Type a text prompt and get a rigged 3D character performing that motion, using fal video models, GVHMR pose recovery and Mixamo retargeting.
toyxyz/toyxyz_RGBMocap
A native Windows app that turns video into editable human motion, running AI pose models locally and exporting FBX or VMD.
Open-source alternative to:Rokoko Vision
- Stars
- 866
- Forks
- 116
- Open issues
- 5
- Last commit
- yesterday
What is it?
RGB MoCap is a Windows desktop app for reconstructing human motion from ordinary video. It runs several AI motion-capture models (ExMR, FlowHMR, GVHMR, MAMMA and PromptHMR) locally on an NVIDIA GPU, so you can review the result in a 3D preview before exporting it.
After reconstruction, the app offers foot and palm contact correction with editable contact timings to reduce sliding. Motion can be retargeted onto Mixamo FBX, VRM and PMX humanoid characters, then exported as FBX, or as VMD for MMD workflows. Projects keep separate results for each AI model.
Setup is handled by scripts (install.cmd, RGBMoCap.cmd) and a Model Manager for downloading weights. Some models require account registration and license acceptance, and the full set of models takes tens of gigabytes of disk space.
Key features
- 01Reconstructs one or several people and previews body mesh and bones.
- 02Corrects foot and palm contact sliding with editable contact timings.
- 03Retargets motion to Mixamo FBX, VRM and PMX humanoid characters.
- 04Exports animation as FBX, or PMX character motion as VMD for MMD.
- 05Supports multi-camera capture through the MAMMA model.
Use cases
- Turning reference footage into animation for a Unity, Blender or game character.
- Creating MMD dance motion from a video using PMX characters and VMD export.
- Prototyping motion capture for animation work without a dedicated mocap studio.
- Comparing outputs from several pose-estimation models on the same clip.
Quick start
- 1
Install Git for Windows and clone the repository
git clone https://github.com/toyxyz/toyxyz_RGBMocap.git - 2
Enter the project folder
cd toyxyz_RGBMocap - 3
Prepare the runtimes and dependencies
install.cmd - 4
Start the app and keep the console open
RGBMoCap.cmd
Things to know
- Windows only, requiring 64-bit Windows and a compatible NVIDIA GPU with OpenGL 4.3.
- Installed models and runtimes need about 63 GiB of disk space.
- Some AI model files require account registration and manual download.
- Optional MAMMA and PromptHMR runtimes need Visual Studio C++ Build Tools and CUDA Toolkit 12.8.
In collections
Topics
FAQ
- What is toyxyz_RGBMocap?
- A native Windows app that turns video into editable human motion, running AI pose models locally and exporting FBX or VMD. RGB MoCap is a Windows desktop app for reconstructing human motion from ordinary video. It runs several AI motion-capture models (ExMR, FlowHMR, GVHMR, MAMMA and PromptHMR) locally on an NVIDIA GPU, so you can review the result in a 3D preview before exporting it.
- Is toyxyz_RGBMocap free to use? What license does it have?
- toyxyz_RGBMocap is open source under the Other license. Check the license terms before using it in commercial products.
- Is toyxyz_RGBMocap still maintained?
- Actively maintained. The last commit to toyxyz_RGBMocap was on Nov 7, 2026.
- What can I use toyxyz_RGBMocap for?
- Turning reference footage into animation for a Unity, Blender or game character. Creating MMD dance motion from a video using PMX characters and VMD export. Prototyping motion capture for animation work without a dedicated mocap studio. Comparing outputs from several pose-estimation models on the same clip.
- How do I install toyxyz_RGBMocap?
- Install Git for Windows and clone the repository: git clone https://github.com/toyxyz/toyxyz_RGBMocap.git Enter the project folder: cd toyxyz_RGBMocap Prepare the runtimes and dependencies: install.cmd Start the app and keep the console open: RGBMoCap.cmd
- Where can I get toyxyz_RGBMocap?
- The source code and installation instructions are on GitHub: https://github.com/toyxyz/toyxyz_RGBMocap
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Retargets human motion from a video or still photo onto an Adobe Mixamo-rigged 3D character, exporting a skinned .glb file.
Free, open-source markerless motion capture system that records multi-camera video and reconstructs 3D human movement.
Windows C++ port of OpenAI's Whisper speech recognition that runs inference on any Direct3D 11 GPU via DirectCompute.
A cross-platform C++20 3D library with a three.js-style API, OpenGL and Vulkan backends, Python bindings, and physics and sensor simulation.
Official implementation of FlowHMR, a generative model that reconstructs physically plausible 3D human motion from monocular video.