Text encoder: https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/text_encoders/qwen_3_4b.safetensors
VAE: https://huggingface.co/Comfy-Org/z_image_turbo/blob/main/split_files/vae/ae.safetensors
V2.0: Nodes has been fixed.
V1.0 Reactor version:
Reactor version : In some installations Reactor has some dependencies issues that could break your ComfyUI installation if you don't know very well what you're doing, So make sure to make a backup of your VENV folder before using the Reactor version.Commissions Open.
Description
FAQ
Comments (28)
which facedetailer node is this from?
facetailer from comfyUI-Impact-Subpack node
and restoreface from the Reactor node.
It some installations Reactor has some issues installing dependencies, So make sure to make a backup of your VENV folder before installing.
@Hikarias bricked my install unfortunately
how do you install this model?
Download the models from the description and put them inside your Comfyui folder.
The model goes inside the unet folder.
Vae inside vae folder
and Text encoder inside your Clip folder.
This is just a workflow for running the new model Z_Image_turbo.
How to use JSON files in comfyui
unzip the file, Drag the .json file into the browser with the comfyui window open.
Hi, after installing missing nodes for the workflow, these 2 nodes still came up missing :
ReActorRestoreFace & LTXVFilmGrain
Can I just disable these two ?
Edit : I think I cured the ReActorRestoreFace problem by installing Comfyui-ReActor node. Now its just the film grain.
Edit : Fixed the LTXVFilmGrain by installing comfyui-LTXVideo node. Also needed to manually replace a file according to their github bug report. Skill issue on my behalf. I now have Jennie Blackpink in my computer. Great workflow though!
How your final using about "FilmGrain" node ? thanks
wow, so simple yet so powerful. thanks
eh Workflow encrypt?
Download it again, Fixed
Got it working great! Will you be doing an IMG to IMG version in the future? Also anyway to integrate reactor face swap from a reference image using this workflow? I am new to comfyui and find it difficult compared to forge, but your workflow is working perfectly for me now after following omnyom
This workflow with those nodes literally broke my comfyui environment. Idk why, but i have to start and install all from scratch. I'm furious.
I think it's the reactor node, That's the problem with Comfyui; some nodes are incompatible with certain packages, and not everyone agrees on which version to use.
You can remove the reactor node and it should be work fine.
That's why I always make a copy of my VENV folder.
Don't you worry,happened to me twice c:
@Hikarias I learned it the hard way hehe. After 5 hours i managed (idk how) to recover 90% of all my workflows and configurations, but i don't have the guts to try and install again the dependences to this and others nodes, at least for a while xD
@DarkGirlAngel I've changed the problematic nodes in the latest version, you could try it if you want, of course making a backup first just in case.
mannnnnnnnnnnn
People really should look and manage requirements.txt by visual inspection before updating nodes. Also avoid using ComfyUI Manager to update nodes. Just "git pull github-node-repository-url" and manage your nodes by yourself and never again have broken ComfyUI install.
I am doing this everytime when trying to use china's workflows :'D chatgpt helps to prevent errors but it is a hard work
They must have trained this model on the Flux dataset, it has the exact same Flux "sameface". Prompt adherence is OK, but good luck getting anything that does not look like everyone else's images.
works really well, thanks so much.... ❤️
Superb workflow, thank you
Can you tell me what the problem might be?
Failed to validate prompt for output 9: * UltralyticsDetectorProvider 28:1: - Return type mismatch between linked nodes: model_name, received_type(CLIP) mismatch input_type(['bbox/face_yolov8m.pt', 'bbox/hand_yolov8s.pt', 'segm/face_yolov8m-seg_60.pt', 'segm/person_yolov8m-seg.pt']) * SAMLoader 28:0: - Return type mismatch between linked nodes: device_mode, received_type(MODEL) mismatch input_type(['AUTO', 'Prefer GPU', 'CPU']) - Return type mismatch between linked nodes: model_name, received_type(IMAGE) mismatch input_type(['sam_vit_b_01ec64.pth']) * UltralyticsDetectorProvider 28:2: - Return type mismatch between linked nodes: model_name, received_type(VAE) mismatch input_type(['bbox/face_yolov8m.pt', 'bbox/hand_yolov8s.pt', 'segm/face_yolov8m-seg_60.pt', 'segm/person_yolov8m-seg.pt']) * FaceDetailer 28:3: - Value 458428318300401.0 bigger than max of 16384: guide_size - Required input is missing: clip - Value 5.0 bigger than max of 1.0: denoise - Required input is missing: model - Required input is missing: image - Value 4.0 smaller than min of 64: max_size - Failed to convert an input value to a FLOAT value: bbox_crop_factor, center-1, could not convert string to float: 'center-1' - Required input is missing: positive - Required input is missing: negative - Value 5.0 bigger than max of 1.0: bbox_threshold - Failed to convert an input value to a INT value: steps, simple, invalid literal for int() with base 10: 'simple' - Required input is missing: vae - Failed to convert an input value to a INT value: drop_size, False, invalid literal for int() with base 10: 'False' - Failed to convert an input value to a INT value: cycle, , invalid literal for int() with base 10: '' Output will be ignored