Stability AI's Stable Video 4D and multi-angle video generation with AI`

Stability AI unveiled Stable Video 4D, a new AI model that can turn single-object videos into multiple videos from eight different angles. This has potential applications in game development, video editing, and virtual reality.

https://stability.ai/news/stable-video-4d

Here are the reasons why this is significant:

1. Novel multi-angle video generation: Stable Video 4D can transform a single input video into multiple novel-view videos from 8 different angles/views. This is a major step forward from previous image-to-video or single-view video generation models.

2. Efficient 3D representation: The model enables efficient optimisation of dynamic 3D representations of subjects in videos, which has important implications for fields like game development, video editing, and virtual reality

3. Improved consistency: Unlike previous approaches that required combining multiple diffusion models, Stable Video 4D generates multiple novel-view videos simultaneously. This results in better consistency across spatial and temporal axes.

4. Speed and performance: The model can generate 5 frames across 8 views in about 40 seconds, with full 4D optimisation taking 20-25 minutes. This represents state-of-the-art performance for such complex video generation tasks.

5. Flexibility and customisation: Users can specify desired camera angles, allowing for tailored outputs that meet specific creative or technical needs.

6. Open-source availability: The model is available on Hugging Face, allowing researchers and developers to build upon and improve the technology.

Potential industry impact:

The technology has significant potential applications in gaming, film, virtual reality, and other visual media industries, potentially transforming how 3D content is created and manipulated.

Stability AI is actively refining the model to handle a wider range of real-world videos, indicating potential for even broader applications in the future.