It really works ;)
Description
FAQ
Comments (19)
Works great! I have question on lora merge weights though. Do you recommend .60? Is 1.00 strength too much for merging into a checkpoint?
recomended 0.6 or less. in my test more than 0.6 make big pixels ....
any workflow to merge two flux loras to create a new character?
Can you send me a workflow for text 2image? I have generated a model according to this workflow, but there is no usable workflow that can make it work smoothly
Hello. Flux with multi LoRA loader workflow | ComfyUI Workflow (openart.ai)
First run is always slow, next go to 34s per image. (RTX 3090)
@Aderek514Thank you very much for your reply.Also, I would like to ask if there is any integration of workflow that do not include clips and VAE(Because I have already downloaded the CLIPs and VAE files)？ so that it can save my very thin hard drive space.
@vigee88 This model has include vae and clip, u dont need this. Only use checkpoint loader like in workflow posted before.
@Aderek514 Sorry, I made a mistake in my expression earlier. What I mean is, I want to get a merged model that does not include CLIP and VAE. What workflow can complete it.Because this kind of model can save hard disk space, and most downloaders already have separate CLIP and VAE files on their hard drives.
@vigee88 u need to download python script and run it with special options, without comfyui
Do you know if there is a way to merge just 2 lora file to create 1 lora?
I have never looked for it, it is rumoured to be there.
I'm obviously a month late to this comment but you can always add the two loras you want to merge to the base model one at a time, and then subtract back out the base model to end up with a merged lora, just with extra steps.
@schonsense I also am obviously a few maoth late to your comment but thx lol
i have a question, i use this wf but result not good with big pixel, i merge 20 lora into the Fluxdev1
set lowest str of loras, or one of your lora is bad res and make bad to all
and in testing workflow use cfg=1 for new checkpoint, look on my wf: Aderek Flux fp8 workflow - upscale, watermark, detailer | ComfyUI Workflow (openart.ai)
@Aderek514 but i have big problem, my flux model in undentifield in "Load diffusion model" node, dualclip and vae too, how to fix this sir, i'm using google colab and i put model in "comfyui/models/unet", it's not working, ty sir
@VNStorms try to load into "checkpoints" or use checkpoint loader, not load diffusion model
@Aderek514 yeah, ty sir, i will test this