I just thought I'd share a workflow I made for Flux using one and/or two upscalers for high res images. I wrote explanatory notes in the workflow itself.
You might need to install custom nodes, just hop on the manager and install them!
Description
FAQ
Comments (26)
Hi!. Please upload the json file too. This is just a screenshot.
The screenshot/image has the workflow embedded in it by the way, just drag it on your ComfyUI window
Thank you!
@HamedEmine Thank you - I didn't even know that was a thing!
You're most welcome! Glad I could help!
You might want to implement FluxGuidance to it. The cfg 1.0 currently gives especially human faces a kind of plasticy and unnatural look. Otherwise great job.
Thank you Vestu, I realized that I was missing out when I decided to ditch the Flux Guidance node completely, (as I recently found out: https://live.staticflickr.com/65535/53908898505_64fc547ff8_o.png ) I'm currently working on another workflow that uses that node, among other things
@HamedEmine Some people recommended guidance 3.5
@Vestu It depends on what you're trying to achieve, for example, I found out that a Guidance of 1.8 is excellent for realistic portraits. And going higher makes the skin unrealistic looking.
PS: In this first version of the workflow, the Guidance is set to 3.5 by default because the node is missing
@HamedEmine True, I noticed it too. guidance 3.5 makes thinks veer towards more contrast, sleek and bright look. 1.8-3.0 range is worth experimenting when wanting something more realistic, softer shadows, contrast, surfaces etc. more blended look perhaps
@Vestu Exactly!
Booo. Installing the custom nodes needed for this workflow crashed ComfyUI and it no longer starts.
I'm so sorry to hear this, were you able to fix it?
It seems that I was still comfortable even after the installation. Reports an error: ImportError: tokenizers>=0.21,<0.22 is required for the normal functioning of this module, no tokenizers==0.20.3 were found. Try: "pip install transformers -U" or "pip install -e".[dev]" if you are working with git main (((
Nice screenshot. No workflow in the file though.
Hello, the screenshot has the workflow embedded in it, just drag the screenshot to your ComfyUI and it will automatically load the workflow!
Drag the png image into your workflow space to load the workflow
I as sooo confused. thanks!
i got high rez fix error how cani fix this?
Select 1 to enable and select the first upscaler.. only number 2 works .. any solution
It is funny - switch 1 doesn't work but I ignore it and connected again to a new no. 4 :D
that image of the robot! I feel like it could go somewhere. Carniverous murder robots, thats kind of new.
How to get HighRes-script fix to work? I just dont see where this (-1) reference is.. the seed shows a huge large number, and I gues my diffusion_pytorch isnt correct? which should I use?
Prompt outputs failed validation HighRes-Fix Script: - Value -1 smaller than min of 0: seed - Value not in list: preprocessor: 'none' not in ['_'] - Value not in list: use_controlnet: 'False' not in ['_'] - Value not in list: control_net_name: 'InstantID\diffusion_pytorch_model.safetensors' not in []
I have the problem with image save, i don´t know why!
some one can help me preaty pelase?
When i download this workflow, i just get an image of the workflow, not the actual file. is there something wrong with folder?
Hello, nope, there is nothing wrong with the folder, the workflow is embedded in the image, so just drag the image to your ComfyUI and it will automatically load it