Skip to main content
affpapa.org
RU EN
skill CX Codex

kimodo-motion-diffusion — Claude Code skill: 3D motion from text

A skill that wires the Kimodo kinematic diffusion model into your agent for 3D human and humanoid robot motion from text prompts and constraints. Expect roughly 17GB of VRAM.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add reason-machines/trending-skills --skill kimodo-motion-diffusion -a claude-code
Open on GitHub

What kimodo-motion-diffusion does

This skill gives the agent Kimodo, a kinematic motion diffusion model trained on mocap data. You drive it with text and kinematic constraints: full-body keyframes, end-effector position and rotation, 2D root paths and waypoints. Model weights pull from Hugging Face the first time you run something.

Two CLI tools ship with it. kimodo_gen generates motion from the command line; kimodo_demo opens a web demo at http://127.0.0.1:7860 with a timeline editor, path and waypoint editing, and 3D visualization. A Python API is there too — Kimodo, ConstraintSet, FullBodyKeyframe, EndEffectorConstraint. Output lands as NPZ, as a qpos CSV for the G1 robot in MuJoCo, or as an AMASS-compatible NPZ for SMPL-X.

Linux is the supported host; on Windows you go through Docker. Without a GPU of about 17GB VRAM nothing generates — the authors point at RTX 3090, 4090 or A100. Install into a clean environment: pip install -e . inside a clone of the repo, or build the Docker image.

What it helps with

  • generate a 3D human motion from a text description
  • produce a walking animation for a Unitree G1 robot
  • pin keyframes and a hand position inside a motion
  • export motion to AMASS or qpos formats
  • chain several prompts into one transition sequence

Example prompts for the agent

  • Generate a human motion for the prompt "a person walks forward at a moderate pace", 4 seconds long
  • Make the G1 robot walk and save qpos to ./outputs/walk.csv
  • Launch the Kimodo interactive demo so I can place keyframes

How to install kimodo-motion-diffusion

1Skills CLI (Vercel)

npx skills add reason-machines/trending-skills --skill kimodo-motion-diffusion -a claude-code

2Manually: put the files in place

~/.claude/skills/kimodo-motion-diffusion/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/kimodo-motion-diffusion/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about kimodo-motion-diffusion

How do I install kimodo-motion-diffusion in Claude Code?

Run npx skills add reason-machines/trending-skills --skill kimodo-motion-diffusion -a claude-code. The skill lands at ~/.claude/skills/kimodo-motion-diffusion/SKILL.md.

What does kimodo-motion-diffusion do?

It exposes the Kimodo model to your agent: 3D motion for humans and humanoid robots from text and kinematic constraints, with export to NPZ, qpos CSV and AMASS.

What does the skill need to run?

Linux, a GPU with about 17GB VRAM (RTX 3090, 4090 or A100), and the Kimodo package installed. Windows works through Docker.

Similar skills

All skills for AI and ML →
Nova3D RareSense Turns a prompt into 3D models and scenes with named parts: a GLB, a preview link and a Blender script. Useful when an agent has to build an object piece by piece. CX Codex +1 ★ 732repo stars blender-motion-state-inspection affaan-m Inspects Blender characters and retargeted animations using bone coordinates and scene geometry instead of guesswork from screenshots. CX Codex +1 5,000installs stable-diffusion-image-generation Orchestra Research A skill that walks your agent through Stable Diffusion image generation with HuggingFace Diffusers: text-to-image, img2img, inpainting, ControlNet. CX Codex +1 1,000installs nvidia-skills NVIDIA This plugin finds NVIDIA skills for GPU acceleration, CUDA, AI agents, model training, robotics, Physical AI, Omniverse, and simulation. ★ 3,748repo stars srdf earthtojake This skill creates and validates MoveIt2 SRDF files paired with a URDF. Use it for planning groups, robot semantics, and review-ready handoff. CX Codex +1 10,400installs fine-tuning-serving-openpi Orchestra Research Workflows for fine-tuning and serving Physical Intelligence OpenPI models (pi0, pi0-fast, pi0.5). Guides setup, norm stats calculation, and policy inference servers. CX Codex +1 861installs

From the same set199

freecodecamp-curriculum Get guidance on freeCodeCamp development, from running the project locally to authoring and validating curriculum challenges. CX Codex +1 1,400installs openmaic-classroom A skill that generates interactive lessons: slides, quizzes, simulations and project work come out of a multi-agent pipeline. Useful when you need a course from a topic or document. CX Codex +1 1,400installs cmux-terminal-multiplexer Equips AI agents with CLI commands for terminal split orchestration and headless Chromium browser actions via surface handles. CX Codex +1 1,400installs nightingale-karaoke This skill guides an agent through a Rust karaoke app with vocal separation, timed lyrics, player profiles, and pitch scoring. CX Codex +1 1,300installs grimmory-self-hosted-library Helps AI agents configure and run the self-hosted Grimmory book manager in Docker, set up e-reader sync, and extend its Spring Boot backend. CX Codex +1 1,300installs taiwan-md-knowledge-base A skill for maintaining an open knowledge base about Taiwan built on Astro v5, with Chinese and English articles and a D3.js graph. Useful if you write or translate content for that project. CX Codex +1 1,300installs zeroboot-vm-sandbox This skill runs agent code inside a real KVM virtual machine through Zeroboot, forking a template in under a millisecond. CX Codex +1 1,300installs slug-font-rendering Provides reference HLSL shaders for GPU vector font rendering with the Slug algorithm. Useful for drawing glyphs at any scale without texture atlases. CX Codex +1 1,300installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor