top of page
Search

3/15

  • Writer: issac zhang
    issac zhang
  • Mar 15, 2024
  • 1 min read

试了几个有着不同的节点的workflow,发现都不能达到输入video reference 然后输出正确版画风格的要求。输出的video都存在着人体错误,动画不正常的等多种缺陷(详见video1,3),但是测试了其他的几个大模型和lora结合之后,发现效果不错(见video2),video2能良好结合动画和动作的参考并输出结果(prompt对齐影响如何暂时没有做测试),所以我推断输入的video reference不能很好的转成版画风格可能并不是prompt、controlnet、parameter、ipdapter的原因,更可能是单纯的lora和大模型的问题。

ree
"The image on the left is for action reference, while the image on the right is for style reference."

 
 
 

Recent Posts

See All
20250702 Supervision Meeting

Here is the Speech draft. P1. As you can see, here is the refined custom workflow for AI-integrated digital woodblock print animation,...

 
 
 

Comments


© Powered by Zicheng Zhang

bottom of page