aligned with text, RGB video, speech, and music. Building on this data, we introduce Movens-Zero, a scalable understanding-generation-execution architecture that supports joint learning across modalities and real-time closed-loop execution. Task-mixed flow-matching training jointly optimizes a single action expert across seven generation tasks, enabling direct motion generation from text, speech, music, and video. Multimodal Real-Time Chunking connects overlapping chunks under continuously updated instructions and robot states for continuous closed-loop execution. We further introduce Movens-Intent, a four-track benchmark with 8,192 expert-verified condition-motion pairs across text, speech, music, and video. Experiments show consistent gains across all four modalities when scaling execution-verified data from 1% to 100% and action-expert capacity from 64M to 911M parameters, while demonstrating that data quality and cross-modal supervision improve generation and execution.
Figures & tables
Figure 1: Overview of Movens-Zero , a scalable understanding–generation–execution architecture built on Movens-5M . It translates multimodal human intent directly into physically executable humanoid behaviors in real time.
Dataset
Open
# Sequences
Hours
Frames
Diversity
Quality checks
Aligned modalities
Source
Human
Humanoid
Types
Scenes
Constraints
Policies
Text
Music
Speech
RGB
Human motion datasets
AMASS ( Mahmood et al., 2019 )
✓
11.3K
–
40.0h
≈ 4.32M
–
1
✗
✗
✗
✗
✗
✗
MoCap agg.
LAFAN1 ( Harvey et al., 2020 )
✓
77
–
4.6h
496.7K
15
1
✗
✗
✗
✗
✗
✗
MoCap
KIT-ML ( Plappert et al., 2016 )
✓
3.9K
–
11.2h
4.03M
–
1
✗
✗
✓
✗
✗
✗
MoCap agg.
HumanML3D ( Guo et al., 2022 )
✓
14.6K
–
28.6h
2.06M
–
1
✗
✗
✓
✗
✗
✗
MoCap agg.
Table 1: Comparison with representative human and humanoid motion datasets. “Types” counts published dataset-level motion categories. “Scenes” counts coarse environments. “Constraints” and “Policies” count physical-constraint and control-policy families used in curation and execution verification, with real-world collection counted as one constraint setting. “Agg.” denotes aggregation. “–” marks unavailable and inapplicable entries.
Figure 2: Overview of the Movens-5M data construction pipeline. Raw web videos are segmented and filtered, semantically annotated, and reconstructed as SMPL-family estimates canonicalized to SMPL-X. Closed-loop physical refinement, humanoid retargeting, and tracking verification produce high-quality, executable humanoid motions.
Figure 3: Overview of our data-augmentation pipeline. Text design expands motion conditions through category-preserving rewriting, long- horizon composition, and targeted extreme-motion design; Movens-Zero synthesizes the corresponding robot trajectories; and closed-loop tracking with semantic-consistency verification retains executable caption–motion pairs.
Figure 4: Statistics of Movens-5M . (a) Empirical sequence-duration density. (b) Hierarchical motion composition: the inner ring shows seven coarse classes, and the outer ring shows their two largest fine-grained types plus the remainder. (c) Composition of the 16 scene groups. (d) Motion–scene association among the five largest retained motion classes and scene groups; each cell gives the percentage of a motion class occurring in the corresponding scene group.
Figure 5: Motion-space visualization and examples. (a) Shared two-second G1 motion-space projection for eight datasets. Each colored view contains 10,000 windows; the seven comparators are random samples, while the Movens-5M display subset is grid-prioritized on frozen coordinates. (b) Fine-grained motion types in Movens-5M. (c) Corresponding source videos. See Appendix A.1 for more detailed dataset examples.
Dataset
Dataset Scale
Reference Motion Quality
Closed-loop Rollout Quality
Hours
Frames
# Humanoid Sequences
Temporal Quality
Foot–Support Consistency
Kinematic
Tracking Fidelity
Execution Validity
Jerk ↓
MSNR ↑
C-Slide ↓
PFC ↓
Non- Skating ↑
Non- Floating ↑
Non- Pen. ↑
Joint Feas. ↑
MPJPE ↓
g-MPJPE ↓
Evel↓
Eacc↓
Tracking Score ↑
Fall ↓
J-Limit ↓
MoCap-based datasets
AMASS ( Mahmood et al., 2019 )
40.0
≈ 4.32M
—
121.55
14.01
0.345
0.916
71.17
72.97
75.82
75.61
41.59
695.83
0.326
4.260
0.292
7.10
19.40
LAFAN1 ( Harvey et al., 2020 )
4.6
496.7K
—
194.32
11.65
0.356
0.798
46.63
23.88
97.55
41.27
153.28
3028.86
0.592
7.713
0.084
63.64
98.70
OmniRetarget ( Yang et al., 2025 )
4.0
≈ 432K
2.1K
204.61
13.11
0.221
0.250
80.69
66.27
100.00
100.00
59.23
443.92
0.322
5.866
0.197
13.70
32.30
Table 2: Complete-sequence dataset motion quality and G1 execution across MoCap-based and video-based datasets. Scale columns report full datasets or releases. Bold marks within-group best results. Higher MSNR means less high-frequency residual relative to the smoothed motion signal, but it should be interpreted jointly with jerk because motion bandwidth and over-smoothing also affect it. — denotes unavailable entries. Evaluation settings and dataset-specific qualifications are detailed in Appendix E.3 .
Figure 6: Examples from the Movens-Intent benchmark, illustrating human intent expressed through (a) text, (b) speech, (c) music, and (d) video. Each panel pairs a user request and its input condition with four frames from the corresponding G1 reference motion. The speech transcript makes the spoken content readable, and the waveforms show the complete audio recordings.
Figure 7: Architecture of Movens-Zero . Its understanding–generation–execution hierarchy translates multimodal human intent directly into physically executable humanoid behaviors in real time: an omni-modal backbone encodes interleaved multimodal inputs, infers human intent, and produces context tokens; a flow-based transformer generates overlapping state-conditioned motion chunks; and a low-level controller executes the resulting references in simulation and the real world.
Method
Motion Generation Quality
Tracking Fidelity and Execution Failures
FID ↓
Matching ↓
R@1 ↑
R@2 ↑
R@3 ↑
Diversity ↑
C-Slide ↓
Jerk ↓
MPJPE ↓
g-MPJPE ↓
Evel↓
Eacc↓
Fall ↓
J-Limit ↓
OMG benchmark
GENMO ( Li et al., 2025a )
43.78
1.33
12.99
23.83
30.86
1.28
2.66
173.54
50.35
150.71
6.91
2.51
1.46
29.10
HY-Motion ( Wen et al., 2025 )
24.68
1.30
31.84
48.73
57.71
1.29
12.20
112.56
69.05
287.71
11.94
3.24
4.98
55.57
TextOp ( Xie et al., 2026 )
59.91
1.30
11.52
19.43
25.00
1.15
3.80
219.67
38.00
99.08
6.56
2.62
2.15
20.70
Kimodo-SMPLX-RP ( Rempe et al., 2026 )
38.44
1.31
26.27
39.36
46.48
1.25
4.52
189.39
43.51
198.96
7.80
2.64
2.64
23.63
Table 3: Text-conditioned motion generation on the OMG and Movens-Intent benchmarks. R@K, Fall, and J-Limit are percentages. FID and C-Slide are scaled by 10−2 and 10−1 , respectively.
Method
World-space
Camera-space
WA-MPJPE 100 ↓
W-MPJPE 100 ↓
RTE ↓
Jitter ↓
Foot-Sliding ↓
PA-MPJPE ↓
MPJPE ↓
PVE ↓
Accel ↓
RICH
GENMO ( Li et al., 2025a )
75.3
118.6
1.9
15.0
6.7
39.1
66.8
75.4
4.1
WHAM ( Shin et al., 2024 )
113.0
191.7
4.3
18.6
3.7
44.3
80.0
91.2
5.3
TRAM ( Wang et al., 2024 )
103.5
168.0
2.5
18.5
12.5
44.1
70.5
78.7
5.4
GVHMR ( Shen et al., 2024 )
78.8
126.3
2.4
12.8
3.0
39.5
66.0
74.4
4.1
Table 4: Video-conditioned motion reconstruction on RICH and EMDB in SMPL space, and on Movens-Intent in G1 space.
Method
Music Alignment and Target Motion Quality
Target Physical Quality and Tracker Fidelity
BeatAlign ↑
FID-k ↓
FID-g ↓
PFC ↓
C-Slide ↓
Jerk ↓
MPJPE ↓
g-MPJPE ↓
Evel↓
Eacc↓
Fall ↓
J-Limit ↓
OMG benchmark
GENMO ( Li et al., 2025a )
0.54
207.39
17.70
0.08
0.13
185.99
39.79
58.51
3.54
1.80
0.00
2.05
LODGE ( Li et al., 2024 )
0.58
80.46
2.66
0.21
0.30
213.86
48.60
101.19
7.55
3.06
1.59
17.95
Bailando ( Siyao et al., 2022 )
0.59
328.15
16.85
0.19
0.17
103.60
40.02
80.01
4.55
1.48
0.23
7.95
Movens-Zero (Ours)
0.60
53.44
2.31
0.10
0.11
87.25
37.79
50.39
3.40
1.75
0.00
15.32
Table 5: Music-conditioned humanoid motion generation on the OMG and Movens-Intent benchmarks. Fall and J-Limit are percentages.
Method
Motion Generation Quality
Tracking Fidelity and Execution Failures
FID ↓
Matching ↓
R@1 ↑
R@2 ↑
R@3 ↑
Diversity ↑
C-Slide ↓
Jerk ↓
MPJPE ↓
g-MPJPE ↓
Evel↓
Eacc↓
Fall ↓
J-Limit ↓
GENMO ( Li et al., 2025a )
77.30
1.35
14.70
23.83
31.45
1.21
0.86
237.09
67.68
134.19
2.94
1.95
0.10
11.18
HY-Motion ( Wen et al., 2025 )
30.66
1.25
36.67
52.49
62.40
1.33
1.44
49.35
45.63
117.10
3.30
0.94
1.03
13.62
TextOp ( Xie et al., 2026 )
105.46
1.48
5.91
10.35
14.89
0.90
1.43
89.54
38.05
129.41
3.04
1.10
0.20
5.42
Kimodo-SMPLX-RP ( Rempe et al., 2026 )
45.18
1.29
25.00
37.26
47.17
1.25
1.60
242.13
36.89
102.85
3.93
2.17
0.24
9.28
Kimodo-G1-RP ( Rempe et al., 2026 )
38.84
1.29
25.00
37.21
46.24
1.23
1.76
215.91
35.50
89.83
3.28
1.87
0.39
8.25
Table 6: Speech-conditioned humanoid motion generation on the Movens-Intent benchmark. R@K, Fall, and J-Limit are percentages. FID and C-Slide are scaled by 10−2 and 10−1 , respectively.
Figure 8: Qualitative overview of (a) video, (b) music, (c) speech, and (d) text conditioning. Each panel shows six temporal samples of the ground-truth motion (reference motion for music), Movens-Zero , and four baselines. Full comparisons appear in Appendix F.1 . Conditioning instructions appear in the headers; amber callouts indicate selected visible pose differences.
Movens-5M fraction
Unique hours
Text R@3 ↑
Video MPJPE ↓
Audio BeatAlign ↑
Speech R@3 ↑
Tracking Score ↑
1%
122
12.72
132.82
0.23
13.98
0.21
3%
367
15.81
129.31
0.29
17.88
0.28
10%
1,223
24.92
110.55
0.38
21.76
0.36
30%
3,669
41.80
87.90
0.45
54.34
0.50
100%
12,231
62.57
40.90
0.61
74.30
0.72
Table 7: Data scaling with the XL action expert (910.5M parameters).
Scale
Action expert
Val. FM Loss ↓
Text R@3 ↑
Video MPJPE ↓
Audio BeatAlign ↑
Speech R@3 ↑
Tracking Score ↑
S
64M
0.487
10.30
108.50
0.20
25.37
0.18
M
188M
0.392
22.96
92.80
0.26
41.25
0.24
L
500M
0.186
47.80
73.84
0.53
63.98
0.40
XL
911M
0.159
62.57
40.90
0.61
74.30
0.72
Table 8: Model scaling of Movens-Zero on the full Movens-5M training set.
Figure 9: Joint data–model scaling. (a–f) Six metrics versus action-expert size for five training-data fractions. (g) Relative improvements from 10% to 100% training data at each fixed model size, normalized by the corresponding 10% value. Reductions in FM loss and MPJPE, and increases in the other scores, are positive.
