> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openspline.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Run on your GPU

> Install the GPU service and stream your first local avatar.

On Linux or WSL2 with an NVIDIA GPU:

```bash theme={null}
curl -fsSL https://raw.githubusercontent.com/openspline/openspline/main/install.sh | sh
sh ~/.local/share/openspline/run.sh
```

Open [http://localhost:7860](http://localhost:7860) for the demo. Then use the same service from your agent:

```bash theme={null}
pip install openspline
```

```python theme={null}
from openspline import Openspline

async with Openspline().avatar("portrait.jpg", quality="low") as avatar:
    print(avatar.viewer_url)
    await avatar.stream(agent.audio_stream(), sample_rate=24000)
```

Open the printed viewer URL. The default service URL is `http://localhost:7860`.

PyTorch uses CUDA 12.8 for Blackwell and CUDA 12.6 for older NVIDIA GPUs. AMD and Apple GPUs are unsupported. See [GPU workers](/reference/gpu-workers) for multiple GPUs and High quality.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.