InstantMesh: 3D generator — what it does and what you need to run it

InstantMesh is an open source 3D generator created by the Tencent ARC team. The tool tackles the image-to-3D task, meaning it reconstructs a 3D mesh from a single input image. The authors emphasize efficiency and use a sparse-view approach.
What it does
- Generates a 3D mesh from a single image.
- By default, exports the result in .obj format with vertex colors.
- When the --export_texmap flag is specified, it can export the model with a texture map; the authors warn that this will increase processing time.
- The code is written in Python and relies on the diffusers library.
What you need to run it
- Platform: as described by the author, CUDA is required. The exact recommendation is: Python>=3.10, PyTorch>=2.1.0, and CUDA>=12.1.
- Memory and storage: the largest model weights file is 1.6 GB, and all model weights files total 6.8 GB. The authors do not specify system RAM requirements.
- License: both the code and model weights are distributed under the Apache-2.0 license.
- Hosting: the code is available in the repository github.com/TencentARC/InstantMesh, and the model weights are on huggingface.co/TencentARC/InstantMesh. This is a local tool; the authors do not provide a cloud version.
Who it is for
The tool will be useful for developers and researchers who want to integrate 3D generation into their pipeline or fine-tune the model. The open Apache-2.0 license allows using the code and weights in commercial projects. The flag for exporting a texture map gives a choice between speed and detail of the final model.
It is not suitable for those looking for a ready-made cloud solution without environment setup — InstantMesh needs to be run locally on a machine with a suitable GPU and CUDA drivers. Users without experience with Python and PyTorch may also find it difficult to deploy the tool, as the authors do not provide a graphical interface.
InstantMesh occupies the niche of a fast mesh generator from a single snapshot. It does not require a cloud subscription, but it does impose specific hardware and software requirements that are worth checking before installation.
Fact sheet
Repository · Model on HuggingFace
| Task | image-to-3d source |
|---|---|
| Code license | Apache-2.0 source |
| Weights license | apache-2.0 source |
| Platform | CUDA per the author’s description source |
| Largest weights file | 1.6 GB source |
| All weights files | 6.8 GB source |
| Export formats | obj per the author’s description source |
| Library | diffusers source |
| Language | Python source |
| Last code change | 2025-01-03 source |
| Repository created | 2024-04-10 source |
| GitHub stars | 4485 source |
|---|---|
| Forks | 500 source |
| Downloads per month | 18595 source |
| Model updated | 2024-04-11 source |
Values are collected automatically from official sources and were checked on 2026-08-19. Each one links to its source, and values taken from the developer’s pages also carry a verbatim quote — hover over the note. Pricing and versions are shown as of the check date and change most often; verify on the vendor’s site before buying.



