If you pick the SVD model, the clip will play for 14 frames, while the SVD-XT model takes it to 25 frames. As mentioned earlier, the video will play for 4 seconds post-render, so the length of the generated video will not matter too much for now.
Stability AI has posted a video on Youtube, showcasing the capabilities of Stable Video Diffusion, and the quality is surprisingly great. As far as the extent of motion is concerned, you are can only create a slow panning shot or a slight bobbing of the head of the subject. AI Artists have been craving an animation model on Stable Diffusion, and now that it is here, it will be interesting to see wha artists come up with this new feature.
Now, if you are intrigued by this tool, you will have to get onto a waitlist, and once you receive the invite, you can start using this tool. At the moment, you can access and read Stability AI’s white paper on this tool to see how the project came to be and what it encapsulates. The reason for the delay in deployment is that Stability AI is trying to use as many publicly accessibly video content to train their models. This is because Getty Images had sued Stability AI for using their proprietary images for training their AI Art generation tool. Stability AI want to be a bit more careful in this case.
