---
slug: getting_started/installation
title: 🔧 Installation
---

<a id="installation"></a>

Start with [Hardware and runtimes](/FastVideo/getting_started/hardware) to distinguish NVIDIA RTX and
data center GPUs, DGX Spark, Apple Silicon MLX, and AMD ROCm.

The installation paths below cover:

- [NVIDIA CUDA](/FastVideo/getting_started/installation/gpu)
- **NVIDIA DGX Spark / GB10 (ARM64 + CUDA 13)** — [install](/FastVideo/getting_started/installation/spark),
  [performance](/FastVideo/getting_started/installation/spark_performance),
  [pair two Sparks](/FastVideo/getting_started/installation/spark_pair)
- [Apple silicon (MLX)](/FastVideo/getting_started/installation/mlx)
- [AMD / ROCm support and current limits](/FastVideo/getting_started/installation/amd)

## Quick Installation {#quick-installation}

### Using uv (recommended) {#using-uv-recommended}

Use uv as the default environment manager for faster and more stable installs.
The commands below target NVIDIA CUDA 12; use `UV_TORCH_BACKEND=cu130` on
CUDA 13. Apple silicon users should follow the [MLX install guide](/FastVideo/getting_started/installation/mlx).

```bash
# Create and activate a new uv environment
uv venv --python 3.12 --seed
source .venv/bin/activate

UV_TORCH_BACKEND=cu126 uv pip install fastvideo
```

### Using Conda (alternative) {#using-conda-alternative}

```bash
# Create and activate a new conda environment
conda create -n fastvideo python=3.12 -y
conda activate fastvideo

UV_TORCH_BACKEND=cu126 uv pip install fastvideo
```

### From source {#from-source}

FastVideo pins PyTorch 2.12.0. Select its CUDA build explicitly with
`UV_TORCH_BACKEND=cu126` for CUDA 12 or `UV_TORCH_BACKEND=cu130` for CUDA 13.

```bash
git clone https://github.com/hao-ai-lab/FastVideo.git
cd FastVideo
UV_TORCH_BACKEND=cu126 uv pip install -e .

# optional: install flash-attn
uv pip install flash-attn --no-build-isolation -v
```

Alternative with Conda environment (still drives installs through `uv`):

```bash
UV_TORCH_BACKEND=cu126 uv pip install -e .
uv pip install flash-attn --no-build-isolation -v
```

## Requirements {#requirements}

- **Python**: 3.10-3.12 is the tested and recommended range (the commands
  above pin 3.12)
- **NVIDIA GPUs**: CUDA 12.6+ with compute capability 7.0+
- **Apple Silicon**: macOS 14.0+ with M1/M2/M3/M4 chips
- **CPU**: x86_64 architecture (for CPU-only inference)

## Next Steps {#next-steps}

- [Quick Start](/FastVideo/getting_started/quick_start) - Generate your first video
- [Inference Cookbook](/FastVideo/cookbook) - Choose a maintained recipe
- [Configuration](/FastVideo/inference/configuration) - Learn about configuration options
- [Examples](/FastVideo/inference/examples/examples_inference_index) - Explore scripts and notebooks