Recently, Stability AI announced the launch of a revolutionary video processing technology—Stable Video4D. This technology can transform a single-perspective video into new videos from eight different angles, providing creators with unprecedented flexibility and creativity.
Stable Video4D is built on the foundation of the company's previous Stable Video Diffusion model. Unlike converting images into videos, the new model can accept video inputs and generate multiple new-perspective video outputs, achieving a significant leap from image-based video generation to full 3D dynamic video synthesis.
When used, users only need to upload a video and specify the desired 3D camera positions, and Stable Video4D can generate eight new-perspective videos, providing a full range of multi-angle views. Currently, the model can generate eight perspectives of a 5-frame video in about 40 seconds, with the entire 4D optimization process taking approximately 20-25 minutes.
Compared to previous methods, Stable Video4D can simultaneously generate multiple new-perspective videos, significantly improving consistency in both spatial and temporal axes. This not only ensures consistency of objects across multiple perspectives and timestamps but also achieves a more lightweight 4D optimization framework.
Stability AI stated that Stable Video4D is currently in the research phase and is expected to be widely applied in areas such as game development, video editing, and virtual reality in the future. The company is actively optimizing the model to handle a broader range of real-world videos.
Stable Video4D is now available on the Hugging Face platform. Stability AI looks forward to further enhancing the potential of this technology to create realistic multi-angle videos through continuous research and development. The company will continue to collaborate with researchers, experts, and the community to drive technological innovation and continuously improve model performance.
Model URL: https://huggingface.co/stabilityai/sv4d