The Dataset Viewer has been disabled on this dataset.

Kimodo Avatar Rigs

Two Unreal Engine 5 Mannequin skeletons and one rigid-parts robot, in glTF binary form, hosted so the Kimodo Avatar Motion Studio can offer them as built-in rigs. The tool fetches them on click; nothing else reads this repo.

They live here rather than beside the app because a Space repo rejects any commit containing a binary file, and a dataset is the repo type built for them.

File Size Model
ue5-manny.glb 5.2 MB UE5 male Mannequin, "Manny"
ue5-quinn.glb 4.9 MB UE5 female Mannequin, "Quinn"
soma-robot.glb 1.2 MB Rigid-parts robot fitted to the SOMA-30 skeleton

The two Mannequins are skinned meshes with an 80-bone skeleton, two JPEG textures, and no animation.

soma-robot.glb is a different kind of file

No skeleton, no skin weights, no textures. It is 22 rigid nodes, each named exactly as a driven SOMA-30 joint names it — Hips, Spine1, Spine2, Chest, Neck1, Head, Left/RightShoulder, Arm, ForeArm, Hand, Leg, Shin, Foot, ToeBase — holding the merged geometry that rides on that joint. Driving it is two writes per joint and no skinning at all:

node.position.copy(jointWorldPosition);
node.quaternion.copy(jointGlobalRotation);

which is exactly the shape of Kimodo's output: 30 global rotations plus a root position. The geometry inside each node is already the bind offset from its joint, and each node sits at that joint's SOMA rest position, so there is nothing to compute.

The eight joints Kimodo does not drive (Neck2, Jaw, the eyes and the fingertip markers) have no node, which is why the fingers ride rigid with the hand: two fingertip markers cannot pose a five-finger chain, so posing them would be invention.

Credit and licence

Redistributed under CC BY 4.0, which is the licence the author published them under. Each file also carries its own author, licence and source URL in its glTF asset.extras, so the credit travels with the file.

File Author Source
ue5-manny.glb SketchPunk UE5 Male Mannequin — Manny — Web Ready
ue5-quinn.glb SketchPunk UE5 Female Mannequin — Quinn — Web Ready
soma-robot.glb the repo owner original model, fitted to SOMA by blender-fit-robot.py

soma-robot.glb carries no licence of its own yet. Its asset.extras leaves the field blank rather than assert one, and the CC BY 4.0 in this repo's front matter was written for the two Mannequins. If you want to reuse the robot, ask first.

The Mannequin design itself is Epic Games'. These are SketchPunk's published derivatives, and it is their CC BY 4.0 licence this redistribution relies on.

Bone names, and the trap in them

The skeleton is the standard Unreal hierarchy — pelvis, spine_01…spine_05, neck_01, neck_02, head, clavicle_l, upperarm_l, lowerarm_l, hand_l, thigh_l, calf_l, foot_l, ball_l, plus the *_twist_* bones and a full set of finger chains.

Every name carries a numeric suffix. The Sketchfab glTF exporter appends the node index to make names unique, so the file actually contains pelvis_78, spine_01_61, upperarm_l_28, thigh_l_77, ball_l_70. Anything matching bones by name needs to strip a trailing _<digits> — but only as a fallback after an exact match fails, because plenty of real Unreal names legitimately end in a number (spine_01, neck_01, thumb_01_l) and stripping eagerly turns spine_01 into spine.

Direct use

https://huggingface.co/datasets/MazenVR/kimodo-rigs/resolve/main/ue5-manny.glb
https://huggingface.co/datasets/MazenVR/kimodo-rigs/resolve/main/ue5-quinn.glb
https://huggingface.co/datasets/MazenVR/kimodo-rigs/resolve/main/soma-robot.glb

Both answer 200 with Access-Control-Allow-Origin: *, so a browser can fetch them cross-origin:

const response = await fetch(
  'https://huggingface.co/datasets/MazenVR/kimodo-rigs/resolve/main/ue5-quinn.glb',
);
const gltf = await new GLTFLoader().parseAsync(await response.arrayBuffer(), '');

The dataset viewer is off: these are 3D assets, not tabular data, and there is no split for it to show.

Downloads last month
281