How to Access Stable Video Diffusion

Introduction

As of July 24, 2025, our Stable Video Diffusion API has been officially deprecated. This means the hosted API endpoints are no longer available. This article explains how to set up self-hosting, allowing you to continue using our Stable Video Diffusion model. 

For the complete details about this endpoint becoming deprecated, please refer to our Release Notes page: Release Notes

How to Access SVD Through Self-Hosting

To self-host Stable Video Diffusion, follow the steps below:

1. Review Licensing and Access Requirements
   Visit the Core Models page and review the license terms for accessing Stable Video Diffusion.

2. Download the Model Weights
   Select the appropriate engine version (e.g., SVD 1.1) and download the model weights from our Hugging Face repository.

3. Get the Reference Code
   Access the official reference implementation on GitHub under the Generative-Models repository.
   This codebase provides the necessary scripts and utilities for running SVD locally.

4. Review the README
   Read through the repository’s README for setup instructions, environment requirements, and usage notes.

5. Locate the Sampling Script
   The script used for Stable Video Diffusion is located at:
   generative-models/scripts/sampling/simple_video_sample.py

6. Run SVD Locally
   Using the script and model weights, you can run the engine on your local machine to generate short videos from input images.

If you require additional support in setting up this self hosted model please create a Support Ticket and a member of our team will assist.