Image-to-3D
A generative pipeline that converts a single 2D image into a textured 3D model (GLB / USDZ). VULK uses TRELLIS and Hunyuan3D to turn product photos and reference images into rotatable WebGL meshes for hero sections.
Image-to-3D
Image-to-3D is the task of reconstructing a textured 3D mesh from a single (or few) input photographs, producing standard formats like GLB, GLTF, OBJ, or USDZ. Modern systems combine multi-view diffusion, sparse-view neural reconstruction (NeRF / Gaussian splatting), and texture baking to output meshes in tens of seconds — usable directly in three.js, React Three Fiber, model-viewer, or AR Quick Look.
In VULK, image-to-3D is wired into the generation agent. When a user uploads a product photo and asks for a "rotating 3D hero", the pipeline calls TRELLIS or Hunyuan3D-2.1, stores the resulting GLB in Cloudflare R2, and emits a React Three Fiber scene that loads the asset with <useGLTF> plus orbit controls and ambient lighting. The model file is decimated for the web (~1-3 MB) and uses Draco / Meshopt compression.
AI Video Background
A short looping video — generated by a text-to-video model — placed behind a website hero section to add cinematic motion without licensed footage. VULK routes across 25+ video models including Veo 3.1, Sora 2 Pro, Runway Gen-4, and Kling 3.
Text-to-3D
Generative pipeline that turns a text prompt into a textured 3D mesh. VULK uses text-to-3D for hero objects, product mockups, and game props — output as GLB and rendered via React Three Fiber.