bunpav

Image to 3D

An image to 3D converter that reconstructs a full mesh from one photo

Drop a product shot or a concept sketch. bunpav reconstructs geometry, UVs, and material from a single frame — no photogrammetry rig, no turntable video.

image → 3dmesh_0203.glb

source: reference-chair.jpg

verts 6,120 · tris 11,880depth reconstructed

how it works

From one photo to an importable mesh

1. Upload a clean reference photo

Even lighting, a plain background, and a three-quarter angle that shows depth all reconstruct better than a flat, front-on shot with harsh shadows.

2. bunpav reconstructs geometry from the frame

The converter infers depth and fills in unseen surfaces to build a full mesh, then unwraps UVs and projects a texture from the source image.

3. Review the back and the edges

The camera never saw the back of the object — that's where the model is guessing. Check occluded surfaces before treating the result as final.

4. Export to your pipeline

GLB for Unity, Unreal, and Blender; FBX or OBJ for other DCC tools; USDZ for Apple AR Quick Look; STL if you're printing the object physically.

who it's for

For anyone who has a reference but not a 3D file

Product mockups

Photograph a physical object and get a clean 3D version back for renders, AR previews, or print-on-demand listings.

Concept sketches

Turn a character turnaround or prop sketch into an initial 3D block-in a modeler can refine, instead of starting from a blank viewport.

Reference-accurate props

Recreate a real-world object in-game — a specific chair, tool, or container — without eyeballing proportions from a photo.

pricing

Same credits, same export range as text-to-3D

No subscription — starting at 9 dollars for 100 credits, shared across both generation modes.

faq

Questions, answered

How does image-to-3D conversion work?

bunpav estimates depth from the photo, infers the geometry the camera couldn't see, and reconstructs a full mesh with UVs and a texture pulled from the source image — all from one frame, no turntable or multi-camera rig required.

What kind of photo works best?

A well-lit, plain-background shot at a three-quarter angle. Avoid harsh shadows, cluttered backgrounds, and pure front-on angles that hide depth cues the reconstruction needs.

Can I convert a concept sketch instead of a photo?

Yes — clean linework with consistent proportions across a turnaround sheet converts well. Heavily rendered or ambiguous shading can confuse depth estimation the same way a bad photo would.

What happens to parts of the object the camera never saw?

The model infers them from context, and that's the least reliable part of any single-image reconstruction. Always check occluded surfaces — the back of a chair, the underside of a prop — before shipping the asset.

What formats can I export the converted model to?

GLB, FBX, OBJ, USDZ, and STL — the same export range as bunpav's text-to-3D generator, covering game engines, DCC tools, AR, and 3D printing.

go deeper

Related reading

launching soon

Upload your first photo before the public launch