Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > ComfyUI-SkinToken

ComfyUI Extension: ComfyUI-SkinToken

Repo Name

ComfyUI-SkinToken

Author
rizzlord (Account age: 2297 days)
Nodes
View all nodes(1)
Latest Updated
2026-07-16
Github Stars
0.05K

How to Install ComfyUI-SkinToken

Install this extension via the ComfyUI Manager by searching for ComfyUI-SkinToken
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter ComfyUI-SkinToken in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

ComfyUI-SkinToken Description

ComfyUI-SkinToken automatically rigs 3D character meshes by predicting skeletons and skinning weights. It can transfer rigs to detailed meshes and export the result as GLB or FBX files.

<p align="center"> <img src="assets/banner.png" alt="ComfyUI-SkinToken banner" width="100%">

ComfyUI-SkinToken Introduction

ComfyUI-SkinToken adds automatic 3D character rigging to ComfyUI. It wraps the SkinTokens project in custom nodes so you can take a character mesh, generate a skeleton and skinning weights, and export a rigged GLB or FBX file.

A skeleton is the set of joints and bones that lets a character pose or move. Skinning weights tell the mesh how strongly each bone should move its nearby surface. Creating both by hand can take substantial time, especially when you are testing many characters. ComfyUI-SkinToken predicts them from the mesh geometry, so you can create a starting rig without placing every bone manually.

The extension can also transfer a generated rig onto your original, higher-detail mesh, adjust the skeleton’s origin for workflows that use a feet-level pivot, and rename bones using Mixamo or Unreal Engine 5 conventions. The author provides these options to make the resulting character easier to carry into the next stage of your workflow. You should still inspect the exported rig and test important poses before using it in a final animation or game asset.

How ComfyUI-SkinToken Works

Think of rigging as building a movable frame inside a character and then deciding which parts of the character follow each piece of that frame. ComfyUI-SkinToken handles those two tasks in sequence:

  1. Read the mesh. A ComfyUI mesh-loading node supplies a TRIMESH, which represents your character’s 3D surface.
  2. Predict the skeleton. The model estimates joint positions, how the joints connect, and bone lengths from the mesh geometry.
  3. Predict skinning weights. It assigns bone influences to the mesh’s vertices. For example, vertices near an elbow may need to follow both the upper arm and forearm.
  4. Prepare and export the result. Blender processes the rigged character and exports it as GLB or FBX. The extension can run Blender in a background process without opening its graphical interface.

A basic ComfyUI workflow looks like this:

[Load 3D Model] → [TRIMESH] → [SkinToken Rig] → [Rigged GLB or FBX]

Connect a mesh loader’s TRIMESH output to SkinToken Rig, choose your export settings, and run the workflow. If your source is a detailed mesh, you can enable use_transfer to apply the predicted rig back to the original geometry.

ComfyUI-SkinToken Features

Rig generation and mesh handling

  • Automatic skeleton generation: Predicts joints, their hierarchy, and bone lengths. This gives you a rigging starting point without manually placing bones.
  • Skinning weight prediction: Determines which bones influence each vertex. group_per_vertex limits the number of bone influences per vertex; its default is 4. Allowing more influences may make some bends smoother but can increase the work needed to use the rig.
  • Rig transfer: use_transfer, off by default, transfers the generated rig back onto the original mesh. Try it when preserving the detail of your source character matters.
  • Optional skin postprocessing: use_postprocess, off by default, applies voxel-based skin postprocessing. Compare an export with and without it to see which better suits your mesh.
  • Mesh cleanup and smoothing: The export workflow can merge nearby vertices and apply angle-based smooth shading. smooth_angle defaults to 55.0; it controls the angle threshold used for smooth shading.

Skeleton placement and naming

  • Bottom-center origin: bottom_center_origin is off by default. Turn it on to place the skeleton origin at the bottom-center of the mesh’s bounding box, which can be useful when you want the character’s pivot near its feet.
  • Skeleton templates: skeleton_template defaults to Keep model names. You can choose Mixamo or Unreal Engine 5 naming conventions when those names better fit your downstream workflow. Renaming bones does not replace the need to check that the generated skeleton suits your character and intended animation.

