Open JasonS09 opened 1 month ago
It's meant for video2video, I'm not force resizing that input so you probably want to use this node to resize to closest compatible dimensions:
Also if you didn't try inputting only the video without start or end images. I don't recognize that error though, so I'm unsure of the cause.
I still get the error, here's my workflow
Also, are the image inputs (image_start, image_end, full_input_video) meant to be exclusive with each other? Or can I use all the inputs simultaneously?
Using them with full video would only replace the first or last frame, so not very useful. Have you tried different frame counts? It might need even number or something.
I just tried with 24, I still get the same error but the numbers change:
RuntimeError: The expanded size of the tensor (220) must match the existing size (348) at non-singleton dimension 3. Target sizes: [2, 16, 19296, 220]. Tensor sizes: [2, 16, 1, 348]
Yeah now I'm getting that too, maybe Comfy changed something with the text encoder since this definitely worked before.
Nevermind sorry, it's because the image encoder can only take single image, not the whole video.
Hello, tried to run vid2vid again using the example, this time it didn't work and got the same error. Can you reproduce it?
I noticed the EasyAnimate Sampler includes a full_input_video input. However, no matter what I try, I get this error if I pass a video as input.
RuntimeError: The size of tensor a (7) must match the size of tensor b (6) at non-singleton dimension 2