Variant
Text R@3 ↑
Video MPJPE ↓
Audio BeatAlign ↑
Speech R@3 ↑
Tracking Score ↑
Data curation
Raw recovered motion
48.72
126.82
0.48
35.31
0.41
+ visual and motion filtering
54.36
84.57
0.52
49.43
0.48
+ closed-loop physical refinement
57.84
64.91
0.55
64.35
0.58
+ tracking verification
61.92
57.62
0.58
70.23
0.67
+ distribution balancing (full)
62.57
40.90
0.61
74.30
0.72
Table 9: Ablations of data curation and modality-aligned training on the Movens-Intent benchmark.
Appendix figures & tables38 assets
Supplementary material from the paper’s appendix.
Appendix
Gate
Signal
Retention rule
Human relevance
Qwen3-VL yes/no likelihood
shuman≥0.90
Temporal rate
RGB-window class posterior
Slow or normal playback/encoding rate
Articulated motion
Body-normalized limb amplitude
Alimb≥0.08
Temporal integrity
Human/background flow statistics
No robust discontinuity test fires
Executability
Selected rollout tracking score
q≥0.20
Record completeness
Metadata and derived artifacts
All mandatory annotations, motions, rollout states, logs, and paths are present and readable
Appendix
Table 10: Hard gates used by the audited web-video assembly. Audio–visual consistency and generic perceptual scores are retained as diagnostics rather than restrictive final gates.
Figure 10: Representative Movens-5M examples: Daily, with (a1) gardening and (a2) mopping.
Figure 11: Representative Movens-5M examples (continued): Fitness, with (b1) alternating lunges and (b2) bodyweight squats. Each half-page case shows four video streams at six clip-relative times.
Figure 12: Representative Movens-5M examples (continued): Sports, with (c1) tennis swing and (c2) boxing. Each half-page case shows four video streams at six clip-relative times.
Figure 13: Representative Movens-5M examples (continued): Dance, with (d1) hip-hop and (d2) traditional dance. Each half-page case shows four video streams at six clip-relative times.
Figure 14: Representative Movens-5M examples (continued): Locomotion, with (e1) walking outdoors and (e2) running. Each half-page case shows four video streams at six clip-relative times.
Figure 15: Representative Movens-5M examples (continued): Performance, with (f1) drumming and (f2) fire-staff performance. Each half-page case shows four video streams at six clip-relative times.
Figure 16: Representative Movens-5M examples (continued): Gesture, with (g1) pointing gesture and (g2) stop gesture. Each half-page case shows four video streams at six clip-relative times.
Figure 17: Non-video fields for gardening record 00036673 : (a) source frames corresponding to Figure 10 (a1), (b) all non-caption metadata fields, including automatic human attributes, with original field names and values, and (c) the complete original video caption.
Figure 18: Non-video fields (continued): (d) the three original temporal intervals and their whole-body descriptions, and (e) upper-limb, lower-limb, hand, and face descriptions across all three intervals. The columns share the same temporal order, and all motion descriptions reproduce the original motion_caption text.
Metric
Threshold
Foot contact score
>0.6
Root jerk
<50m/s3
Position high-frequency ratio
<0.003
Minimum pelvis height
>0.6m
Maximum pelvis height
<1.5m
Pelvis distance to base of support
<6cm
Appendix
Table 11: Physics-aware motion filtering metrics and thresholds.
Scale
Hidden width D
Depth
Heads
Head dimension
DiT parameters
Action-expert parameters
S
768
3
12
64
56.1M
64.2M
M
1,152
4
18
64
174.0M
187.9M
L
1,536
6
24
64
479.2M
500.1M
XL
1,920
7
30
64
881.4M
910.5M
Appendix
Table 12: Action-expert architectures used in the model-scaling experiments. DiT parameters count the transformer backbone, while action-expert parameters count the complete trainable action side.
Task
Streams
Exposure
Input
Text-to-Motion
1
19.48%
Generate a motion sequence that matches the following description. Description: [text description].
Music-to-Motion
2
2.83%
[Music]
(Music, Text)-to-Motion
2
2.83%
[Music] Generate a dance motion that follows the rhythm and energy of this music.
Speech-to-Motion
1
9.74%
[Spoken command]
(Speech, Text)-to-Motion
1
9.74%
[Spoken command] Parse the utterance in the audio and move according to that instruction.
Video-to-Motion
3
27.69%
[Video]
Appendix
Table 13: Seven-task mixture used for action-expert training. A “stream” denotes a source-specific dataset, whereas a “task” denotes a conditioning objective. Sampling probabilities are computed from the enabled training-split sizes.
