TRELLIS
TRELLIS is a revolutionary AI model developed by Microsoft Research that generates high-quality 3D assets from text descriptions or single 2D images using a novel Structured Latent Diffusion architecture.
Key Highlights
3D Model Generation from Text and Image
Capability to create detailed and textured 3D models from both text descriptions and a single image.
High-Detail Textured Meshes
Produces industry-standard textured mesh outputs with PBR material support for professional-ready models.
Multi-View Consistency
Advanced multi-view algorithm ensuring the generated 3D model looks consistent and correct from every angle.
Wide Format Support
Exports in common 3D formats like GLB, OBJ, and FBX for compatibility with Blender, Unity, and Unreal.
About
TRELLIS is a revolutionary AI model developed by Microsoft Research that generates high-quality 3D assets from a single 2D image. Released in December 2024, TRELLIS achieves groundbreaking results in image-to-3D conversion and offers rapid prototyping capabilities particularly suited for game development, product design, and virtual reality applications. The model provides significant advantages over other solutions in the field in both speed and quality.
TRELLIS's technical architecture is built on Structured LATents (SLAT), a novel representation format that encodes 3D geometry, texture, and rendering information in a structured latent space for efficient and high-quality generation. The training process utilizes Objaverse and similar large-scale 3D datasets. The diffusion-based generation pipeline follows a multi-stage process working from image to SLAT representations and then to mesh, texture, and radiance field outputs. The model supports two different input modes: text-to-3D and image-to-3D generation.
In terms of performance, TRELLIS delivers impressive results. Image-to-3D generation completes in approximately 12 seconds on an A100 GPU, representing a significant speed improvement compared to InstantMesh's 30-second duration. An F-Score of 0.473 is achieved on the GSO dataset, representing a notable improvement when compared to One-2-3-45's score of 0.311. Generated 3D models can be output in mesh, texture map, and radiance field formats.
TRELLIS finds applications in game development pipeline asset generation, e-commerce 3D product visualization, architectural visualization, virtual and augmented reality content creation, and digital twin generation. Its fast generation time provides significant time savings in iterative design processes. It offers an efficient solution particularly for projects requiring high-volume 3D content generation.
TRELLIS is available as open-source under the MIT license. Model weights, training code, and inference pipeline are accessible via GitHub. Built on PyTorch, it is optimized for NVIDIA GPUs. Demos and pre-trained models are available through Hugging Face. A user-friendly Gradio interface provides a browser-based 3D generation experience.
TRELLIS is a significant work demonstrating the power of structured latent representations in single-image 3D generation. Compared to Wonder3D's cross-domain attention approach and SPA3D's point cloud alignment technique, TRELLIS holds a competitive position in both speed and output quality. Microsoft Research's strong research infrastructure reflects the model's technical depth and continuous development potential. Its generation speed and multi-format output support make TRELLIS an ideal choice for professional workflows.
Looking more closely at TRELLIS's technical innovations, the advantages offered by the SLAT (Structured Latent) representation format over other approaches in the field become more apparent. SLAT encodes 3D space on a structured voxel grid, preserving both local geometric details and overall structural coherence. This representation format enables the diffusion model to operate effectively in 3D space and enhances both mesh quality and texture details of generated models. The model's multi-output format support is a significant advantage: users can obtain mesh, Gaussian splatting, and radiance field outputs from the same generation and select the format that suits their needs. TRELLIS also supports conditional generation, allowing guidance through text prompts or reference images. Microsoft Research's active development process on TRELLIS means continuous improvements and new feature additions. The project's high star count on GitHub and community participation serve as concrete indicators of the model's impact in the field, reflecting strong adoption among researchers and practitioners alike.
Use Cases
Game Asset Creation
Accelerating design iterations by creating rapid 3D asset prototypes during game development.
E-Commerce 3D Product Images
Creating 360-degree viewable 3D models of products for online stores.
Architectural Visualization
Creating prototypes for rapid 3D modeling and visualization of architectural concepts.
Education and Simulation
Producing rapid 3D object and scene models for educational materials and simulation environments.
