This is a flux workflow for XY plotting loras against redux, mainly for testing how different redux strengths affect outputs with the option to see how loras can alter the redux conditioning.
I'm sorry how messy it is I haven't had any luck with efficiency nodes and flux, so I had to be unorthodox with joining everything together. I've tried to make it simple to use for people who don't care what's going on and just want it to work, hopefully I succeeded (I've included instructions in the workflow).
This workflow works on my 8gb VRAM gpu, regardless of the size of the XY plot as it does each image 1 at a time. I was able to generate a 9x11 plot with no issues, just took a couple hours lol.
Description
FAQ
Comments (4)
which exact CLIP model to use this please? Mine is crashing for some reason.. Just want to be sure I am using the exact right one..
Thanks! cool workflow
I'm assuming you're referring to the CLIP vision model, the link in the readme seems to no longer work as the publisher changed their name. Here is a new link to the model I use: https://huggingface.co/Comfy-Org/sigclip_vision_384/blob/main/sigclip_vision_patch14_384.safetensors
this helped me so much not only how to understand x-y plots in comfy, but because i was forced to research and learn how to use every part of what's going on in here to get it to work properly with what i was using it to accomplish, that because of this workflow, it stepped up my understanding of the entire generative ai process, even parts of automatic1111. I know that sounds overexaggerated for what this resource is, but, it fundamentally helped me understand more the relationship between the coding of .py files and how the computer interprets and uses them in comfy, so thanks man.
this led me down a rabbit hole that gave me a deeper understanding of all things generative AI related, and i know that's not probably your intention, but i thought would be a cool thing to hear you helped someone accomplish, so for whatever that's worth, and if it's not worth much to you, i also tipped you a couple hundred buzz. thanks again, cheers.
This means a lot, thank you. I went down a rabbit hole myself to create the workflow and had quite a fun time doing it, so it's nice knowing that it's helped someone else interested in how AI works :)