Skip to content

Releases: Xingxun7777/blender-kimodo-motion

Release list

v0.1.0 — Initial Release

v0.1.0 — Initial Release Latest
Latest

Choose a tag to compare

View all tags
@Xingxun7777 Xingxun7777 released this 17 Apr 00:47
v0.1.0
60c7f67

First public release.

Highlights

  • One-click Windows installer (Python 3.12 + venv + PyTorch cu128 + fbxsdkpy + kimodo)
  • FBX-SDK file-level retarget with bone-structure hash caching
  • Auto skeleton preset detection (Mixamo / VRoid / MMD)
  • Multi-sample generation (N variants per prompt)
  • Translation API bridge (DeepSeek / OpenRouter / Moonshot / OpenAI-compatible)
  • Chinese / English UI

Requirements

  • Blender 5.0.1+ (4.x rejected at addon load)
  • Windows 10 / 11 x64
  • NVIDIA RTX 20 / 30 / 40 / 50 series (16 GB+ VRAM recommended for the LLaMA-3-8B text encoder)
  • ~50 GB free disk (5 GB venv + 17 GB HuggingFace models + overhead)

Installation

  1. Download kimodo_motion.zip from Assets below.
  2. Blender > Edit > Preferences > Add-ons > Install > pick the zip > Enable.
  3. Open N-panel > Kimodo > Runtime Install > click One-click install runtime (or double-click installer/install.cmd).
  4. Log in to HuggingFace after the venv is built: 
    <venv>\Scripts\python.exe -m huggingface_hub.commands.huggingface_cli login
  5. Select your character armature, type a prompt, click Generate and apply to selected armature.
  6. First generation downloads ~17 GB of models (30 min). Subsequent generations: 8–60 seconds.

Full guide in INSTALL.md and README.md.

Real-machine acceptance test

  • Clean install: 1 m 26 s (8 steps, torch reused, fbxsdkpy pulled from INRIA)
  • Server /generate: 51.4 s (first request, includes LLaMA-3-8B load + diffusion at 50 steps)
  • FBX retarget via subprocess: 0.875 s per sample
  • Output: 127 KB valid FBX binary ("Kaydara FBX Binary" magic), 22 rotation + 1 translation channels applied

License boundary (IMPORTANT)

Component License Distribution
Addon itself MIT bundled in zip
Vendored retarget (vendor/kimodo_retarget/kimodo_retarget_fbx.py) Apache-2.0 bundled with LICENSE-APACHE2.0
fbxsdkpy (Autodesk FBX SDK wrapper) Autodesk FBX SDK LSA never bundled — pulled from INRIA GitLab
kimodo (NVIDIA) Apache-2.0 pip-installed from git
Kimodo-SOMA-RP-v1 weights NVIDIA Open Model License auto-downloaded from HuggingFace
Meta-LLaMA-3-8B-Instruct LLaMA-3 Community License (gated) user accepts + HF token
PyTorch cu128 BSD-3 pip from pytorch.org
Assets 4
Loading