A 2D to 3D image conversion job starts as one flat picture and has to leave as a mesh you can drop into a game. Searchers asking for conversion software usually want a desktop installer that “makes it solid.” What they get, if the tool is honest, is a single-image reconstruction: the front is measured from the photo, and the back is inferred. This how-to is that path for props — upload or generate the 2D source, run it through 3D Studio, export GLB, GLTF, or FBX, and rig only if the object has to walk. Credit figures below were checked 2026-09-21 in src/lib/threed-models.ts and src/lib/rigging/external-rig/autoRigPricing.ts.
What 2D to 3D image conversion actually asks for
Photogrammetry is the science of measuring physical objects from photographs. Stereophotogrammetry estimates 3D coordinates from two or more images taken from different positions, by triangulating shared points. That is a real multi-photo capture method. A one-file 2D to 3D image converter is a different contract: one viewpoint goes in, a watertight mesh comes out, and anything the camera never saw is a guess.
For a crate, lantern, sword, or potion bottle, that guess is often good enough. The silhouette, the material color, and the readable front face carry the prop. For a hero character, the guess is where ears, backpacks, and the far arm go wrong. Treat 2D to 3D image conversion as a prop factory first. If you need the unseen side to match a design sheet, generate that sheet on purpose — A-pose, plain background, one subject — before you spend the mesh credits.
The public research lines up with that split. Tencent’s Hunyuan3D-2.1 repository describes “From Images to High-Fidelity 3D Assets with Production-Ready PBR Material” (verified 2026-09-21, 4,050 stars). Microsoft’s TRELLIS repo is the paper “Structured 3D Latents for Scalable and Versatile 3D Generation,” a CVPR 2025 Spotlight (verified 2026-09-21, 13,690 stars). Those projects are why a browser can offer 2D to 3D image conversion at all. The picker costs below are why you can budget a jam around it.
The 3D Studio path for 2D to 3D image conversion
3D Studio is badged AI Credits in the tools catalog, not Free. The catalog description (verified 2026-09-21 in src/app/_home-v2/_data/tools.ts) chains AI image generation, single-image to textured 3D, humanoid auto-rigging with weight-paint refinement, and text-to-animation, exported as FBX, GLB, or GLTF. The home-page steps are Generate, Lift to 3D, Auto-rig and weight paint, then text-to-animation and export. For a static prop you stop after the lift and the file download.
The visible image lanes, and what a default run costs, come from getCredits in src/lib/threed-models.ts (verified 2026-09-21). Hidden and admin-only entries are left out of this table.
| Model | Default image cost | Views it accepts | Use it when |
|---|---|---|---|
| Hunyuan 3D 3.1 | 30 credits | image-to-3d only | Best-budget textured prop from one PNG |
| TRELLIS 2 | 40 credits at 1024 (35 at 512, 45 at 1536) | image-to-3d only | Single concept mesh, resolution you can dial |
| Tripo v3.1 | 50 credits at default HD (30 with texture off, 40 at standard) | image, text, or multi-image | You have a turntable, or you want HD color from one photo |
| Meshy 7 | 50 credits textured single image (35 base + 15 texture) | image, text, or multi-image | A-pose or T-pose characters, or a multi-view set |
| Rodin 2.0 | 50 credits | image or text | You need the format enum: GLB, FBX, OBJ, USDZ, or STL |
| Pixal3D | 0 credits, VIP only | image-to-3d only | Unlimited drafts if you already have VIP |
Hunyuan 3D 3.1 is the one marked rating: "best-budget". Pixal3D is credits: 0 with creditLabel: "Free unlimited" and vipOnly: true. Everyone else pays per conversion. New accounts start with 100 free credits (home page and account API docs), and CREDITS_PER_DOLLAR = 100 in src/lib/models.ts, so a 30-credit Hunyuan pass is $0.30 of the grant.
If you want the grant math and the voxel branch, the earlier post Spin Image to 3D Model Free stays on that question. The model-roster post Snap Image to 3D Model AI stays on picking among AI mesh models. This page stays on 2D to 3D image conversion: one picture, one mesh, one export.
Prompt a concept, or upload the picture you already have
Two inputs feed the same converter. If the prop already exists as a sprite, a photo, or a concept PNG, drop that file. If you only have a sentence, make the picture first in AI Image Gen, then convert. The image catalog (verified 2026-09-21) lists Nano Banana Pro, GPT Image 2, Grok Imagine, Seedream 5, Flux 2 Pro, and Z-Image Turbo in one panel at /generate.
A usable source image for converting image to 3d model work has four traits:
- One subject. A second object in frame becomes a fused blob.
- A plain background. Busy scenery gets reconstructed as geometry.
- The whole object in frame, including the base. Cropped feet and cropped hilts come back incomplete.
- For anything you will rig, an A-pose or T-pose with limbs off the torso. Meshy 7 exposes Pose Mode with options
a-poseandt-pose. Rodin 2.0 exposes Force T/A-Pose. Turn that on before you generate, not after a crumpled result.
Prompt versus upload is not a quality ranking. Upload wins when the art direction is already locked. Prompt wins when you do not have a drawing and you need a clean silhouette in minutes. Either way the next click is the same: Image to 3D inside 3D Studio, not a separate 2D to 3D image conversion software install.
Multi-view is the upgrade when one photo is not enough. Meshy 7 and Tripo v3.1 list multi-image-to-3d. Hunyuan 3D 3.1, TRELLIS 2, Pixal3D, and Rodin 2.0 do not. Meshy 7’s getCredits uses 45 as the multi-image base and 35 as the single-image base, then adds 15 when Generate Texture is on. Defaults leave texture on, so a turntable pass is 60 credits and a single textured pass is 50. Ultra Mode adds 10 credits and is hidden for multi-view. Remesh adds 15. 8K texture adds 10; 2K and 4K do not.