Hunyuan 3D Pro - Image to 3D Model

Turn any photo of an object into a production-ready 3D model with Hunyuan 3D Pro.

Upload a clean image of the object, run the workflow, and get back a textured GLB file, a flat thumbnail render, an interactive 3D viewer, and an auto-rotating turntable preview. The model is ready for game engines, AR pipelines, 3D printing, or downstream Blender work.

Optional: feed up to seven additional reference views (back, sides, top, bottom, front quarters) for sharper geometry on complex objects.

How do you convert an image to a 3D model with Hunyuan 3D Pro?

Upload a photo of the object to the Load Image node. Single-subject shots on a neutral background work best. Run the workflow. After 30 to 60 seconds, you get a GLB file, a thumbnail PNG, an interactive 3D viewer, and an auto-rotating turntable. Copy the fal.media URL to download the GLB to your computer.

Load Image Your source photo. Square aspect ratio works best. Aim for a clean shot of one object with even lighting and a neutral background. Avoid harsh shadows on the subject, busy backgrounds, multiple objects in frame, or people standing behind the subject. The model latches on to whatever it sees most clearly, so a clean reference image makes a big difference.

generate_type Default Normal handles standard mesh generation for most cases. Switch to other modes only if you have a specific reason (high-poly hero asset, low-poly mobile target, etc.).

enable_pbr Off by default for speed. Turn on if you need PBR-textured materials for game engines or AR pipelines. Slower (about 90 seconds vs 30 to 60) but the materials look correct under any lighting setup downstream.

face_count 500,000 polys default. Drop to 100,000 for lightweight web or mobile assets, bump to 1,000,000 for hero shots and detailed photogrammetry-style models.

Multi-view inputs (optional) Seven unconnected slots on the Hunyuan node: back_image, left_image, right_image, top_image, bottom_image, left_front_image, right_front_image. Wire in additional Load Image nodes if you have photos from multiple angles. Even 2 to 3 reference views improve reconstruction quality noticeably.

What is Hunyuan 3D Pro good for?

Generating game assets, AR product visualizations, and 3D-printable models from a single reference image. Strong when you need watertight meshes for printing, clean topology for game engines, or quick concept models for previz. The GLB exports cleanly to Blender, Unity, Unreal, and any standard 3D pipeline.

Game developers use this to generate inventory items, props, environment pieces, and background characters without commissioning custom 3D work. AR and VR teams turn product photos into 3D viewables for shopping apps and product pages. 3D printing hobbyists convert reference photos of figurines, toys, or design objects into printable STLs (export the GLB through Blender or an online converter first).

For hero assets that need pristine topology or specific UV layouts, you'll still want a human modeler. Hunyuan 3D Pro produces clean meshes but not always the optimized quad topology a AAA game studio pipeline expects. Treat it as a high-quality starting point that saves hours of blockout work, not a complete replacement for traditional modeling.

Best results come from clean reference photos. Studio-style shots with a single subject and neutral background give the model the strongest signal. Cluttered scenes, half-occluded subjects, or images with people behind the target object all degrade output quality.

FAQ

How long does Hunyuan 3D Pro take to generate a model? Around 30 to 60 seconds for a standard mesh, depending on face count. Enabling PBR textures adds another 30 seconds or so. The output appears in both the interactive viewer and the turntable animation, with a fal.media download URL displayed in a text node for grabbing the GLB file.

Can I use multi-view images for better quality? Yes. The Hunyuan node has 7 optional reference inputs (back, left, right, top, bottom, left_front, right_front). Wire in extra Load Image nodes with photos from those angles and the reconstruction sharpens noticeably. Even 2 to 3 additional views help. Use evenly-lit shots from consistent camera distance for best results.

How long do the GLB download URLs last? The fal.media URLs that host the GLB file expire after a few hours to a day. Download promptly if you want to keep the model. The local file on Floyo's storage (shown in the Local GLB File Path node) sticks around longer and is what downstream workflows in your Floyo account reference.

What's the difference between Preview3D and Preview3DAnimation? Both display the same GLB. Preview3D is a static interactive viewer where you drag to rotate and scroll to zoom, good for inspecting details. Preview3DAnimation auto-rotates the model like a turntable, good for portfolio screenshots, all-sides quick checks, or recording showcase GIFs without manual interaction.

Can the output be used for 3D printing? Yes, the GLB outputs a watertight mesh suitable for 3D printing. Most printers want STL or OBJ, so you'll need to convert the GLB through Blender or an online tool. Check the mesh in Blender first for any small geometry artifacts before sending to the slicer, since AI-generated meshes occasionally have non-manifold edges.