Generation controls

The default settings are a reasonable place to begin. These controls affect how the model selects a generated skeleton:

SettingDefaultWhat changing it does
top_k5Limits generation choices to the highest-ranked candidates at each step.
top_p0.95Uses a probability threshold to narrow the choices considered during generation.
temperature1.0Lower values generally make results more consistent; higher values allow more variation.
repetition_penalty2.0Discourages repeated choices during generation.
num_beams10Sets the width of the model’s search for a result. A higher value can favor a more consistent selection.

If you are comparing repeated runs of the same character, try lowering temperature or increasing num_beams. Change one setting at a time so you can tell which adjustment affected the rig.

Output and workflow options

SettingDefaultPurpose
file_formatglbChoose glb or fbx for the exported character.
save_fileTrueSave the exported file to the output directory.
filename_prefix3D/SkinToken_Set the beginning of the output filename.
deviceautoSelect automatic device handling or cuda. Inference requires a CUDA-capable GPU.
ckpt_namegrpo_1400.ckptSelect the checkpoint used for rig prediction.

The extension provides two nodes:

  • SkinToken Download Models retrieves the required checkpoints from Hugging Face into ComfyUI/models/skintoken/. Its outputs report the download status and model directory.
  • SkinToken Rig performs the rigging. It outputs a rigged TRIMESH, the exported file path, an internal SKINTOKEN_ASSET for chaining nodes, and a backend label indicating bpy or blender_headless.

Required checkpoints are also downloaded automatically on the first run. Export uses Blender 4.0 or newer; it can run in the background without a Blender window.

ComfyUI-SkinToken Models

The provided documentation identifies grpo_1400.ckpt as the default rigging checkpoint. Select it through ckpt_name unless you have another checkpoint available in that dropdown. No additional checkpoint names or comparisons are documented here.

The underlying SkinTokens project supplies the rigging model. The documentation also identifies Qwen3-0.6B as its language-model backbone for token generation. You do not need to choose between these as separate character-rigging modes: the checkpoint selection is the model control exposed by SkinToken Rig.

Troubleshooting ComfyUI-SkinToken

The rigging node does not receive my model. What should I check?

  1. Confirm that your mesh-loading node produces a TRIMESH output.
  2. Connect that output to SkinToken Rig.
  3. Run the workflow again. A file path or another output type is not a substitute for the node’s TRIMESH input.

A checkpoint is unavailable or the model download does not complete. What should I check?

  1. Run SkinToken Download Models and read its status output.
  2. Check that the models_dir output points to ComfyUI/models/skintoken/.
  3. Confirm that the default ckpt_name is grpo_1400.ckpt, or select a checkpoint that is actually available.

Blender export cannot start. What should I check?

  1. Confirm that Blender 4.0 or newer is available.
  2. If Blender is not found automatically, set SKINTOKEN_BLENDER_BIN to the Blender executable’s path.
  3. Run the workflow again. The extension can export through a background Blender process, so an open Blender window is not required.

My detailed character loses detail in the rigged result. What can I try?

Enable use_transfer and compare the new export. This option transfers the generated rig onto the original mesh to preserve its geometry.

My character bends awkwardly. Which setting should I try first?

Inspect the skeleton and skinning around the problem joint. You can then try increasing group_per_vertex from its default of 4 and compare the deformation. This setting allows more bones to influence a vertex, but it does not guarantee a better result for every mesh.

The character’s pivot is not near its feet. Is there a setting for that?

Enable bottom_center_origin. It places the skeleton origin at the bottom-center of the mesh’s bounding box. Check the result on your particular character before using it downstream.

Will choosing a Mixamo or Unreal Engine 5 template make the rig animation-ready?

The templates change bone naming conventions. After export, inspect the generated bones and test the poses or retargeting workflow you intend to use.

Learn More about ComfyUI-SkinToken

ComfyUI-SkinToken Related Nodes

RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

ComfyUI-SkinToken detailed guide | ComfyUI