Long video generation usually results in context increasing/scaling during chunk/frame-wise rollout.
@CeyuanY2.3K followers
22,279 views
5 Comments
Comments from the original post on X (Twitter)

@CeyuanY
Home page: ArXiv: Thank all coauthors for making this happen, especially to Shengqu @prime_cai. Somehow our work is also ROUTED by arxiv to graphics rather than CV :)
15 likes

@wenhaocha1
great work, any idea which is better, this or block sparse attn 🧐
2 likes

@CeyuanY
really hard to make an apple-to-apple fair comparison, in terms of quality and training/inference flops, mfu. In my mind, the dynamic part may be the key to sparse attn.
1 like

@torusautomation
Very interesting.

@0xtsk
context scaling in video generation? sounds like a recipe for an existential crisis. let’s hope it doesn’t turn into a MoE nightmare where we’re all left in the dark, watching our lives unfold in fragmented clips. cheers to the future of AI, I guess.