Stability AI announces a pivotal advancement in the realm of generative video models with the launch of Stable Video Diffusion. Representing a significant milestone akin to the pioneering Stable Diffusion for images, this state-of-the-art AI video model is now available for research preview.
The model’s code is openly accessible on GitHub, while the necessary weights to operate the model locally are hosted on the Hugging Face platform. The technical intricacies and capabilities of this model are detailed comprehensively in the accompanying research paper.
Must Read: AI Safety Summit 2023: The 6 highlights you must know
Adaptability stands as a hallmark feature of this innovation. Stable Video Diffusion demonstrates versatility across multiple downstream tasks, including the synthesis of multiple views from a single image, offering opportunities for refinement through fine-tuning on multi-view datasets.
The vision is to expand this foundational model into a family of models, akin to the ecosystem that has evolved around stable diffusion for images.
The release unveils two variants of the image-to-video model, capable of generating sequences of 14 and 25 frames, customizable to frame rates ranging from 3 to 30 frames per second.
In external evaluations, these models have showcased superiority over existing closed models, as evidenced by user preference studies conducted at the time of release.
Despite its remarkable capabilities, it’s important to note that at this stage, Stable Video Diffusion is exclusively intended for research purposes and not for real-world or commercial applications. Stability AI is keen on refining and improving the model based on user insights, emphasizing the significance of feedback regarding safety and quality to enhance its eventual release.
This latest addition, Stable Video Diffusion, joins Stability AI’s expansive suite of open-source models spanning various modalities such as image, language, audio, 3D, and code. This diverse portfolio underscores Stability AI’s commitment to augmenting human intelligence through AI innovation.
Must Read: OpenAI ChatGPT Voice rolled out for Free: check out how to use it on Mobile.
Stable Video Diffusion marks a paradigm shift in the domain of generative video AI, offering a glimpse into its potential applications across industries, from advertising and education to entertainment and beyond. As Stability AI continues to refine and evolve this model, the anticipation grows for its eventual wider release, promising far-reaching implications for AI innovation and application.
Must Read: Bard’s YouTube Integration: AI ChatBot Can now watch Videos for you
