# TripoSG — card for AI agents

1.5B rectified-flow model that turns a single image into a detailed 3D mesh.

Category: AI 3D model generators · maker VAST AI Research (Tripo) · checked 2026-10-25

## Price
- Price: Free
- Free tier: Open weights, free to download.

## License
MIT — https://github.com/VAST-AI-Research/TripoSG/blob/main/LICENSE

- commercial: **yes** (Commercial use OK)
  > to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  (https://github.com/VAST-AI-Research/TripoSG/blob/main/LICENSE)
- attribution: **no** (No credit needed)
  > The above copyright notice and this permission notice shall be included in all
  (https://github.com/VAST-AI-Research/TripoSG/blob/main/LICENSE)
- in app: **yes** (Can ship inside your app)
  > in the Software without restriction, including without limitation the rights
  (https://github.com/VAST-AI-Research/TripoSG/blob/main/LICENSE)

## Rules to follow
- Keep the MIT license file with the code or weights you ship.
- Needs a CUDA GPU with more than 8 GB of VRAM (model card).

## How to connect
- Download only
- Repository: https://github.com/VAST-AI-Research/TripoSG
- Website: https://huggingface.co/VAST-AI/TripoSG

## Notes
Hugging Face API, Oct 25, 2026: 8,874 downloads, 152 likes; GitHub VAST-AI-Research/TripoSG: 1,860 stars.

## Numbers (AgentGid data)
- Gid Score: 15
- GitHub stars: 1,924

## Lines to paste into your prompt or AGENTS.md

```
- TripoSG (AI 3D model generators). 1.5B rectified-flow model that turns a single image into a detailed 3D mesh. Price: free; commercial use OK; no credit needed.
  Connect: download from https://huggingface.co/VAST-AI/TripoSG (no API).
  Rule: Keep the MIT license file with the code or weights you ship.
  Rule: Needs a CUDA GPU with more than 8 GB of VRAM (model card).
  License: https://github.com/VAST-AI-Research/TripoSG/blob/main/LICENSE (checked 2026-10-25; details: https://agentgid.com/assets/triposg/)
```

## Links
- AgentGid page: https://agentgid.com/assets/triposg/
- Is it free for commercial use? https://agentgid.com/assets/triposg/commercial-use/
- Source: https://huggingface.co/VAST-AI/TripoSG
- Source: https://github.com/VAST-AI-Research/TripoSG
- Source: https://github.com/VAST-AI-Research/TripoSG/blob/main/LICENSE
- Source: https://huggingface.co/api/models/VAST-AI/TripoSG


## TripoSG: a detailed 3D shape from one picture

TripoSG is a free AI model that turns one picture of an object into a 3D model file. It works with photos, cartoons and sketches, and gives a shape without colors. It runs on your own computer and needs an NVIDIA graphics card with at least 8 GB of memory.

### Where to use it
- A 3D shape from a product photo: Turn one picture of a chair, toy or gadget into a GLB file you can open in Blender or a game engine.
- Props for games and visual effects: Make 3D stand-ins or base shapes for game props and effects shots from concept art, then paint and tune them in your 3D tool.
- A lighter model for the web: Ask for fewer triangles so the file stays small enough for a web page or a phone game.
- A 3D sketch from a doodle and words: Draw a rough doodle, add a short text like "a cat with wings", and get a quick 3D draft to try ideas.
- Shapes for 3D printing drafts: Make a first shape from a drawing of a figurine or part, then fix and close it in a 3D tool before printing. (idea, not from the official pages)
- Turnable items on a portfolio page: Show the GLB file on a page with a 3D viewer so visitors can turn the object with the mouse. (idea, not from the official pages)

### How to start
- Get the code from GitHub, make a Python 3.10 environment, install PyTorch for your CUDA version, then the project requirements.

```
git clone https://github.com/VAST-AI-Research/TripoSG.git
cd TripoSG
pip install -r requirements.txt
```

- Run the script on a picture. The model files are downloaded from Hugging Face the first time.

```
python -m scripts.inference_triposg --image-input assets/example_data/hjswed.png --output-path ./output.glb
```


### License
The TripoSG code and model are under the MIT license, so paid use is allowed if you keep the license notice with copies. The background cutter the script downloads, RMBG-1.4, is free only for non-commercial use.

### Good to know
- You need an NVIDIA graphics card with at least 8 GB of memory. The script is set up for such a card only.
- The result is a shape only, without colors or textures. Paint it in a 3D tool or use a model that adds textures.
- The script also downloads RMBG-1.4 to cut the object out of the photo. That model is for non-commercial use unless you buy a license from BRIA.
- It was trained on 2 million picked image and shape pairs, as the makers report.
- There is no paid service or key in this project; the makers link free online demos on Hugging Face.

Checked 2026-11-12 against: https://github.com/VAST-AI-Research/TripoSG, https://github.com/VAST-AI-Research/TripoSG/blob/main/README.md, https://github.com/VAST-AI-Research/TripoSG/blob/main/LICENSE, https://github.com/VAST-AI-Research/TripoSG/blob/main/requirements.txt, https://github.com/VAST-AI-Research/TripoSG/blob/main/scripts/inference_triposg.py, https://github.com/VAST-AI-Research/TripoSG/blob/main/scripts/image_process.py, https://huggingface.co/VAST-AI/TripoSG, https://huggingface.co/VAST-AI/TripoSG/raw/main/README.md, https://huggingface.co/api/models/VAST-AI/TripoSG, https://huggingface.co/api/models/VAST-AI/TripoSG-scribble, https://huggingface.co/briaai/RMBG-1.4/raw/main/README.md

### Mistakes to avoid
- Don't: Installing the default PyTorch from PyPI and running on a machine without an NVIDIA card → Do: Install the PyTorch build for your CUDA version from download.pytorch.org and run on a CUDA GPU with 8 GB or more (The script puts the models on "cuda" and the README asks for the matching CUDA build.)
- Don't: Shipping paid work made with the default script without thinking about RMBG-1.4 → Do: Give it pictures with a clean transparent background so the cut-out step is not used, or get a commercial RMBG license (The script downloads and loads RMBG-1.4, which is non-commercial; a picture with a usable transparent layer skips that model.)
- Don't: Running the script from inside the scripts folder (python inference_triposg.py) → Do: Run it from the project root as a module, python -m scripts.inference_triposg (The README runs it as a module from the root so imports and the pretrained_weights/ folder resolve.)
- Don't: Putting the raw GLB with hundreds of thousands of faces on a web page → Do: Pass --faces (for example 5000) to shrink it while it is made (Without --faces (default -1) the mesh is not reduced.)

### Questions
- Does it give colors or textures? No. TripoSG makes the shape only; textures come from another tool or model.
- Can I run it on a Mac or a laptop without NVIDIA? The makers list a CUDA graphics card with at least 8 GB of memory as the requirement, and the script is written for such a card.
- Can I use the 3D models in a paid game? The TripoSG code and model are MIT licensed. For paid work, cut out the background yourself or get a commercial RMBG license, because RMBG-1.4 is non-commercial.
- Which file does it save? A GLB file at the path you give with --output-path (./output.glb by default).
- Does it work with drawings, not only photos? Yes. The makers say it handles photos, cartoons and sketches, and a separate scribble model takes a doodle plus a text prompt.

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.
