DEV Community

Cover image for TRELLIS is still the lead Open Source AI model to generate high-quality 3D Assets from static images
Furkan Gözükara
Furkan Gözükara

Posted on

1 1 1 1 1

TRELLIS is still the lead Open Source AI model to generate high-quality 3D Assets from static images

TRELLIS is still the lead Open Source AI model to generate high-quality 3D Assets from static images — Some mind blowing examples — Supports multi-angle improved image to 3D as well — Works as low as 6 GB GPUs

![Image description](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/xip0xv8ahe0yqmgpchrd.png) ![Image description](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/gb96suqrsutn6it6wq46.gif) ![Image description](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/3mk5f8h20m11l5lc4yfy.gif) ![Image description](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/1d7lwuq56d8mc7hk8zzw.gif)

TRELLIS is a large 3D asset generation model. It takes in text or image prompts and generates high-quality 3D assets in various formats, such as Radiance Fields, 3D Gaussians, and meshes. The cornerstone of TRELLIS is a unified Structured LATent (SLAT) representation that allows decoding to different output formats and Rectified Flow Transformers tailored for SLAT as the powerful backbones. We provide large-scale pre-trained models with up to 2 billion parameters on a large 3D asset dataset of 500K diverse objects. TRELLIS significantly surpasses existing methods, including recent ones at similar scales, and showcases flexible output format selection and local 3D editing capabilities which were not offered by previous models.

Paper : https://arxiv.org/abs/2412.01506

FEATURES

  • High Quality: It produces diverse 3D assets at high quality with intricate shape and texture details.

  • Versatility: It takes text or image prompts and can generate various final 3D representations including but not limited to Radiance Fields, 3D Gaussians, and meshes, accommodating diverse downstream requirements.

  • Flexible Editing: It allows for easy editings of generated 3D assets, such as generating variants of the same object or local editing of the 3D asset.

  • TRELLIS-image-large Large image-to-3D model 1.2B — we are using this

VIDEO CHAPTERS

  • 0:00 Showcase: Single Image to Stunning 3D Model with Trellis

  • 0:26 Introducing Trellis: The Ultimate Image-to-3D Generator by Microsoft

  • 0:59 Trellis Advanced Features: Multi-Angle Input & Custom Gradio App

  • 1:37 GPU Compatibility & Enhanced Gradio App Functionality

  • 2:00 Gradio App Deep Dive: Generations, Presets, Video & Mesh Control

  • 2:48 Batch Processing, Metadata, & One-Click Installer with Flash Attention

  • 3:17 VRAM Modes (High/Low) & Project Acknowledgements

  • 3:42 One-Click Installer: Development Effort & Library Compilation

  • 4:20 How to Get Trellis: Download & Pre-Installation Steps

  • 4:45 Installation Guide: Folder Setup & Running the .bat File

  • 5:23 Automated Installation Process: Libraries, App & Model Downloads

  • 5:54 Installation Complete: Model Download & Folder Size (13.5GB)

  • 6:56 Running Trellis: Launch Options for Different GPUs & VRAM

  • 7:22 First Launch: App Initialization & GPU Monitoring (nvitop)

  • 8:08 Using Trellis: Uploading Image & Auto Background Removal

  • 8:34 Core Generation Settings: Seed, Steps, Guidance, Mesh Detail

  • 9:17 Example 1 (Car Image): Generation & Output Analysis

  • 9:42 Optimizing Output: Better Image & Higher Texture Resolution

  • 10:02 Behind the Scenes: Generation Stages & Output File Management

  • 10:34 Showcase: High-Detail Model from Optimized Settings (Lion)

  • 11:35 Enhancing Quality with Multi-Angle Image Input

  • 11:53 Local Performance Check & Cloud Deployment (RunPod, MassedCompute)

  • 12:46 Result: Multi-Angle Input Generation & Final Remarks

  • 13:07 How to Get Trellis & Support Information

Sentry image

Make it make sense

Only get the information you need to fix your code that’s broken with Sentry.

Start debugging →

Top comments (0)

Build gen AI apps that run anywhere with MongoDB Atlas

Build gen AI apps that run anywhere with MongoDB Atlas

MongoDB Atlas bundles vector search and a flexible document model so developers can build, scale, and run gen AI apps without juggling multiple databases. From LLM to semantic search, Atlas streamlines AI architecture. Start free today.

Start Free

AssemblyAI Challenge (ends July 27, $3k prizes)

Build with Universal-Streaming, AssemblyAI's most advanced real-time transcription API and win from our $3,000 prize pool! Universal-Streaming is ultra fast (300ms latency!), ultra accurate, and offers intelligent endpointing to keep conversations flowing naturally.

Check out the challenge

DEV is bringing live events to the community. Dismiss if you're not interested. ❤️