VideoCrafter1: Open Diffusion Models for High-Quality Video Generation
image-20231108102745812
image-20231108104712911
包括2个关键部分:video vae 和 video diffusion
image-20231108111414839
image-20231108112046340
image-20231108112244180
image-20231108112640660
image-20231108110036296
需要利用来自CLIP image ViT (clip image encoder)的最后一层的全部面片patch的token Fvis = {fi}K i=0
Fimg = P(Fvis)
image-20231108114344516
image-20231108114424527
image-20231108114431546
image-20231108114758612
image-20231108114824763
image-20231108114657469
image-20231108115035248
image-20231108115147302
image-20231108115209613
image-20231108115243336
本文分享自 iResearch666 微信公众号,前往查看
如有侵权,请联系 cloudcommunity@tencent.com 删除。
本文参与 腾讯云自媒体同步曝光计划 ,欢迎热爱写作的你一起参与!