Skip to content
SPECIFICTION
Get listed

Hi3D vs PartCrafter

Hi3D

hi3d.ai

PartCrafter

github.com

Side by side

FigureHi3DPartCrafter
Price fromFreeFree
Free planYesYes
Free trial——
Open source——
WebYes—
Windows—Yes
Linux—Yes
iOSYes—
AndroidYes—
APIYes—
Self-hosted—Yes
Plans published41
Spec score · AI 3D Model Generators8.1 (#12)8.3 (#8)
SupportThe pricing page provides [email protected] for subscription help and lists Discord, email support, and dedicated support among plan support options.The repository invites users to contact the maintainer by email or open an issue with questions or suggestions.

Plans

Hi3D

  • FreeFree
  • MaxPrice not published
  • ProPrice not published
  • UltraPrice not published

Hi3D pricing →

PartCrafter

  • PartCrafter (MIT-licensed open-source software)Free

PartCrafter pricing →

In their own words

Hi3D

Hi3D is an AI image-to-3D tool that turns images into models using Math Magic’s Sparc3D and Ultra3D models. It supports both single-image and multi-view generation, along with text-to-3D, AI texturing, 3D relief generation, and a tool for dividing characters into printable parts. Exports include FBX, GLB, OBJ, STL, and USDZ, and models can be sent directly to Bambu Studio and OrcaSlicer. The wider integration list includes Snapmaker, Creality Print, ElegooSlicer, ComfyUI, and Blender. Developers and enterprises can use its Open Platform API; API operations are priced by model and resolution, with one credit listed at $0.020. The Free plan includes 100 lifetime credits, up to 10 asset generations, one concurrent task, and storage for seven models; those models are public under CC BY 4.0. Paid plans include private and commercial use, and the listed pricing note says paid plans start at $19.90/mo. Hi3D Scan supports video capture or gallery upload for model reconstruction. Platforms include Android, iOS, web, and API.

Visit hi3d.ai

PartCrafter

PartCrafter is a free, open-source model that generates multiple 3D parts and objects from a single RGB image. It can handle individual objects and complex scenes, producing distinct meshes with meaningful part structure without requiring a pre-segmented image. Its architecture represents parts with disentangled latent tokens and uses hierarchical attention to pass information within and across parts. Local inference scripts can remove backgrounds and optionally apply style transfer to adapt real-world photos toward the rendered-image domain used for training. An optional vision-language feature can suggest the number of parts; Gemini is the default provider, and the provider architecture can be extended. Pretrained object-level and scene-level weights are hosted on Hugging Face. Installation is from the GitHub repository in a Python environment, and inference requires a CUDA-enabled GPU with at least 8 GB of VRAM. The linked Hugging Face demo is paused, though its page says users can ask the authors to restart it. The repository lists an MIT license and provides the software as-is, without warranty.

Visit github.com