This is the official repository for Instant Skinned Gaussian Avatars for Web, Mobile and VR Applications (SUI 2025 Demo Track).
Try our online demo at https://naruya.github.io/gaussian-vrm/
For more details, check out our project page at https://gaussian-vrm.github.io/
🎉 We’ve released sample avatars and original scan data!
Check out the Sample Avatars section below for details.
Gaussian-VRM (GVRM) is a three.js implementation of Instant Skinned Gaussian Avatars. GVRM is built on top of three-vrm and gaussian-splats-3d. GVRM can handle avatars as standard three.js objects, allowing you to directly reuse VRM format avatar operations (such as movement and animations).
For detailed usage instructions, please check 👉 Gaussian-VRM Examples 👈
// GVRM
const gvrm = await GVRM.load('./assets/author.gvrm', scene, camera, renderer); // 1/3
await gvrm.changeFBX('./assets/Idle.fbx'); // 2/3
renderer.setAnimationLoop(() => {
gvrm.update(); // 3/3
renderer.render(scene, camera);
});The three steps are:
- Load GVRM file
- Change animation
- Update every frame
That's all! (Super easy!) The full JavaScript code can be seen below:
import * as THREE from 'three';
import { GVRM } from 'gvrm';
const canvas = document.getElementById('canvas');
const renderer = new THREE.WebGLRenderer({ canvas });
renderer.setSize(640, 480);
const scene = new THREE.Scene();
const camera = new THREE.PerspectiveCamera(65, 640 / 480, 0.01, 100);
camera.position.set(0, 0.4, 1.5);
// GVRM
const gvrm = await GVRM.load('./assets/author.gvrm', scene, camera, renderer); // 1/3
await gvrm.changeFBX('./assets/Idle.fbx'); // 2/3
renderer.setAnimationLoop(() => {
gvrm.update(); // 3/3
renderer.render(scene, camera);
});For more usage, please check 👉 Gaussian-VRM Examples 👈
(coming soon!)
Six sample avatars and original scan data are available on Google Drive:
🔗 Sample Avatars (Google Drive)
These avatars are released under the MIT License,
as long as they are not used in ways that violate public order or morality.
If you would like to use animation files with the sample avatars, the easiest way is to download them from Mixamo.
Recommended download settings:
- Format: FBX ASCII (.fbx)
- Skin: Without Skin
- Frames per Second: 60
- Keyframe Reduction: None
-
Source Code This repository's source code is licensed under the MIT License.
👉 Unlike other related research, this work does not use SMPL, any deep learning models,
or mesh optimizers with restrictive licenses — therefore, it can be released under the MIT License! 🎉🎉 -
Assets (
./assetsand./examples/assets/directories) The files under these directories are not covered by the MIT License.
They are provided solely for research purposes and may not be used, modified, or redistributed without explicit permission.Certain sample avatars are separately provided under the MIT License — see the Sample Avatars section above for details.
This work was supported by the Ochiai Pavilion at the Osaka/Kansai Expo 2025.
This work was supported by JSPS KAKENHI Grant Number 23KJ0284.
The VRM model in this repository is freely usable for any purpose, except standalone redistribution of the original, unmodified model. The model can be found at JOKER's store. Thanks!
@misc{kondo2025instantskinnedgaussianavatars,
title={Instant Skinned Gaussian Avatars for Web, Mobile and VR Applications},
author={Naruya Kondo and Yuto Asano and Yoichi Ochiai},
year={2025},
eprint={2510.13978},
archivePrefix={arXiv},
primaryClass={cs.CG},
url={https://arxiv.org/abs/2510.13978},
}


