Lightweight WebGL Renderer - Real-time 3D Gaussian Splatting Render Engine for LAM and LAM_Audio2Expression.
Our rendering engine is now available as an NPM package.
๐ Getting Started
-
Install dependencies
-
Run development server
-
Open in browser
http://localhost:5173/
๐ฆ Core Features
โจ Key Highlights
- Real-time rendering support: Interactive LAM avatars with LAM_Audio2Expression facial animations
- NPM package integration: Use with
npm install gaussian-splat-renderer-for-lam
๐ฆ NPM Package Usage
Installation
npm install gaussian-splat-renderer-for-lam
Basic Usage Example
import * as GaussianSplats3D from 'gaussian-splat-renderer-for-lam'; const div = document.getElementById('GaussianRenderer'); const assetPath = './asset/arkit/p2-1.zip'; //the real asset address const render = await GaussianSplats3D.GaussianSplatRenderer.getInstance(div, assetPath);
๐ก Advanced Features
- Facial animations: Load
/asset/test_expression_1s.jsonfor expression control- Custom avatars: Generate new models via LAM and replace ZIP files
๐ Related Projects
copy
| Feature | Repository |
|---|---|
| Chat integration | OpenAvatarChat |
| Avatar generation | LAM |
| Expression generation | LAM_Audio2Expression |
๐บ Demo
The demo shows that LAM_Audio2Expression leverages audio input to generate ARKit blendshapes-driven facial expressions, powering ultra-realistic 3D avatars generated by LAM, and rendered with this render engine.
Gradio_Demo.mp4
The demo highlights a low-latency SDK for real-time interactive chat avatars, developed using the same rendering engine as the core framework.
Chat_Demo.mp4
๐งช Example Files
- Avatar model:
./asset/arkit/p2-1.zip - Expression data:
./asset/test_expression_1s.json
๐ License
This project is licensed under the MIT License.