Style processing and upscaling video created in iClone

[jsmorgan9573]

Hello friends, I’m @jsmorgan9573 I have been making short stop frame animation clips and skits with friends and family since the 1960s (it was 8mm film back then)
These days I use iClone to create avatars of my grandkids and have produced some pretty fun MP4 videos.
I've been playing with ComfyUI for a few weeks and it's absolutely amazing. I love it, but I have a big problem.
I'm 70 this year, I was a process controls engineer for more than 40 years until I contracted Cancer in 2009. I had the surgery, 8 months of chemotherapy and came through just fine. In 2019 the cancer returned and it was a little more of a struggle than the previous ordeal in that the chemo took more of a toll on my cognitive skills and short term memory (they call it chemo brain)
The reason I'm revealing all this is to illustrate the reasons I'm having so much trouble learning how to code a workflow of what I'm trying to do.
I'm reaching out to all the brilliant coders here hoping someone could help me out. I've managed to get a few workflows to function after hours of hunting down the required models and nodes called for, but most of them fail either because I can't find the required modules or when I do find them I don't know which comfyui folder to put them in.
It's becoming very frustrating.
My goal is to create a video to video workflow that will take my iClone productions (mp4's) and process them through a styler and upscaler to make them appear less cartoonish and more like real world scenes.
I'm not even sure it's possible to do. I even rotoscoped an mp4 into hundreds of single frame images hoping to process each image separately and then stich them back into a video (without much luck)
I'm working with an Asustek ROG strix PC, Intel(R) Core(TM) i7-14700F processor, 32GB ram, Nvidia GeForce RTX 4060i with 8GB of vram. I would truly appreciate any help or insight offered.

Thanks for listening,
JS Morgan
(mailto:jsmorgan9573@gmail.com)

[ReinerBforartists]

Hey,

In case you tell us which methods fails, somebody here might be able to help you out with this specific problem.

In general, there is not the one solution as i just discovered in the last days. You can do upscaling by a simple lanczos in a upscale image node. That's like upscaling an image in Photoshop. You can upscale by a upscaling model. This gives you a bit more detail. You can upscale by a second KSampler. Which basically means you dream over the whole video, and add new details. Then there is special upscaling nodes like Supir, which promises upscaling quality like in Topaz Video enhancer or Davinci Resolve Super Upscaling. CSSR is another player here, but super slow. I stopped my experiments after two hours then when it told me that it is at frame 30 of 180 ...

So the first thing you need to decide is what is more important to you. Quality or speed. And if you really want to do it in Comfy. As told above, there is Topaz Video Enhancer. Or DaVinci Resolve Super Upscaling in the commercial version.

I personally favourite two upscaling workflows at the moment. One is a mixture of simple upscaling and upscaling by a model. This one is pretty fast. But simple. And a workflow from banodoco, which does upscaling across IP Adapter, Controlnet, AnimateDiff LCM and Ultimate SD Upscale. Called supermegabadassupscaler. I am not even sure if this is available outside of his discord.

The first method can be found in my CogVideoX workflow at Civitai. https://civitai.com/models/1117606
For the workflow from banodoco you have to join his Discord i guess. I don't want to paste workflows from other people into discussions. It is not made by me.
Kind regards
Arunderan