One of the main investments in the technology market consists of creating more generative AI models. And now, Stability AI has announced the development of a new tool of this type.
Stable Video Diffusion’s main task is to generate videos through requests made in text. In practice, it is capable of creating recordings of 14 or 25 frames, with a reproduction rate of 3 to 30 fps.
Another possibility is the development of images with a wide variety of angles, within the same scene. The results can be either with a more realistic appearance or in an animation style. The company even released an article about the creation. See an excerpt from it:
“Introducing Stable Video Diffusion – a latent video diffusion model for state-of-the-art, high-resolution text-to-video and image-to-video generation. Recently, latent diffusion models trained for 2D image synthesis have been transformed into generative video models by inserting temporal layers and fine-tuning them on small, high-quality video datasets. However, training methods in the literature vary widely and the field has not yet agreed on a unified strategy for video data curation. In this paper, we identify and evaluate three different stages for successfully training video LDMs: text-to-image pre-training, video pre-training, and high-quality video fine-tuning.”
Stability AI
Generative video AI is still in the testing phase, with a focus on privacy and security improvements, before commercial availability. Even so, Stability AI even published a waiting list for anyone who wants to test its new tool.