Pros & Cons
Pros
- Innovative 3D generation with Microsoft's SLAT (Structured Latent) representation
- Output as Radiance Fields, 3D Gaussians, and mesh from single image
- Trained on 500K+ high-quality 3D models
- Research accepted as CVPR 2025 Spotlight
- PBR materials, transparency, and detailed texture support
Cons
- Requires Linux and minimum 24GB GPU memory
- H100 GPU recommended for full capacity TRELLIS.2
- Setup and operation require technical knowledge
- Not yet fast enough for real-time 3D generation
Technical Details
Parameters
Unknown
Architecture
Structured Latent Diffusion
Training Data
Objaverse
License
MIT
Features
- Text-to-3D
- Image-to-3D
- Textured meshes
- High detail
- GLB export
- Multi-view consistency
- PBR materials
Benchmark Results
| Metric | Value | Compared To | Source |
|---|---|---|---|
| Üretim Süresi (Image-to-3D) | ~12 saniye (A100) | InstantMesh: ~30 saniye | TRELLIS Paper (arXiv:2412.01506) |
| F-Score (GSO Dataset) | 0.473 | CRM: 0.402 | TRELLIS Paper |
| Novel View PSNR | 22.8 dB | LGM: 20.5 dB | Papers With Code |
| Mesh Kalitesi (Chamfer Distance) | 0.034 | TripoSR: 0.048 | TRELLIS Paper |
Available Platforms
Frequently Asked Questions
Related Models
InstantMesh
InstantMesh is a feed-forward 3D mesh generation model developed by Tencent that creates high-quality textured 3D meshes from single input images through a multi-view generation and sparse-view reconstruction pipeline. Released in April 2024 under the Apache 2.0 license, InstantMesh combines a multi-view diffusion model with a large reconstruction model to achieve both speed and quality in single-image 3D reconstruction. The pipeline first generates multiple consistent views of the input object using a fine-tuned multi-view diffusion model, then feeds these views into a transformer-based reconstruction network that predicts a triplane neural representation, which is finally converted to a textured mesh. This two-stage approach produces significantly higher quality results than single-stage methods while maintaining generation times of just a few seconds. InstantMesh supports both text-to-3D workflows when combined with an image generation model and direct image-to-3D conversion from photographs or artwork. The output meshes include detailed geometry and texture maps compatible with standard 3D software and game engines. The model handles a wide variety of object types including characters, vehicles, furniture, and organic shapes with good geometric fidelity. As an open-source project with code and weights available on GitHub and Hugging Face, InstantMesh has become a popular choice for developers building 3D asset generation pipelines. It is particularly useful for game development, e-commerce product visualization, and rapid prototyping scenarios where fast turnaround and reasonable quality are both important requirements.
LGM
LGM (Large Gaussian Model) is a 3D generation model developed by researchers at Peking University that produces high-quality 3D objects from single images or text prompts in approximately five seconds using 3D Gaussian Splatting representation. Released in 2024 under the MIT license, LGM combines multi-view image generation with Gaussian-based 3D reconstruction in an end-to-end framework. The model first generates multiple consistent views of the target object using a multi-view diffusion backbone, then a U-Net-based Gaussian decoder predicts 3D Gaussian parameters from these views to construct the full 3D representation. Unlike mesh-based approaches, the Gaussian Splatting output enables real-time rendering with high visual quality including accurate lighting, transparency, and reflective surface effects. LGM supports resolutions up to 512 pixels for the generated views and produces detailed 3D content with clean geometry and vivid textures. The model can be used for both image-to-3D conversion from photographs and text-to-3D generation when paired with a text-to-image model as a front end. As an open-source project with code and pre-trained weights available on GitHub, LGM is accessible to researchers and developers for both academic study and practical applications. The model is particularly suited for interactive 3D visualization, virtual reality content, game asset prototyping, and any scenario where real-time rendering of generated 3D content is required. LGM demonstrates that Gaussian Splatting provides a compelling alternative to traditional mesh representations for AI-generated 3D content.
Meshy
Meshy is a proprietary AI-powered 3D generation platform developed by Meshy AI that creates detailed, production-ready 3D models from text descriptions and images. The platform combines text-to-3D and image-to-3D capabilities with advanced AI texturing features, positioning itself as a comprehensive solution for rapid 3D content creation. Meshy uses a transformer-based architecture that generates textured 3D meshes with PBR-compatible materials, making outputs directly usable in game engines like Unity and Unreal Engine without additional processing. The platform offers multiple generation modes including text-to-3D for creating objects from written descriptions, image-to-3D for converting photographs into 3D models, and AI texturing for applying realistic materials to existing untextured meshes. Generated models include proper UV mapping, normal maps, and physically based rendering materials suitable for professional workflows. Meshy provides both a web-based interface and an API for programmatic access, making it accessible to individual artists and scalable for enterprise pipelines. The platform is particularly popular among game developers, animation studios, and AR/VR content creators who need to produce large volumes of 3D assets efficiently. As a proprietary commercial service launched in 2023, Meshy operates on a subscription model with free tier access for limited generations. The platform continuously updates its models to improve output quality, topology optimization, and texture fidelity, competing directly with other AI 3D generation services in the rapidly evolving market.
Meshy v4
Meshy-4 is a historical model version retired for API requests. Check the retirement boundary, migration dependencies and current plan terms before starting new work.
Quick Info
Structured Latent Diffusion