SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.15546 · 2026

Kimodo: Scaling Controllable Human Motion Generation

Jan Kautz, Ye Yuan, Davis Rempe, Xue Peng, Sanja Fidler, Haotian Zhang, Jiefeng Li, Yifeng Jiang, Chen Tessler, Michael Huang, Chaeyeon Chung, Umar Iqbal, and 12 more

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 17 functions out of this paper's own repositories and ran 12 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
nv-tlabs/kimodo canonical 12 of 17
FunctionStatusWhere it lives
angle_to_Y_rotation_matrix Ran nv-tlabs/kimodo/kimodo/geometry.py
code served (permissive licence) · get_code("d6b4add804976434")
cont6d_to_matrix Ran nv-tlabs/kimodo/kimodo/geometry.py
code served (permissive licence) · get_code("42c47c90157ca5e4")
create_pairs Ran nv-tlabs/kimodo/kimodo/constraints.py
code served (permissive licence) · get_code("495275fac958edb6")
get_beta_schedule Ran nv-tlabs/kimodo/kimodo/model/diffusion.py
code served (permissive licence) · get_code("0c0a342a6d097b28")
get_env_var Ran nv-tlabs/kimodo/kimodo/model/common.py
code served (permissive licence) · get_code("1261e7b3e8e04b68")
get_skeleton_display_name Ran nv-tlabs/kimodo/kimodo/model/registry.py
code served (permissive licence) · get_code("a5998a861ba6879c")
get_skeleton_key_from_display_name Ran nv-tlabs/kimodo/kimodo/model/registry.py
code served (permissive licence) · get_code("ca4f17050d7f391c")
load_checkpoint_state_dict Ran nv-tlabs/kimodo/kimodo/model/loading.py
code served (permissive licence) · get_code("e95c6635f9e67e38")
matrix_to_cont6d Ran nv-tlabs/kimodo/kimodo/geometry.py
code served (permissive licence) · get_code("0ae30a547c08a556")
pad_x_and_mask_to_fixed_size Ran nv-tlabs/kimodo/kimodo/model/backbone.py
code served (permissive licence) · get_code("2609d3cb0326a521")
parse_prompts_from_meta Ran nv-tlabs/kimodo/kimodo/meta.py
code served (permissive licence) · get_code("a06db7b4a4f76637")
resolve_target Ran nv-tlabs/kimodo/kimodo/model/common.py
code served (permissive licence) · get_code("debe7d25330cd587")
get_env_var Not yet run nv-tlabs/kimodo/kimodo/model/loading.py
code served (permissive licence) · get_code("a9ce08fb03d39b34")
get_skeleton_display_names_for_dataset Not yet run nv-tlabs/kimodo/kimodo/model/registry.py
code served (permissive licence) · get_code("0ef3379f93d586a4")
instantiate_from_dict Not yet run nv-tlabs/kimodo/kimodo/model/loading.py
code served (permissive licence) · get_code("006930e776ca886b")
load_model Not yet run nv-tlabs/kimodo/kimodo/model/load_model.py
code served (permissive licence) · get_code("9864a44c22701ca7")
materialize_value Not yet run nv-tlabs/kimodo/kimodo/model/common.py
code served (permissive licence) · get_code("d4eeec6d214fdc47")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

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.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2603.15546")
get_code_for_paper("2603.15546")
have("2603.15546")

Connect an agent — have() is free.