Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
Tensor<VideoFrames> -> Tensor<ContinuousLatents>
Encode raw video frames into a continuous latent vector space.
Problem it solves
High-dimensional video pixel data is too computationally intensive for transformer or diffusion backbones to process directly.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.