Skip to content
SPECIFICTION
Get listed

Crisp3D vs PartCrafter

Crisp3D

crisp3d.ai

PartCrafter

github.com

Side by side

FigureCrisp3DPartCrafter
Price from$2.66/moFree
Free planYesYes
Free trial——
Open source——
WebYes—
Windows—Yes
Linux—Yes
Self-hosted—Yes
Plans published71
Spec score · AI 3D Model Generators8.3 (#7)8.3 (#8)
PurposeCrisp3D turns text prompts, photos, sketches, and concept images into 3D models for printing.PartCrafter generates multiple structured 3D parts and objects from a single RGB image in one pass.

Plans

Crisp3D

  • FreeFree
  • 1,000-credit top-up$22, paid once
  • Pro$31.92 a year
  • Ultra$95.92 a year
  • 5,000-credit top-up$99, paid once
  • 10,000-credit top-up$150, paid once
  • Basic$191.04 a year

Crisp3D pricing →

PartCrafter

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

PartCrafter pricing →

In their own words

Crisp3D

Crisp3D is a web tool for creating 3D models for printing from text prompts, photos, sketches, and concept images. It also offers AI texturing, 3D segmentation, and AI retopology to simplify dense geometry. Users can upload supported existing models to preview colors and finishes, split parts, add joints, or simplify geometry. Listed AI models include Tripo3D, Meshy, Hunyuan3D, and Hi3D; supported slicers include Bambu Studio, OrcaSlicer, Creality Print, ElegooSlicer, and Snapmaker Orca. Exports are available in GLB, FBX, OBJ, STL, and 3MF. The free plan permits one concurrent task and limited AI models; its models are public, non-commercial, and unavailable for download. Paid plans start at $25.90/mo. Paid users can choose private visibility, and commercial use is subject to the applicable license, platform terms, and rights in input materials. Subscription credits refresh each billing cycle and do not carry over. Crisp3D says uploaded files, prompts, and generated outputs are not used to train or improve AI models unless users opt into a separate program.

Visit crisp3d.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