Workflow for XL/Illustrious/NoobAI Models

Using the new Detailer at 4:3 resolution results in it being unable to refine the image at certain point, leaving this weird image tearing where 4/5 of the image properly got Detailed while the 1/5 doesn't. What happened here?

Hello. Thank you for providing this. I opened it, but it appears the contents of the Standard and Basic workflows are empty. Also, it seems like Advanced and Advanced TIPO are reversed? I am writing to inform you.

There is a problem with install Z-Tipo:
How to install TIPO custom node? It have an error.
How to fix this?

I found a way to solve the z-tipo-extension installation problem (if the GPU is not essential for TIPO): install the CPU wheel even on Python 3.13:

python.exe -m pip install -U pip setuptools wheel

python.exe -m pip install --upgrade --force-reinstall --no-cache-dir llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu --prefer-binary

Thank you so much for the constant workspace updates!

As someone who's dumb as bricks at this, I just wanna say thanks for the easy to use workflow <3

This is excellent! But for the life of me, I can't figure out how to use the regional prompter. Putting a red-green-blue image in all three nodes and using simple "1 girl, red dress, red hair" prompts for all prompts just results in one girl with 3 mixed haircolors. Any pointers on correctly using this feature?

Hi there, I just wanted to start by telling you your workflow is fantastic! I was hoping you would be able to help me with the regional prompting. I'm using the main positive and negative for the scene, then using two colors, red and green, to try and apply loras to the characters in the scene. However the loras are still mixing. Would you happen to have any tips or recommendations? Thank you!

Nice Job :)

After installing comfyi, it always gets stuck at 'Restarting backend to apply changes' What should I do?

I've been using V31 for a few days and ran into an interesting problem.

After ~100 generations, the Negative node from comfyui-impact-pack causes color oversaturation on generations.

It only happens with Illustrious models (switched models for a time and didn't notice any oversaturation).

Restarting nor fully closing comfyui doesn't seem to fix it. However, if I load an earlier generation and use that workflow then the colors are back to normal.

I bypassed then deleted most of the workflow testing to see where the fault was. Positive and Negative were the last custom nodes left so I figured why not and bypassing Negative resulted in a fix.

I don't know much about these nodes or the workflow so I can't say why this is happening.

I really like the detailer but sometimes it becomes too aggressive. Like "turning male face into female" aggresive. Usually happened when the male face, either due to the character themselves or influenced by style Lora, looks kinda like female-ish. But could still happened even to male with very manly face.

I thought I could prevent this but using lora of the respective male character but turned out it's not. I haven't troubleshooting deeply like turning off the loras that I usually use, so it might be just my setting.

The lora loader (lora manager) does not see any loras I have in my lora folder. Only some older stuff I deleted from the folder.

Is there a way to set it up so the detailers occur after the upscaler? Been finding that the detailers will do their work and the upscaler undoes all the fine precision and results in a larger picture of lower quality than it received :(

Something to consider adding to the workflow (if compatible): LoRA Stack Weight Optimizer - Perfect Blends by LatteLeopard

Does anyone know why when I decide to upload a picture on the basic one as reference, first it doesn't follow the picture and second it adds a weird discoloring to the faces only...

I have a normal setup from lora manager workflow that works but I am trying to get something with more detail