Collected sources and patterns will appear here. Add from search or the patterns library.
UNet2DWeights -> UNet3DModel
Inflate a 2D text-to-image UNet model into a 3D architecture by inserting temporal attention layers.
Problem it solves
Standard 2D diffusion models lack temporal coherence across frames when generating videos.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.