AI/ML
How to Deploy WAN 2.1 on GCP for AI Video Workflows
Introduction
For effective workflows, Alibaba's open-source Wan 2.1 video model thrives on Google Cloud Platform (GCP). This guide describes how technical people can utilize Google's AI infrastructure to install it on GCP using Compute Engine as of March 13, 2025.
Prerequisites
- GCP Account: With Compute Engine enabled.
- Hardware: A2-highgpu-1g (A100, 40GB VRAM).
Deployment Steps
Step 1: Create VM
- Machine Type: A2-highgpu-1g.
- Boot Disk: Ubuntu 20.04, 50GB.
Step 2: Connect
SSH:
gcloud compute ssh wan2-1-vm --zone us-central1-a
Step 3: Install and Download
Setup:
sudo apt update && sudo apt install -y python3-pip gitgit clone https://github.com/Wan-Video/Wan2.1.gitcd Wan2.1huggingface-cli download Wan-AI/Wan2.1-T2V-14B --local-dir ./modelspip3 install -r requirements.txt
Step 4: Run Wan2.1
Execute:
python3 inference.py --model_path ./models/Wan2.1-T2V-14B --prompt "A stormy ocean" --output ocean.mp4
Requirements
- Instance: A2-highgpu-1g
- VRAM: 40GB
- Cost/Hour: $3.674
Conclusion
Wan 2.1 on GCP’s A2 instances delivers high performance video generation as of March 2025. Ideal for developers, this setup harnesses Google’s GPU power, ensuring efficient AI video workflows.
Ready to transform your business with our technology solutions? Contact Us today to Leverage Our AI/ML Expertise.
AI/ML