DreamBox Wan Video Generator

A Docker setup for running the Wan Video Generator using DiffSynth-Studio's WanVideoPipeline, optimized for NVIDIA RTX 3090 GPUs.

Quick Start (Step-by-Step Guide)

Prerequisites

Before you begin, make sure you have installed:

• Docker Desktop (for Windows/Mac) or Docker Engine (for Linux)
• NVIDIA Container Toolkit (for GPU support)
• An NVIDIA GPU with at least 24GB VRAM (RTX 3090 or equivalent)

Installation Steps

1. Start Docker

   • Launch Docker Desktop if you're on Windows/Mac
   • On Linux, ensure the Docker service is running:

     sudo systemctl start docker

2. Open a Terminal/Command Prompt

   • Windows: Press Win+R, type cmd and press Enter
   • Mac: Open Terminal from Applications/Utilities
   • Linux: Open your terminal application

3. Navigate to Where You Want to Install

   • Use the cd command to go to your preferred directory:

     # Example: installing in your Documents folder
     cd Documents

4. Clone the Repository

   git clone https://github.com/YourUsername/DreamBox_Wan
   cd DreamBox_Wan

5. Build the Docker Image

   • This step downloads the models automatically (may take 5-15 minutes depending on your internet speed)

   docker compose build

6. Run the Container

   docker compose up -d

7. Access the UI

   • Open your web browser
   • Navigate to: http://localhost:7860
   • You should see the Wan Video Generator interface

8. When Finished

   • To stop the container:

   docker compose down

Requirements

• Docker
• Docker Compose
• NVIDIA Container Toolkit
• NVIDIA GPU with at least 24GB VRAM (RTX 3090 or equivalent)

Troubleshooting

Common Issues

1. Docker isn't running

   • Make sure Docker Desktop is running (check for the icon in your system tray)
   • On Linux: sudo systemctl status docker to check if the service is active

2. GPU not detected

   • Ensure NVIDIA drivers are up to date
   • Verify NVIDIA Container Toolkit is properly installed:

     nvidia-smi

   • This command should display your GPU information

3. Permission issues with Docker

   • Windows: Run Command Prompt as Administrator
   • Linux: You may need to add your user to the docker group:

     sudo usermod -aG docker $USER
     # Then log out and back in

4. Port already in use

   • If port 7860 is already being used by another application, modify the port in docker-compose.yml

5. Not enough VRAM

   • If you have less than 24GB VRAM, try using the tiling option in the UI
   • Reduce the number of frames and inference steps

Checking Logs

If you encounter issues, check the container logs:

docker compose logs

Usage Guide

1. Enter a prompt describing the video you want to generate
2. Upload a conditioning image
3. Adjust the parameters as needed:
   • Number of frames
   • FPS (frames per second)
   • Guidance scale
   • Inference steps
   • Seed value (-1 for random)
   • Enable/disable tiling (helps with memory usage)
4. Click "Generate Video"
5. The generated video will be saved to the output directory

File Storage

The Docker setup includes one volume mount:

• ./output:/app/output - For storing generated videos

All generated videos will be saved in the output directory in your project folder.

RTX 3090 Optimization

This setup includes specific optimizations for RTX 3090 GPUs with 24GB VRAM:

• Flash Attention 2 for efficient attention computation
• BFloat16 precision for optimal performance/memory balance
• VRAM management features to maximize available memory
• Xformers memory efficient attention when available

Technical Details

• Uses DiffSynth-Studio's WanVideoPipeline for video generation
• Employs Flash Attention 2 and torch.bfloat16 precision for RTX 3090 GPUs
• Optimizes VRAM usage with vram_management feature
• Generates videos based on text prompts and conditioning images