Model and motion
Optimization
Context features
Qwen2.5-Omni-3B, frozen
2,048-D
Optimizer
AdamW
β=(0.9,0.999)
Action expert
6 blocks, 24 heads
width 1,536; head dim. 64
Learning rate
10−4
minimum 5×10−7
Action-expert params
479.2M DiT
500.1M total
Schedule
cosine
1,000-step warmup
Motion/state token
40-D
q01–q99 norm. + clip
Weight decay / ϵ
0
10−8
Progress / time embed.
5-D / 256-D
—
Maximum steps
106
seed 292285
Temporal and modality sampling
Distributed execution
Appendix
Table 14: Training and inference settings for the L action expert.
Figure 19: Text benchmark examples: (a1) arm movements with turning and a leg lift; (a2) repeated squat-and-jump motion. Each case pairs a user instruction and the original text description with six G1 reference frames, read from left to right.
Figure 20: Speech benchmark examples (continued): (b1) stepping into a wide squat with hands at the chest; (b2) standing from sitting and moving the arms. Each case shows a user instruction, the original transcript and complete speech waveform, and six G1 reference frames. Audio and reference motion use separate time axes.
Figure 21: Music benchmark examples (continued): (c1–c2) user requests to dance to two music inputs. Each case pairs a dance request and complete music waveform with six G1 reference frames in a single row. Waveform labels report audio time; frame labels report reference time.
Figure 22: Video benchmark examples (continued): (d1) an indoor dance sequence; (d2) a fitness sequence. Each instruction asks the humanoid to imitate the person in the video. Input-video and G1-reference rows share six sampled times within each case.
Dataset
Sampled
Sources
Ref.
Flat
Slide
Contact
Exec.
BONES-SEED
1000
1000
1000
1000
988
988
1000
OmniRetarget
1000
1000
1000
1000
928
938
1000
PHUMA
1000
1000
1000
1000
1000
1000
1000
OMG-Data
1000
1000
1000
1000
297
325
1000
AMASS
1000
1000
1000
1000
997
999
1000
LAFAN1
77
77
77
77
61
63
77
Appendix
Table 15: Coverage of the independent complete-sequence cohorts. Sampled counts accepted sequences after technical rejection; Sources counts distinct selected recordings. Ref. and Flat count valid G1 sequences on the common floor, Slide counts sequences with measurable support-contact intervals, Contact counts sequences with at least one G1-kinematic sole-contact frame in the shared [−0.02,0.03] m band, and Exec. counts completed SONIC rollouts including failures. Non-Skating interval coverage is supplied in the result manifest.
Figure 23: Video Case 01: raising one arm from a low supported posture before transitioning to forward support.
Figure 24: Video Case 02: repeated changes in whole-body inclination.
Figure 25: Video Case 03: floor exercise with changing support and leg configurations.
Figure 26: Speech Case 04: stepping into a wide stance, lowering into a squat, and bringing the hands to the chest.
Figure 27: Speech Case 05: extending the right arm outward while the left arm remains lowered.
Figure 28: Speech Case 06: walking forward with the arms swinging naturally at the sides.
Figure 29: Text Case 07: placing one hand on the hip while the other arm remains down.
Figure 30: Text Case 08: walking forward, stopping, and bringing both hands together in front of the chest.
Figure 31: Text Case 09: moving both hands to the waist and returning the arms to the sides.
Figure 32: Text Case 10: walking forward with alternating steps and slight arm swing.
Figure 33: Text Case 11: raising bent arms, widening the stance, and swaying the hips while swinging the fists.
Figure 34: Music Case 12: dance with changes in stance, arm placement, and body orientation.
Figure 35: Music Case 13: dance with changing arm gestures and lower-body stance.
Figure 36: Music Case 14: dance with coordinated arm, leg, and facing-direction changes.
Figure 37: Six live-video-conditioned motions executed on Unitree G1. Each capability card shows twelve numbered pairs; every pair combines an earlier performer input with the later G1 response at a corresponding action phase and prints both source times. The time differences are display alignment offsets chosen for qualitative pose comparison, not measurements of end-to-end system latency.
Figure 38: Six Internet-video-conditioned motions executed on Unitree G1. Each capability card shows twelve numbered reference–execution pairs selected at clear corresponding action phases. Both crops in a pair come from the same composite-video timestamp; remaining pose differences reflect humanoid retargeting and dynamics rather than a timing shift introduced by the figure.
