Image-Blaster is a generative 3D asset pipeline and web viewer that transforms a single input image into a fully interactive 3D scene. The system uses Claude AI agents to orchestrate a multi-step workflow: image analysis and object extraction, background removal (plate generation), world generation via World Labs (gaussian splats, colliders, panoramas), 3D object generation via FAL (Hunyuan3D or Meshy), and sound effect generation via ElevenLabs. All generated assets are stored locally with a deterministic indexed naming convention and request metadata for resumability. The frontend is a React 19 + Three.js (R3F) application with physics (Rapier), post-processing (Bloom, DoF, motion blur), splat rendering (SparkJS), and a full scene editor with undo/redo, transform controls, and asset library. The architecture is a monorepo with a Vite dev server that serves assets and provides a CRUD API for scene projects. While the system is primarily an application, it also contains a data pipeline aspect in its asset generation scripts, which follow a consistent lifecycle of submission, polling, download, and metadata writing. The codebase demonstrates strong engineering practices with reusable patterns like the FAL queue wrapper, indexed file resolution, camera gesture hooks, and error boundaries. However, it is heavily dependent on external cloud APIs (FAL, World Labs, ElevenLabs) and Claude agent tooling, with no offline fallback or mock mode.
TRELLIS.2 is a research codebase for state-of-the-art image-to-3D generation using a novel sparse voxel representation (O-Voxel) and flow matching with 4B parameters. It provides full training and inference pipelines for generating high-fidelity 3D meshes with PBR materials from single images. The repository includes custom CUDA extensions for voxel processing, a flexible sparse tensor library, and comprehensive dataset tooling for large-scale 3D asset curation (Objaverse-XL, Sketchfab). While primarily a research project with training code and data pipelines, it also offers Gradio web demos and Python APIs for inference. The codebase demonstrates strong engineering in its modular design (pipelines, mixins, config-driven training) but suffers from some research-typical issues like code duplication, hardcoded paths, and fragile dependency management.
WorldGen is a Python library and demo application for generating 3D scenes from text prompts or single images in seconds. It orchestrates a pipeline of state-of-the-art models: a fine-tuned FLUX diffusion model to generate 360° panoramic images, DA-2 for monocular depth estimation on equirectangular images, and optionally OneFormer for semantic segmentation and LaMa for inpainting to handle background occlusions. The 3D scene is produced as a Gaussian splatting point cloud (via a custom RGBD-to-Gaussian conversion) or as a textured mesh. An experimental mode uses Apple's Sharp model for higher-quality Gaussian splats by processing cubemap faces. The codebase demonstrates strong engineering practices for a research project: modular pipeline design, use of submodules for dependencies, custom diffusers pipelines with VAE tiling modifications for seamless panorama generation, and a Viser-based interactive viewer. While the core value is the integrated pipeline, several utility modules (equirectangular projection utilities, splat file handling) are cleanly isolated and reusable in other 360° or 3D vision projects.