How to install Stable Video Difussion?

For those into AI drawing, Stable Diffusion is a familiar term.

Since Stability AI released Stable Diffusion (SD), it has gained a lot of popularity. Expanding its expertise in the image domain, Stability AI recently introduced Stable Video Diffusion. Now, you can create a video with just one image!

Installation Guide for Stable Video Diffusion

Stability AI has open-sourced two models:

• SVD Model: Generates videos at a frame rate of 14 with a resolution of 576x1024.
• SVD-XT Model: An upgraded version with finetuning, maintaining the resolution but capable of a frame rate of 25.

Reminder: To run Stable Video Diffusion on your local computer, you need a GPU with at least 20GB of memory.

Step 1: Download the Codebase

Click this link to download the codebase.

Step 2: Set Up Python Environment

Python version needs to be 3.10 or higher due to third-party dependencies. We recommend creating a new environment using conda.

Activate the environment and install the necessary dependencies.

Finally, install sgm by running it in the project's root directory. If you want to train the model, you'll also need to install sdata, but I won't cover that here due to my computer's configuration limitations.

Step 3: Model Download Links

• SVD Model Download Link: SVD Model
• SVD-XT Model Download Link: SVD-XT Model

After downloading the models, place them in the checkpoints directory.

Step 4: Run the Algorithm

You can use Streamlit to run the algorithm.