Text to Video: The Next Leap in AI Generation

[@a16z]

Topics Covered:
00:00 - Text to Video: The Next Leap in AI Generation
02:41 - The Stable Diffusion backstory
04:35 - Diffusion vs autoregressive models
07:17 - The benefits of single step sampling
10:55 - Why generative video?
13:10 - Understanding physics through AI video
14:53 - The challenge of creating generative video
18:43 - Data set selection and training
21:24 - Structural consistency and 3D objects
23:51 - Incorporating LoRAs
28:47 - How should creators think about these tools?
31:41 - Open challenges in video generation
32:35 - Infrastructure challenges and future research

[@TheGameIsOverCy]

Nice one, thanks ! Can't wait to see what's coming next !

[@tyc00n]

text to video is a sideshow, simulation making datasets for robotics is the main event