TripoSG-scribble Open-source 3D Model Generation Tool - Drive Creative Design and Prototype Building with Scribbles and Text Prompts

Triposg Scribble

Developed by VAST-AI

TripoSG-scribble is an AI tool that rapidly generates 3D models from scribble images and text prompts. As a variant of TripoSG, it is suitable for creative design and rapid prototyping.

3D Vision Open Source License:MIT #Scribble to 3D #Rectified Flow Transformer #High-Fidelity 3D Generation

Downloads 129

Release Time : 4/16/2025

Model Overview

This model integrates a rectified flow transformer and cross-attention mechanisms to convert scribble sketches and text descriptions into high-quality 3D meshes, supporting various applications such as creative design and game asset generation.

Model Features

Rectified Flow Transformer

Utilizes a rectified flow (RF) transformer architecture to achieve stable linear trajectory modeling, improving the accuracy and efficiency of 3D generation.

Image Feature Conditioning

Processes image feature conditions through cross-attention mechanisms, optimizing the integration of scribbles and text prompts.

CFG Distillation Technique

Employs CFG distillation to enhance inference efficiency. When using, set CFG=0 for significantly improved generation speed.

Model Capabilities

Scribble to 3D Model

Text-Prompted 3D Generation

High-Fidelity 3D Shape Generation

Use Cases

Creative Design & Art

Rapid Prototyping

Quickly generate 3D prototypes from simple scribbles and text descriptions, accelerating the design process.

Produces high-quality 3D meshes suitable for further refinement.

Gaming & VFX

Game Asset Generation

Rapidly generate 3D models and VFX assets for games, improving development efficiency.

Creates high-fidelity 3D models compatible with game engines.

Product Design

Product Visualization

Convert product sketches into 3D models for display and testing.

Quickly generates interactive 3D product prototypes.

🚀 TripoSG-scribble - Fast 3D Shape Prototyping with Scribble and Prompt

TripoSG-scribble can convert a scribble image and a text prompt into a 3D shape. It is a variant of TripoSG, which is a state-of-the-art image-to-3D generation foundation model. TripoSG uses large-scale rectified flow transformers to generate high-fidelity 3D shapes from single images.

✨ Features

Convert a scribble image and a text prompt to high-quality 3D meshes.
Suitable for creative and design applications, gaming and VFX asset creation, prototyping and visualization.

📦 Installation

The README doesn't provide specific installation steps, so this section is skipped.

💻 Usage Examples

Basic Usage

For detailed usage instructions, please visit our GitHub repository.

📚 Documentation

Model Description

Model Architecture

TripoSG utilizes a novel architecture combining:

Rectified Flow (RF) based Transformer for stable, linear trajectory modeling
Advanced VAE with SDF-based representation and hybrid geometric supervision
Cross-attention mechanism for image feature condition
1.5B parameters operating on 2048 latent tokens

For inference efficiency, TripoSG-scribble is different from TripoSG in:

TripoSG-scribble is a CFG-distilled model and should be used with CFG=0
TripoSG-scribble is trained with 512 latent tokens

Intended Uses

This model is designed for:

Converting scribble image and text prompt to high-quality 3D meshes
Creative and design applications
Gaming and VFX asset creation
Prototyping and visualization

Requirements

CUDA-capable GPU (>8GB VRAM)

🔧 Technical Details

TripoSG utilizes a novel architecture combining a Rectified Flow (RF) based Transformer for stable, linear trajectory modeling, an advanced VAE with SDF-based representation and hybrid geometric supervision, a cross-attention mechanism for image feature condition, and 1.5B parameters operating on 2048 latent tokens. For inference efficiency, TripoSG-scribble is a CFG-distilled model (should be used with CFG = 0) and is trained with 512 latent tokens.

📄 License

This project is under the MIT license.

🔗 About

TripoSG-scribble is developed by Tripo, VAST AI Research, pushing the boundaries of 3D Generative AI. For more information:

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご