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.
Key Highlights
FlexiCubes Clean Mesh Extraction
Uses FlexiCubes instead of marching cubes for mesh extraction, producing cleaner topology with more regular face distributions suitable for game engines and animation
Multi-View Consistency Pipeline
Generates consistent multi-view images of the object before 3D reconstruction, ensuring geometric accuracy through cross-view information fusion
UV-Mapped Textured Output
Produces meshes with proper UV mapping and texture images ready for immediate use in standard 3D workflows, game engines, and rendering applications
State-of-Art Open-Source Quality
Represents the current state of the art in open-source single-image 3D reconstruction, exceeding quality of simpler models like TripoSR and Shap-E
About
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 2024, InstantMesh combines a multi-view diffusion model with a large reconstruction model (LRM) and FlexiCubes-based mesh extraction to produce clean, well-textured 3D assets suitable for downstream applications. The model represents a significant leap in quality and usability within the open-source single-image 3D reconstruction space.
The pipeline operates in multiple stages. First, a multi-view diffusion model generates consistent views of the object from multiple angles based on the input image. These generated views are then fed into a sparse-view reconstruction network that processes the multi-view images to produce a 3D representation. Finally, FlexiCubes extraction converts this representation into a high-quality polygonal mesh with UV-mapped textures, producing assets with cleaner topology than methods using marching cubes extraction. Each stage builds upon the output of the previous one, minimizing quality loss from input to output and providing an integrated production process.
InstantMesh's use of FlexiCubes for mesh extraction is a key differentiator. Unlike marching cubes, which produces meshes with irregular topology and many unnecessary vertices, FlexiCubes generates meshes with more regular face distributions and better-defined surface details. This results in meshes that are more suitable for use in game engines, animation pipelines, and 3D printing without extensive post-processing. The differentiable nature of FlexiCubes allows the mesh extraction process to be optimized end-to-end during training, which directly improves final output quality by producing sharper edges and more accurate geometry.
The model produces textured meshes with corresponding UV maps and texture images, providing immediately usable 3D assets. The quality of both geometry and texture typically exceeds that of simpler feed-forward models like TripoSR or Shap-E, though generation takes somewhat longer due to the multi-stage pipeline. Output meshes can be exported in OBJ, GLB, and PLY formats and are fully compatible with standard 3D workflows. Texture maps are generated at sufficient resolution and do not require additional texture work for most applications.
InstantMesh's multi-view generation stage can utilize Zero123++ or similar models, and this modular design allows each component to be independently updated or replaced. The model was trained on the Objaverse dataset and demonstrates strong generalization capability across diverse object categories. It can successfully handle a wide range of inputs from product photography to artistic illustrations, producing consistent results for objects of different styles and complexity levels.
Released under the Apache 2.0 license, InstantMesh is fully open-source with pre-trained weights available on Hugging Face. A live demo on Hugging Face Spaces allows users to try the model without local installation. The model represents the state of the art in open-source single-image 3D reconstruction and is actively used in both research and production environments across the creative industry.
Use Cases
Game Asset Production
Generate game-ready 3D models with clean topology and UV-mapped textures from concept art for rapid asset creation in game development pipelines
3D Content for AR/VR
Create 3D objects from product photos or artwork for augmented reality experiences and virtual reality environments with production-quality mesh output
Digital Twin Generation
Rapidly create 3D digital representations of physical objects from photographs for inventory visualization, documentation, and virtual showcases
Research and Benchmarking
Use as a state-of-the-art baseline for 3D reconstruction research and as a comparison point for evaluating new single-image 3D generation methods
Pros & Cons
Pros
- Creates 3D mesh from a single image in seconds
- Efficient architecture based on LRM (Large Reconstruction Model)
- Open source — accessible on Hugging Face
- Consistent 3D structure with multi-view synthesis
Cons
- Low quality on back faces and unseen areas
- Limited resolution in detailed surface textures
- Difficulty with complex objects (transparent, reflective)
- Irregular mesh topology — not suitable for direct game engine use
Technical Details
Parameters
N/A
License
Apache 2.0
Features
- Single Image to 3D Mesh
- Sparse-View Reconstruction
- Multi-View Generation Pipeline
- FlexiCubes Mesh Extraction
- High-Quality Textured Output
- Open-Source Apache 2.0
- Tencent Research Model
- Hugging Face Spaces Demo
Benchmark Results
| Metric | Value | Compared To | Source |
|---|---|---|---|
| Novel View PSNR | 22.2 dB (GSO) | TripoSR: 21.7 dB | arXiv 2404.07191 |
| Üretim Süresi | ~10 saniye | LGM: ~5 saniye | GitHub InstantMesh |
| SSIM (GSO) | 0.880 | OpenLRM: 0.856 | arXiv 2404.07191 |
| LPIPS (GSO) | 0.125 | TripoSR: 0.138 | arXiv 2404.07191 |
Available Platforms
News & References
Frequently Asked Questions
Related Models
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.
OpenLRM
OpenLRM is an open-source implementation of the Large Reconstruction Model architecture for single-image 3D reconstruction, developed by Zexiang Xu and collaborators. The project provides a fully open and reproducible implementation of the LRM approach, which uses a transformer-based architecture to predict 3D representations from single input images in a feed-forward manner. OpenLRM processes an input image through a pre-trained vision encoder like DINOv2, then feeds the resulting features into a transformer decoder that generates a triplane-based neural radiance field representation, which can be rendered from novel viewpoints or converted to a textured 3D mesh. The entire reconstruction takes only a few seconds on a modern GPU, making it practical for interactive applications and batch processing workflows. Released under the Apache 2.0 license in December 2023, OpenLRM fills a critical gap in the 3D AI research community by providing an accessible reference implementation that researchers can study, modify, and build upon. The model supports various output formats and can be integrated into existing 3D pipelines for applications ranging from game development to e-commerce product visualization. OpenLRM handles diverse object categories including furniture, vehicles, characters, and everyday items with reasonable geometric fidelity. Pre-trained model weights are available on Hugging Face for immediate use. As one of the foundational open-source projects in feed-forward 3D reconstruction, OpenLRM has directly influenced and enabled numerous downstream projects and research efforts in the rapidly evolving single-image 3D generation space.