← Back to BlogGetting Started

Getting Started with Fastwan AI

Learn how to use Fastwan AI for rapid video generation using breakthrough sparse distillation technology.

What You'll Learn

  • • Understanding Fastwan AI model variants and capabilities
  • • Hardware requirements and setup instructions
  • • Step-by-step guide to generating your first video
  • • Performance optimization tips and best practices

Understanding Fastwan AI

Fastwan AI represents a breakthrough in video generation technology, built on advanced sparse distillation techniques that enable unprecedented speed without compromising quality. The system combines video sparse attention (VSA) with distribution matching distillation (DMD) to compress traditional 50-step generation processes into just 3 steps.

Model Selection

FastWan2.1-1.3B

Resolution: 480P
H200: 5 seconds
RTX 4090: 21 seconds

Perfect for rapid prototyping and real-time applications where speed is prioritized.

FastWan2.2-5B

Resolution: 720P
H200: 16 seconds
RTX 4090: ~45 seconds

Higher resolution output suitable for professional content creation.

Hardware Requirements

Recommended: NVIDIA H200

Optimal performance with fastest generation times for production environments.

Consumer: RTX 4090

Excellent performance for personal projects and small-scale commercial work.

Installation

# Install FastVideo framework
git clone https://github.com/hao-ai-lab/FastVideo.git
cd FastVideo
pip install -r requirements.txt

# Download model weights
python download_models.py --model fastwan2.1-1.3b

Generate Your First Video

from fastvideo import FastWanPipeline

# Initialize pipeline
pipe = FastWanPipeline.from_pretrained(
  "FastVideo/FastWan2.1-T2V-1.3B-Diffusers"
)

# Generate video
video = pipe("A bird flying over mountains").videos[0]

Ready to Create?

With Fastwan AI properly configured, you can now generate professional-quality videos in seconds.

Try Live Demo →