V4.1 Updated img4way switch. All inputs are optional instead of required so ControlNet nodes can be disabled without causing an error.
V4.1 Has ControlNet workflow with new custom switche nodes to make swapping between txt2img, img2img, inpainting, and controlnet easy as pie.
Suggest bypassing CNets that are not in use so help with render times.
Custom nodes include: 4way process switch, 4way CNet switch, and VAE switch.Drop NK custom nodes into custom nodes folder.
**You can disable or mute all the ControlNet nodes when not in use except Apply ControlNet, use bypass on Apply ControlNet because the conditioning runs through that node.
Custom nodes used in V4 are: Efficiency Nodes, Derfuu Modded Nodes, ComfyRoll, SDXL Prompt Styler, Impact Nodes, Fannovel16 ControlNet Preprocessors, Mikey Nodes (Save img node), and pythongosssss Custom Scripts.
V3 has all the features of V2 but now includes Wildcard prompter and Style selector. Three custom nodes are included for Aspect ratio and 3way latent switch and VAE switch for easy switching between baked and optional VAEs. Drop nodes in "custom nodes" folder.
V3 also has an improved workflow, I found the lora stacks were distorting the hirez pass. Now the hirez pass has no lora stacks just the loaded checkpoint which seems to improve the output.
V2 has inpainting and custom 3 way switch node for easy swapping between txt2img - img2img - inpainting.
A compact, easy to use txt2img and img2img workflow with 2k and 4k upscale.
Custom nodes used are: Efficiency Nodes, ComfyRoll, SDXL Prompt Styler, Impact Nodes.
I've included the edited aspect node as it does not contain a "custom" aspect ratio. Drop in your custom nodes folder.
Use a 2x Upscaler for the 4k upscale.
Description
FAQ
Comments (25)
What are the .py files for?
Custom nodes that I made.
@Nokode I installed these, tho I'm still missing a few nodes. Found CR nodes, but most of the other items in that list are nowhere in the UI's manager tab that I could see them. Are these in ambiguous packs that doesn't name them hence are harder to find with a simple search?
When loading the graph, the following node types were not found:
CR LoRA Stack
Get image size
Image scale to side
OpenposePreprocessor
Zoe-DepthMapPreprocessor
MiDaS-NormalMapPreprocessor
CR Apply LoRA Stack
Save Images No Display
Nodes that have failed to load will show as red on the graph.
@SuperCakeToast Get image size, image scale to side are derfuu modded nodes, Save image no display is Mikey nodes, and the controlnet nodes are Fannovel16 ControlNet Preprocessors. CR is Comfyroll nodes.
@Nokode IDK what I did wrong but while trying to install dependencies here and there ended up with that massive "to get" list
"Mask Smooth Region
Image to Latent Mask
Image Rembg (Remove Background)
CLIPSeg Masking
Mask Dilate Region
CLIPSeg Model Loader
Image Filter Adjustments
Image to Noise
Image Mix RGB Channels
Image Levels Adjustment
Image Select Channel
Image Size to Number
Image Crop Location
Mask Crop Region
Images to RGB
Image Blend by Mask
Bounded Image Crop with Mask
Bounded Image Blend with Mask
Image Padding
Mask Erode Region
Bounded Image Crop
Constant Number
Number Operation
Image Bounds
Inset Image Bounds
Image Blank"
whn ever i try to use your this workflow, i face below error>>
"When loading the graph, the following node types were not found:
NK_VAESwitch
NK_ImageInputSwitch_4way_v2
NK_InputLatents4Way
Nodes that have failed to load will show as red on the graph."
above missing file you are given in zip file, but i don't know where should i have to place that file to fix above error.
Please mention the location to fix this issue
Thanks a lot for sharing this ! Have you plan a tutorial for explain how to use this workflow ? I'am a comfy beginner and be a little bit confused with it
I tried to make it as easy as possible but if there's anything you're confused about you can ask here and I can do my best to help. Also Scott Detweiler on YouTube uses ComfyUI and has a playlist of the basics and all the way up to more advanced stuff.
@Nokode Thanks ! I have good results with Txt2img and Img2img but have an issue with controlnet : "Error occurred when executing Zoe-DepthMapPreprocessor: expected size to be one of int or Tuple[int] or Tuple[int, int] or Tuple[int, int, int], but got size with types [, ]"
Have you an idea of what's going wrong ?
@Eniotnadgt What depth model are you using? Some of the controlnet models require the "diffControlnetloader".
Thanks, I've enjoyed your work in here, could you ad Ultimate SD upscaler 2x + 2x to reach 4k? thank you!
y como se usa, se agrega a una carpeta con cualquier nombre en el custom_nodes ?
maybe i am late but sorry to ask am getting : When loading the graph, the following node types were not found:
• NK InputLatentsText
Nodes that have failed to load will how as red on the graph. how to make them work ?? thanks
Make sure you have dropped the custom node file into the nodes folder.
@Nokode yeah thats what i did , it still gives this message. do i drop the files in a file named NK ? OR i just drop all files in the custom nods ?
@miketuffin Just drop them all in the custom nodes folder
loaded .json file and dropped all files in custom_nodes folder, but getting this error when loading workspace:
When loading the graph, the following node types were not found:
•Get image size
•Image scale to side
Nodes that have failed to load will show as red on the graph.
and when i press 'Queue Promt', im getting this:
Cannot execute because a node is missing the class_type property.: Node ID '#6'
getting the exact same error, did you find a fix?
Would love to see this rebuilt, why was it archived?
i humbly request that this be brough back. has been my go-to for a while and produced the best results for me
the only things that are not working for me with this workflow now are this
Warning: Missing Node Types
When loading the graph, the following node types were not found:
Get image size
Image scale to side