Motion Capture — Walk Cycles

1,200 mocap walk and run cycles as retargetable BVH — moods, speeds, loads and terrains.

A locomotion library for animators and motion models: clean, loopable walk, run and idle-to-walk transitions captured on a 53-marker setup. Every clip is foot-locked, retargeted to a standard skeleton, and labelled with speed, mood and carried-load so you can filter to exactly the gait you need.

Dataset details

Supported tasks

• Character animation and game locomotion
• Motion-synthesis and motion-matching model training
• Biomechanics and gait reference

Dataset structure

Each item ships as a BVH clip plus a manifest describing the motion:

{
  "id": "walk_00204",
  "bvh": "walk_00204.bvh",
  "action": "walk",
  "speed_mps": 1.4,
  "mood": "confident",
  "load": "none",
  "terrain": "flat",
  "loopable": true,
  "frames": 96,
  "fps": 30
}

Samples

Curation

Captured in a marker-based studio with performers directed through a fixed shot list of gaits and moods. Clips were cleaned (gap-filled, denoised), foot-locked to remove sliding, and retargeted to a documented 23-joint skeleton.

Personal and sensitive information

Skeletal motion data only — no imagery, faces or biometric identifiers. Performer consent on file for commercial ML use.

Considerations & limitations

• Single reference skeleton; extreme proportions need re-retargeting.
• Locomotion-focused — no combat, dance or object interaction.
• Captured on flat ground, so slope and stair adaptation is approximated.

License & provenance

Captured with performer consent; retargeted to a standard rig. The sample archive (full_dataset.zip) is a free 50-item preview; the full set of 1,200 items unlocks with a purchase. Redistribution of the raw dataset is not permitted under either license.

Citation

@misc{aria-forge_00000014,
  title  = {Motion Capture — Walk Cycles},
  author = {Aria Forge Studio (@aria-forge)},
  year   = {2026},
  note   = {TensorForge marketplace dataset, LST-00000014-0000-4000-8000-000000000014}
}