Workflow for XL/Illustrious/NoobAI Models
Any feedback would be appreciated.
See the Article for a Guide:
https://civarchive.com/articles/17339 (probably doesn't work rn)
https://civarchive.com/articles/17339
Key Features:
LoRA
TIPO Prompting
Detail Enhancers
ControlNet Upscale
IPAdapter
HiresFix
OpenPose
Wildcards
ControlNET
Color Match
FBCNN
Regional Prompting
Inpaint
Outpaint
Workflows included:
🟥Advanced: comprehensive text to image workflow
🟨Basic: more basic version of the Standard workflow
🟦Standard: more basic version of the Advanced workflow
🟩Detailer: detailing previously generated or existing images
Experiment and enjoy!
Description
Update from V27→V28
(This is not an exhaustive changelog)
General:
General improvements and bug fixes
Adjusted default values
Changed checkpoint to Fabricated XL v7.0
T2I:
Changed Z-TIPO default values
Basic
Apply Signature added
Detailer:
Apply Signature added
FAQ
Comments (26)
z-tipo just flat out does not work
What's the error message? If it's the failing to import one, I fixed mine by manually installing a prebuilt llama-cpp wheel for my cuda version.
SO I am changing this post from the original because I am an Idiot! I did what you mentioned, Installing the llama-cpp for my cuda. Now everything is right.
So an update on me trying to get z-TIPO to work. Now I have a windows error ([WinError -1073741795] Windows Error 0xc000001d) This pops up in ComfyUI. I have no idea what is going on.
Same here. Comfy (portable) telling me that TIPO is not installed. Whether manager installation seems to work nor direct install via github.
Is it possible to replace TIPO node with another one?
@sweetyalex694Â I solved the problem by istalling the Llama-cpp. That resolved the install issue for me. Also, did you download and add the gguf for TIPO? I had forgotten that as well. Now, I must admit I may have done something wrong with the installation becasue its still not working. THe workflow will work without the node, just don't activate it in the Controllpanel
@Legendaer thanks for the response! can you give a tutorial on how to do this please?
@wackop Install it in your comfyui venv with the option that matches your system
pip install llama-cpp-python --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
or https://abetlen.github.io/llama-cpp-python/whl/<cuda-version>
it works for me. just makes the pics all the similar.
anyone know where i can find these detailers?
There is a section for it in the article. Here is the essential part copy-pasted:
🟥🟨🟩 Hands (hand_yolov9c.pt) 📂/ComfyUI/models/ultralytics/bbox ⚠Required for detailers since it's loaded as the pipe default
https://huggingface.co/Bingsu/adetailer/blob/main/hand_yolov9c.pt
🟥🟨🟩 Faces (face_yolov9c.pt) 📂/ComfyUI/models/ultralytics/bbox
https://huggingface.co/Bingsu/adetailer/blob/main/face_yolov9c.pt
🟥🟨🟩 Eyes (Eyeful_v2-Paired.pt or Eyeful_v2-Individual.pt) 📂/ComfyUI/models/ultralytics/bbox
https://civitai.com/models/178518/eyeful-or-robust-eye-detection-for-adetailer-comfyui
🟥🟨🟩 NSFW (ntd11_anime_nsfw_segm_v5-variant1.pt) 📂/ComfyUI/models/ultralytics/segm
https://civitai.com/models/1313556/anime-nsfw-detectionadetailer-all-in-one
🟥🟨🟩 Body (yolo11m-seg.pt) 📂/ComfyUI/models/ultralytics/segm
@Legendaer sorry i, obviously, didnt read the article. I have other detailers so i thought this would be more of a plug-n-play situation but then i saw that the detailers were "9c" and im like yeah, i dont have those. Lol. Thank you! I apologize for my laziness and oversight
Hello, using v28, finding it really interesting. A question: in the regional prompting group we have three separate "load image" nodes: how are we supposed to use them: loading, for example, three 1024x1024 blank masks with just the red, blue or green region colored? Thanks!
Hi, you can separate the colors in each Load Image node, or load an image containing multiple colors at once, both approaches work. Whatever is colored red in the "Load Image RED" image will be converted into a mask for the red regional prompt, and the same applies to the other colors in their respective Load Image node. And the image size isn't that important, since it will be resized to match the width and height of the generated image.
@Legendaer Thanks a lot - crystal clear.
@LegendaerÂ
Hello, I tried specifying the Red and Green regions in Regional Prompting, then using the Red prompt "1girl,solo" and the Green prompt "1boy,solo," but it didn't work. Any tips?
@jbh0716526Â Try and acknowledge these two people both in the general prompt (together) and in the regional prompts (separately). Something like saying: in this image there are two people, a boy and a girl; RED is a boy, who does this; GREEN is a girl, who does this; it's more structured and less random.
@McBob67Â @McBob67Â
Thank you. I'll try it.
Hello, the latest version of ComfyUI_windows_portable includes Python 3.13.
I'm getting an error with Z-TIPO. What version of ComfyUI are you using?
Please let me know which versions of CUDA, Python, and llama-cpp wheel are you using.
same error
ComfyUI 0.9.2 (right now, but also worked for previous versions)
python 3.13
cu130
https://abetlen.github.io/llama-cpp-python/whl/cu130
@Legendaer Thank you, it works well.
Hello, I really like this workflow but I have 1 issue.
Is there a way to have the Lora Loader scan other folders than the default ComfyUI one? Other Lora Loaders do this fine and pull the loras based on my extra_model_paths.yaml, but the one included in this workflow only pulls from the default ComfyUI Lora folder.
Hi, I'm not sure why it's not working for you. It pulls the loras from the specified paths in the extra_model_paths.yaml file without any issues for me.
I’m using BasicT2I. After upscaling the image, it reverts to the original version and loses all changes made by the face detailer, body detailer, and other detailers. How can I fix this?
I've been using an older, modified basic version of this workflow for a while, and I always found it very convenient and useful. This latest version is more of the same, but enhanced (I don't use all the tools, but I'm sure there are others who do). I'll be using this version for quite some time, so thanks.