well after about a day of the original workflow being up and me clearing out abunch of files i deleted my original workflow so i went to download it back from here and what do you know i uploaded the wrong one to civitai due to my own fault of misnaming of my files. so i took that one down and now reposting the proper one after fixing this. i apologize for this oversite.
this is the proper 4step 3 video looper. with the interpolation fixed.
I wanted to make an easy workflow dedicated to video looping. This is a 4step workflow using high and low models as well as high and low noise loras. we have toggle buttons for resolution, color match, interpolation, as well as upscaling. each video created is set at 81 frames at 24fps.
upload your image, write your prompts, select which features you'd want to use as well as resolution and hit run.
I recommend using the color match feature as sometimes it loses color between video A-B-C.
if people like this and would want me to make a non 4step or a GGUF style let me know and ill look into making those.
I use DaSiWa Boundbite v10 lightspeed safetensors in this, I haven't tested other ones but its what I use and recommend. Use which ever upscale model you prefer I like the 2x anime one for my generations.
Description
changed where the image reference for the color match subgraph is inserted so it takes the color from the resized image rather than the original input image. this should make the color matching feature more consistent.
changed what frames are trimmed, added another interpolation section before the interpolation group so that it adds frames to make the video loop more consistently before turning on interpolation feature. not trimming the frames allowing the rife node to insert a frame between the end and start of where video C loops to A for a smoother more seamless loop. its a small change some might not notice a difference but i do so i wanted to push an update on this through.