Description
Stable Video Diffusion, created by Stability AI, is an open-source AI model for video generation. It is the first foundation model from the company for creating videos and is built upon the Stable Diffusion image model.
Currently, the model is in a research preview phase. The code for Stable Video Diffusion can be found in Stability AI's GitHub repository, while the weights needed to run it locally are available on their Hugging Face page.
The model can be used in different video-related applications and can be easily adapted for multi-view synthesis from one image using multi-view datasets. External evaluations and user studies show that Stable Video Diffusion performs well compared to other closed models. It is important to know that Stable Video Diffusion is currently only for research purposes and not for commercial or real-world use.
Feedback on safety and quality helps improve the model before its official release. Stable Video Diffusion is one of Stability AI's open-source models, covering areas like image, language, audio, 3D, and code.
The company aims to improve human intelligence and regularly updates its models. To keep up with Stability AI's progress, users can subscribe to the newsletter and contact the company to learn about commercial uses.
Stability AI also shares updates on social media platforms like Twitter and Instagram.
Pricing Plans
Releases
Improved Motion Coherence: To enhance motion coherence throughout the generated video, the model uses a motion bucket ID set at 127, leading to smoother transitions and a more realistic depiction of motion. 
Enhanced Resolution: SVD 1.1 produces videos with a resolution of 1024x576 pixels, resulting in clearer and more detailed visuals. 
Open-Source Availability: The model is available on platforms like Hugging Face, providing users the opportunity to access, experiment, and help improve its development.
