In the workflow you try to link to a 13b gguf TTV which actually links to a 14b gguf. Also you say you are using a 13b gguf but the workflow is using a regular diffusion model loader.

兄弟，这个有办法扩张到可以接通面部识别，保持面部一致性的转换视频吗，如果用lora控制的话太困了，训练lora需要很久。所以我们需要个面部识别和图像输入以及怎么实现人脸一致性

Can you do vid2img2vid?

why my result on this workflow are like a cartoon?

I got this error: "The new shape must be larger than the original tensor in all dimensions", could you help me, please?

Error SamplerCustom

Given groups=1, weight of size [1536, 32, 1, 2, 2], expected input[2, 16, 15, 60, 48] to have 32 channels, but got 16 channels instead

dident change anything and loaded all your models

Test in different number of images, image quality gap is particularly large.

this is cool

this workflow has become my new toy :) it's very successful and works very fast. i use it to create different variations of the videos i've produced before. i couldn't get the florence part to work, i got some errors but i don't think it's very necessary.

I also got this error: "The new shape must be larger than the original tensor in all dimensions", could you help me, please?

I already downloaded the clip but it doesn't work

What is the clip file name?

I just get a flickering solid beige video as result...

is it becuase of this |[VideoHelperSuite] - WARNING - Output images were not of valid resolution and have had padding applied"?