I preprocessed some video data and decoded them. I noticed that their quality is poor. At the same time, when I use the image-to-video pipeline from the demo site, the quality doesn't drop. What is the reason for that?
Besides, whichever LTX version I choose, the results obtained from the diffusers pipeline are much worse in RGB quality compared to the official demo.