Mixamo LLM Mocap
Turn a locked-camera video, filmed or AI-generated, into a clean FK animation on any Mixamo character in Blender. GVHMR pose estimation, SMPL-X retargeting, foot locks and fidelity checks, built to be run end to end by an AI agent.
- GitHub stars
- 402 (GitHub stars as of 2026-10-24)
- License
- MIT
- Last commit
- 2026/09/26
- By
- squall01337
Listed from the author's README. GameSkill has not run this end to end yet. Agent support and install steps are as documented by the author.
Install
git clone https://github.com/squall01337/mixamo-llm-mocapPoint your agent at the README pipeline. The default path runs Blender headless; the landmark path uses Blender MCP.
Commands are copied from the project's README. Check the repository before running anything on your machine.
What you get
- estimate_pose_gvhmr.py: GVHMR to SMPL-X, refined on ViTPose keypoints plus heels, toes and zoomed hands
- retarget_smplx.py: rotations onto your Mixamo rig with floor, foot locks and camera-true trajectory
- bl_motion.py: keys the action in headless Blender
- render_review.py and eval_fidelity.py: 2×2 review video and limb/foot-slide/jitter scores
- Two-performer support: --person left|right, setup_duo.py and compare_pair.py
Good fit
- Fight and action animations from AI-generated video
- Indie games using Mixamo rigs without a mocap suit
Not for
- Moving-camera footage; the plate must be locked
- Machines without a CUDA GPU
- Plug-and-play SKILL.md install; it is a scripted pipeline, not a packaged skill
Overview
Mixamo LLM Mocap is a video-to-animation pipeline designed so an agent can run every stage. Feed it a locked-camera clip, including AI-generated video, and it produces a keyed FK action on your Mixamo character, plus a review video and numeric fidelity scores the agent can use to decide whether to iterate.
The default path transfers SMPL-X joint rotations, so forearm twist, spine curve and head orientation are real, and it refits fast limbs on the plate's keypoints so a head-high kick stays head-high. Contacts from the detector, the floor, the image and physics keep planted feet from sliding. An older landmark path, driven through Blender MCP, adds beat sheets, QA gates and mesh-vs-mesh contact checks for two fighters.
Related
- MCP for Blender for the live Blender session the landmark path uses.
- Blender character rigging for preparing rigs.
Requirements
- Blender 5.1+
- An NVIDIA GPU with about 8 GB VRAM (CUDA)
- A Mixamo character
Topics
FAQ
Works well with
Skills the author points to as related or as the better fit for neighbouring tasks.
MCP for Blender (blender-mcp)
Community plugin to control Blender with any LLM. The most-starred game-adjacent MCP server on GitHub.
Blender Character Rigging
Character rigging in Blender 5.2: Rigify, bone layers, IK/FK, robust skinning, correctives and game skeletons for Unity, Unreal and Roblox.
Blender Animation
Animation in Blender 5.2: layered actions, Python keyframing, NLA, mocap retargeting, baking rigs to export skeletons and per-clip export.
Similar skills
Video to Blender (procedural-blender-nodes)
A Blender skill that turns a reference video, image or idea into editable Geometry Nodes, Shader Nodes and compositor graphs, backed by a tested 39-group node library, a linter and a validator. Requires Blender 5.2+.
Scenario Skills
Production-ready images, video, audio and 3D from any agent through the Scenario MCP, plus expert teams for Blender, Unreal and Unity.
Blender Hard-Surface Modeling (Scenario)
Hard-surface modeling in Blender for props, weapons and vehicles: booleans, bevels, weighted normals and support loops, plus shading fixes.
Blender Rigging with Python (Scenario)
Rig characters, creatures and props in Blender through Python: Rigify, weights, IK/FK, drivers, corrective and facial shape keys, and deformation tests.