Here is info about Framepack by Lvmin Zhang. This model uses one single image and/or test prompts to create a frame-by-frame video sequence. Its key advantage is that it avoids drifting (slow degeneration as frames progress).
Here is a publication by the authors of this model, describing the principles:
Lvmin Zhang and Maneesh Agrawala (2025), Packing Input Frame Contexts in Next-Frame Prediction Models for Video Generation. Arxiv.
Another publication: FramePack: O(1) Video Diffusion on Consumer GPUs
Framepack model applications can be found online here:
No comments:
Post a Comment