Figure 39: Six speech-conditioned motions executed on Unitree G1. Each capability card sets the full spoken instruction in parentheses and italics, reports the clip duration, shows twelve numbered robot snapshots with timestamps, and maps those numbers to the complete input-audio waveform.
Figure 40: Six music-conditioned motions executed on Unitree G1. Each capability card names the action, reports the clip duration, and shows twelve numbered robot snapshots above the complete input-audio waveform; matching numbers associate the snapshots with detected audio onsets.
Figure 41: Six text-conditioned motions executed on Unitree G1. Each capability card shows the complete input prompt, clip duration, twelve numbered robot snapshots with their timestamps, and the corresponding numbered locations on the execution timeline.
We present ScaleMoGen, a scale-wise autoregressive framework for text-driven human motion generation. Unlike conventional autoregressive approaches that rely on standard next-token prediction, ScaleMoGen frames motion generation as a coarse-to-fine process. We quantize 3D motions into compositional discrete tokens across multiple skeletal-emporal scales of increasing granularity, learning to generate motion by autoregressively predicting next-scale token maps. To maintain structural integrity, our motion tokenizers and quantizers are explicitly designed so that discrete tokens at every scale strictly preserve the skeletal hierarchy. Additionally, we employ bitwise quantization and prediction, which efficiently scale up the tokenizer vocabulary to preserve motion details and stabilize optimization. Extensive experiments demonstrate that ScaleMoGen achieves state-of-the-art performance, establishing an FID of 0.030 (vs. 0.045 for MoMask) on HumanML3D and a CLIP Score of 0.693 (vs. 0.685 for MoMask++) on the SnapMoGen dataset. Furthermore, we demonstrate that our skeletal-temporal multi-scale representation naturally facilitates training-free, text-guided motion editing.
Inwoo Hwang, Hojun Jang, Bing Zhou +3
Seoul National University · Snap Inc. · Meta Reality Labs
Text-conditioned motion generation is a promising interface for programming humanoid robots, yet current generators are often trained on human motion datasets retargeted to robot morphologies. Although such data provides rich semantic and kinematic priors, it fails to capture the nuances of whole-body tracking controllers, including balance, contact dynamics, actuation limits, and controller-specific failure modes. As a result, generated motions can be semantically plausible but difficult or impossible for the robot to execute. We introduce TEXEDO, a test-time scaling framework for humanoid motion generation that improves motion quality without requiring a stronger underlying generator. Given a text prompt, TEXEDO samples multiple candidate motions from a pretrained text-conditioned generator and selects the best motion that is both executable and task-aligned. The reward model combines a dynamic feasibility verifier, distilled from whole-body tracking rollouts to predict physical executability, with a semantic alignment verifier that measures text-motion alignment in a learned co-embedding space. Our pipeline treats dynamic feasibility as a hard constraint and semantic alignment as the selection objective within the feasible set. Through large-scale simulation studies and real-world deployment on a Unitree G1 humanoid robot, we show that TEXEDO consistently improves both tracking fidelity and text alignment. These results demonstrate that grounded verification is an effective path toward deployable language-guided humanoid motion generation. Project website: https://jianuocao.github.io/TEXEDO/
Jianuo Cao, Yuxin Chen, Yuzhen Song +3
Nanjing University · University of California, Berkeley · Southern University of Science and Technology
High-quality human motion data is becoming increasingly important for applications in robotics, simulation, and entertainment. Recent generative models offer a potential data source, enabling human motion synthesis through intuitive inputs like text prompts or kinematic constraints on poses. However, the small scale of public mocap datasets has limited the motion quality, control accuracy, and generalization of these models. In this work, we introduce Kimodo, an expressive and controllable kinematic motion diffusion model trained on 700 hours of optical motion capture data. Our model generates high-quality motions while being easily controlled through text and a comprehensive suite of kinematic constraints including full-body keyframes, sparse joint positions/rotations, 2D waypoints, and dense 2D paths. This is enabled through a carefully designed motion representation and two-stage denoiser architecture that decomposes root and body prediction to minimize motion artifacts while allowing for flexible constraint conditioning. Experiments on the large-scale mocap dataset justify key design decisions and analyze how the scaling of dataset size and model size affect performance.