TripoSR GitHub Install Helper
Build a local TripoSR install script from the official GitHub workflow. Choose your OS, Python environment, PyTorch/CUDA channel, VRAM tier, and output format, then copy the exact command plan.
TriposRai is not affiliated with Tripo AI, Stability AI, or VAST. This page translates the official TripoSR GitHub setup flow into practical local commands for developers. Always verify the upstream repository before shipping production code. Read the disclaimer.
Quick Answer
Official local setup in five commands
git clone https://github.com/VAST-AI-Research/TripoSR.git cd TripoSR python -m venv .venv && source .venv/bin/activate pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121 pip install --upgrade setuptools && pip install -r requirements.txt Interactive setup builder
Generate a TripoSR GitHub install plan
Select your local environment settings below. The helper will automatically generate your customized installation script, a first-run inference command, and diagnostic tips tailored to your hardware.
Environment Setup
Inference Configuration
Official workflow map
What the generated script does
Clone the source
The helper uses the official VAST-AI-Research/TripoSR repository, then moves into the project directory before installing dependencies.
Match PyTorch and CUDA
The most important dependency choice is the PyTorch wheel. The official README warns that your local CUDA major version should match the PyTorch CUDA build.
Install requirements
TripoSR depends on packages such as omegaconf, transformers, trimesh, rembg, xatlas, and torchmcubes.
Run one image
The official demo command is python run.py examples/chair.png --output-dir output/. Replace the image path after the first successful run.
VRAM Logic
Choose chunk size by hardware, not guesswork
The upstream script exposes --chunk-size for surface extraction and rendering. Smaller chunk sizes reduce VRAM usage and can rescue small GPUs, but they slow down computation.
--chunk-size 4096 Lower peak memory first; speed is secondary.
--chunk-size 8192 Matches the official default and the README's roughly 6GB baseline.
--chunk-size 8192 Keep the default unless you hit OOM on large input batches.
torchmcubes CUDA support error
If you see torchmcubes was not compiled with CUDA support, the official troubleshooting path is to uninstall it, upgrade build tools, and compile it directly from the tatsy/torchmcubes repository:
pip uninstall torchmcubes
pip install --upgrade setuptools
pip install git+https://github.com/tatsy/torchmcubes.gitIf it still fails, confirm the installed PyTorch CUDA wheel and your local CUDA major version are aligned.
Output choices
OBJ, GLB, texture, and Blender handoff
Use OBJ for inspection and cleanup
OBJ is the default. Pick it when the next stop is Blender, MeshLab, retopology, or manual material cleanup.
Use GLB for web previews
Add --model-save-format glb when you want a compact handoff to web viewers, product previews, or game-engine import tests.
Bake texture only when needed
--bake-texture exports a texture atlas instead of relying on vertex colors. It is useful for pipelines that require texture files.
Primary sources
Official TripoSR references
- VAST-AI-Research/TripoSR GitHub repository
Official codebase with setup notes, CLI arguments, demo command, troubleshooting, and MIT license reference.
- stabilityai/TripoSR on Hugging Face
Official model card and checkpoint source for the pretrained TripoSR weights.
- TripoSR technical report
Research paper for the single-image reconstruction model, architecture context, and performance claims.
Next workflow step
After the GitHub install works
FAQ
TripoSR GitHub install questions
What is the official TripoSR GitHub install command?
Clone https://github.com/VAST-AI-Research/TripoSR.git, create a Python environment, install a platform-matched PyTorch build, upgrade setuptools, then run pip install -r requirements.txt.
How much VRAM does the default TripoSR GitHub demo use?
The official README says the default single-image run takes about 6GB VRAM. Smaller chunk sizes reduce VRAM usage but increase computation time.
What command runs a local TripoSR image-to-3D test?
Use python run.py examples/chair.png --output-dir output/ for the official demo image, or replace examples/chair.png with your own input file path.
Can TripoSR export GLB instead of OBJ?
Yes. The official run.py script supports --model-save-format glb. The default output format is obj.
What causes the torchmcubes CUDA error?
The common torchmcubes error usually means torchmcubes was compiled without matching CUDA support. The official README recommends matching your local CUDA major version to the PyTorch CUDA build, upgrading setuptools, uninstalling torchmcubes, and reinstalling it from the tatsy/torchmcubes GitHub repository.