Stability AI, a notable name in the AI startup scene, has recently unveiled its latest innovation, Stable Video Diffusion.
This ground-breaking AI model is designed to animate static images into dynamic videos, a feat achieved by only a handful of models, either open-source or commercial.
Stable Video Diffusion is an extension of Stability's existing Stable Diffusion text-to-image model. However, it's currently in a 'research preview' phase, with users required to agree to specific terms of use. These terms outline the model's intended applications, such as educational or creative tools, and prohibit its use for creating 'factual or true representations of people or events.'
Stable Video Diffusion comprises two models, SVD and SVD-XT. SVD transforms images into 14-frame videos, while SVD-XT increases the frames to 24. Both models can generate videos at a speed ranging from three to 30 frames per second.
The models were trained on a dataset of millions of videos and then fine-tuned on a smaller set of clips. The source of these videos remains unclear, potentially opening up legal and ethical challenges around usage rights.
Despite its impressive capabilities, Stable Video Diffusion has its limitations. It struggles with generating videos without motion or slow camera pans, rendering text, or consistently generating faces and people. However, Stability is optimistic about the models' extensibility and adaptability for various use cases, such as generating 360-degree views of objects.
Looking ahead, Stability plans to develop a variety of models that extend SVD and SVD-XT's capabilities. The company also aims to launch a text-to-video tool, bringing text prompting to the models on the web. The ultimate goal appears to be commercialisation, with potential applications in advertising, education, entertainment, and more.
Despite recent financial challenges and executive departures, Stability AI continues to innovate and push boundaries in the AI sphere. With over $125 million raised and a last valuation at $1 billion, the company is certainly one to watch as it navigates the complex world of AI development and commercialisation.
Made with TRUST_AI - see the Charter: https://www.modelprop.co.uk/trust-ai
Comments