File size: 3,787 Bytes
6130932 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 |
# MLSharp-TD-Bridge (2D to 3D Gaussian Splatting)
> **One-click 2D to 3D Gaussian Splatting Tool for TouchDesigner | Portable AI Bridge**
> **Author:** GemosDodo
[中文版 (Chinese Version)](./README_zh-CN.md) | [English Version](./README.md)
<img width="1780" height="842" alt="image" src="https://github.com/user-attachments/assets/94273fd8-546f-41e1-b8c9-089944bb0463" />
## Introduction
**MLSharp-TD-Bridge** is a portable AI bridge tool built specifically for TouchDesigner. Based on Apple's **Sharp** model architecture, it achieves **one-click conversion of 2D images into high-quality 3D Gaussian Splatting models directly within TouchDesigner**.
Traditional AI generation often involves tedious Python environment configuration, complex dependency installations (CUDA, PyTorch, etc.), and command-line operations. The goal of this project is to completely "automate" and "black-box" this process.
**Designers do not need to write a single line of code. Simply drag in an image, and obtain 3D assets within seconds.**
## 💾 Download
Due to GitHub file size limits (this project includes a fully portable Python environment), please download the complete project files via **Hugging Face**.
> **Hugging Face (High Speed / Global):** [Click here to visit Repository](https://huggingface.co/GemosDodo/MLSharp-TD-Bridge)
>
> *Please go to the "Files and versions" tab and download the `.zip` file.*
---
## Features
* **Zero-Config & Plug-and-Play**
Includes a complete portable Python environment (containing PyTorch, CUDA, gsplat, sharp, etc.). After downloading and unzipping, there is **no need to install Anaconda**. Just run the `.toe` file directly.
* **Fully Automated Workflow**
With a single button click, the script automatically handles: `Save Image` -> `Clean Old Files` -> `AI Inference` -> `Rename Files` -> `Refresh 3D View`.
* **Native TouchDesigner Integration**
* **Input**: Supports any TOP image source (Movie File In, Noise, etc.).
* **Output**: Generated results are automatically loaded back into a Point File In for immediate preview in the TD viewport.
* 🛠 **Offline Privacy**
All calculations are performed locally on your GPU. No internet connection is required, ensuring data privacy.
## Requirements
* **OS**: Windows 10 / 11 (64-bit)
* **GPU**: NVIDIA GPU (Recommended: 6GB+ VRAM, CUDA support required)
* **Software**: TouchDesigner 2022.x or 2023.x and above
## Getting Started
1. **Download the Project**
Download the project zip file from the **Hugging Face** link above and unzip the entire folder.
> **⚠️ Note**: Please ensure the unzipped file path **does not contain non-English characters** (e.g., `D:\MLSharp_Bridge`) to avoid Python path errors.
2. **Launch the Project**
Double-click to run `MLSharp-TD-Bridge-by-GemosDodo.toe`.
3. **One-Click Generation**
Select the `Sharp_3D_Bridge` component. In the custom parameters panel, click the **`Generate`** button.
4. **View Results**
Wait for approximately 30-60 seconds (depending on GPU performance). The generated 3D model will automatically load into the preview window on the right.
## Acknowledgements
This project utilizes the **TouchDesigner 3D Gaussian Splatting Toolkits** open-sourced by **Yea Chen** to visualize the generated Gaussian Splatting 3D models.
This is my first open-source project, standing on the shoulders of giants. This project is non-profit and free for everyone. I hope you enjoy using it in your TouchDesigner learning or creative process!
* **ML_Sharp GitHub**: [https://github.com/apple/ml-sharp](https://github.com/apple/ml-sharp)
* **TouchDesigner 3D Gaussian Splatting Toolkits**: [https://github.com/yeataro/TD-Gaussian-Splatting](https://github.com/yeataro/TD-Gaussian-Splatting) |