Learn to install and use Alibaba's open-source Wan 2.1 AI video generation model in this comprehensive 45-minute tutorial. Master text-to-video, video-to-video, and image-to-video generation with a custom Gradio app that works on systems with as little as 3.5GB VRAM. Follow step-by-step instructions for Windows installation without requiring Linux or WSL, and explore cloud-based alternatives using Massed Compute and RunPod for more powerful GPU access. Compare performance between RTX 3090 TI and RTX 5090 across various model configurations, including the 14 billion parameter model. Discover essential tips for prompt engineering, aspect ratio optimization, and efficient VRAM management. Get detailed guidance on cloud service setup, file management, and interface navigation, complete with demonstrations of generation speeds and output quality across different GPU configurations. Access additional resources through provided GitHub repositories, Discord community, and detailed documentation for troubleshooting and advanced implementations.

Syllabus

0:00 Introduction to Wan AI Video Generation Model
0:28 Wan AI vs. Paid Video Services
1:12 Model Features: Text-to-Video, Image-to-Video, Video-to-Video
1:40 Model Parameters and VRAM Requirements
2:12 Demonstration of Generated Videos: Image to Video
2:40 Demonstration of Generated Videos: Text to Video
3:01 Demonstration of Generated Videos: More Examples
3:48 Importance of Prompts and Negative Prompts
4:06 Windows Installation Tutorial Overview
4:19 Performance Comparison: RTX 5090 vs. RTX 3090 Ti
4:41 Application Compatibility and Virtual Environment
4:59 Cloud Service Options: Massed Compute and Rampod
5:33 Detailed Post and Installation Instructions
6:01 Downloading and Extracting Installer Files
6:45 Running the Windows Installer Important Note: Don't Run as Admin
8:07 Model File Download Options and Instructions
10:12 Troubleshooting Model Download Failures
11:05 Moving Downloaded Models and Starting the Application
11:50 Gradio Interface Overview and Model Selection
12:02 GPU Presets and VRAM Management
12:23 Prompt Enhancement Feature
13:17 Text-to-Video Generation Demo and Speed on RTX 5090
14:13 Performance Comparison on RTX 3090 Ti
15:58 Actual Speed Test No Video Recording
17:04 Output Folder and Multi-line Prompt Support
17:41 Generating Multiple Videos and Random Seeds
18:10 Using Video-to-Video Feature
18:45 Video-to-Video Demonstration and Denoising
19:38 Exploring 14 Billion Parameter Text-to-Video Model
20:37 Performance of 14 Billion Parameter Model on RTX 5090 and 3090 Ti
21:30 Testing 14 Billion Parameter 480 Pixel Image-to-Video Model
22:15 Prompting with Claude AI for Image-to-Video
23:26 Image-to-Video Generation Demo and Speed on RTX 5090 and 3090 Ti
23:45 Crucial Tips: Aspect Ratio and Cropping for Image-to-Video
24:16 Repeating Test for 720 Pixel Image-to-Video Model
24:32 Step Speeds for 720 Pixel Image-to-Video Model on RTX 5090 and 3090 Ti
24:57 Using Wan AI on Massed Compute Cloud Service
25:31 Registering and Setting Up Massed Compute
26:37 Deploying a Machine and Connecting with Thinlinc Client
28:09 File Transfer to Massed Compute Shared Folder
29:32 Installing Wan AI on Massed Compute
30:13 Downloading Models on Massed Compute Fast Download Speed
31:00 Starting the Gradio Interface on Massed Compute
31:20 Accessing Gradio Interface from Local Computer
31:56 Generating Video on Massed Compute and Monitoring Progress
32:46 Downloading Generated Videos from Massed Compute
33:30 Terminating Massed Compute Machine Caution
34:04 Faster Generation Options on Massed Compute H100, A100, L40S
34:31 Running Multiple Applications in Parallel on Massed Compute
35:08 Using Wan AI on RunPod Cloud Service
35:41 Registering and Setting Up RunPod
35:58 Deploying a Pod and Connecting to Jupiter Lab
37:50 Uploading and Extracting Installer on RunPod
38:10 Installing Wan AI on RunPod
38:40 Downloading Models on RunPod Fast Download Speed
40:00 Starting Gradio Interface on RunPod and Accessing Live Share Link
40:37 Testing Generation Speed on RunPod H100 GPU
43:25 Downloading Generated Videos from RunPod
43:55 Stopping and Terminating Rampod Pod Caution and Resume Option