SMOOTH WORKFLOW
UPDATE: Smooth Workflow v.4!
With support for ZImage! Choose between a a workflow for ZImage or Illustrious/Pony generations.
Made it very simple to use. ^^
Both Workflows have to option to save with or without Adetailer/Face Detailer.
All images on the Showcase used the new SmoothMix Ultimate ZImage Turbo model!
ZImage parameters are very different for the Illustrious/Pony models. Make sure to not confuse them.
ComfyUI updates can be very frustating as they can make many extension incompatible/stop working - So I tried to use as little extensions as possible on v.4.
Shout out to Martin Rizzo (@Photographer) for making probably the best 'Save Image' Node compatible with CivitAI. It just works! =D
Used extensions on v.4:
ComfyUI-Impact-Pack and ComfyUI-Impact-Subpack
Small UPDATE: Smooth Workflow v3.1!
Facedetailer should work much better.
Used extensions:
Save image with generation metadata
ComfyUI-Impact-Pack and ComfyUI-Impact-Subpack
Advanced CLIP Text Encode (needed for workflow v.2 and v.3)
ComfyUI-FBCNN (Flexible Blind JPEG Artifacts Removal) (needed for workflow v.3)
UPDATE: Smooth Workflow v3!
Changed the node used for Positive Prompts: and ComfyUI-Impact-Subpack
The image results should propely show the effect of adding weights to LoRAs now!
The Positive Node is now divide 2 areas. Your Positive Prompt goes on the top one!
Use "Select to add LoRA" to open a list of all LoRAs you have installed!
Lets say you select "Smooth Booster v.3" on your list. It will show up on the prompt as <your_Loras_path\Smooth_Booster_v3>.
Remember to set the Weight you want on the end!! <your_Loras_path\Smooth_Booster_v3:0.5> Like so!
Check the showcase image for an example!
The bottom area is used to show the prompt that was executed. Its used for adding Wildcards and see the prompt used in the execution. If you don't use Wildcards don't worry about it. For more info check its Github on the Wildcards Nodes section.
Added extension: Flexible Blind JPEG Artifacts Removal
Added Compression Removal to the "(7) SAVE IMAGE" section.
This is a JPEG artifact/compression removal tool to help clean up the image a little. Its set to "auto detect - enable" by default!
More advanced users can disable it an modify the compression level as they wish.
For an indepth look at it check this Github!
For a even more indepth look check the Paper on Cornell Unniversity.
===========================================================
UPDATE: Smooth Workflow v2!
Added extension: Advanced CLIP Text Encode
Added CLIP to the "(4) IMAGE PARAMETERS" section.
If you are having dificulties recreating images from CivitAI or tools using A1111, like Forge, selecting "stop_at_clip_layer" to -2 will help to recreate images similar to those more easily.
For more advanced users the Advanced CLIP Text Encode was added to the "(3) PROMPTS" section
With these you can change both "weight_interpretation" between A1111, ComfyUI and others. For more info check the GitHub.
==========================================================
Its divided in sections to make it easy for begginers to navigate.
(1) CHECKPOINT / VAE
Very straight foward. Just choose your Checkpoint and VAE.
(2) IMAGE SIZE
Select your image size. I can freely adjust it. By default its set at 832x1216 since its the one used by CivitAI.
(3) PROMPTS
Type your positive and negative promtps here.
To use embeddings use this format: embedding:your_embedding_here
Example: embedding:SmoothNoob_Quality
To use Loras, use this format: <lora:you_lora_here:weight_number_here>
Example: <lora:Smooth_Booster:0.5>
Doing so will make sure CivitAI can detect your embeddings and loras propely.
(4) IMAGE PARAMETERS
Set the number of STEPS and CFG. By default its set as STEPS: 30 and CFG: 5.0.
Choose the Sampler and the Scheduler. By default its set as Sampler: DpmPP2M and Scheduler: Karras
You can also set the SEED in this section.
(5) HIRESFIX
Here the HiresFix will Rescale your image. You can choose the method: "by percentage", "set Width/Weight" or "to longer side - maintaing aspect". By default its set "By percentage" at 40. This will upscale your image to 1328x1944 if you set image size as 832x1216. Choose your favorite method.
Make sure to set your favorite upscaler model as well. By default its set to REMACRI.
(6) ADETAILER
Here you can adjust the face detailer. The default settings work really well already but feel free to make any adjustments.
(7) SAVE IMAGE
Here you can choose the path your images will be saved.
I installed ComfyUI using StabilityMatrix, so I set my path to F:\StabilityMatrix\Images\Text2Img\%date.
Remember to change this in accordance to your PC settings. I recommend leaving the %date since it create a folder with the date the images were created. It helps keeping everything in order.
=======================================================
Other recommend extensions:
Description
Small Update to Adetailer
FAQ
Comments (51)
Great workflow for beginners to learn. Thank you!
This has been great for me as a beginner, thankyou.
One question, is it possible to have the face detailer work for only one face in an image?
The most prominent one preferable.
I sometime find if I have a male and a female character in an image, the face detailer will feminise the male face.
it looks that this commit brakes the workflow:
https://github.com/alexopus/ComfyUI-Image-Saver/commit/3dd8beaf194101fcc8f267c40c73d4b3f5128bb0
in comfyui base dir:
cd custom-nodes && \
git clone https://github.com/alexopus/ComfyUI-Image-Saver.git && \
cd ComfyUI-Image-Saver && \
git checkout 1c0ab65f1e4e9fdd4524b6b1887e16f4f4ebd83c && \
cd ../../
THANKS @dannib in my case I did the same and deactivate ksampler
I'm getting a few conflicts. with scheduler selector in the comfyui impact pack, the latent switch. Also saveimagewithmetadata with comfyui_prnodes. Is this an update issue with this workflow?
It seems several Nodes were updated since I uploaded the workflow. I updated the 3.1 version. See if it work now. Make sure to update everything. ^^
Image-Saver Node updated and was causing conflicts on Scheduler Selector. I update the workflow.
Hi @DigitalPastel I didn't find the updated workflow, I miss something ?
Is anyone familiar with how to do an img to img using this? I tried and failed
I used your workflow when I started generating locally, there were a few things I had to mess around with to get it to work. It was a great crash course for me to learn comfyui and a bit of python and an amazing baseline for me to create my own workflow from. I cannot tell you how grateful I am for you uploading this.
Hello. Keeps stopping on downloading FBCNN models from Github. I tried to look, keep getting 404 errors. Can you please direct me to the right place to download these models?
[15/08/2025]Update: After many tries, I finally was able to download the missing models. It is now working fine, the results are amazing.
I just disabled the node. I'd be interested to know where to DL those model files though...
Recently after latest updates of ComfyUI i recive this Error whenever i try go use this worflow:
Prompt outputs failed validation: KSampler: - Return type mismatch between linked nodes: scheduler, received_type(['simple', 'sgm_uniform', 'karras', 'exponential', 'ddim_uniform', 'beta', 'normal', 'linear_quadratic', 'kl_optimal', 'bong_tangent']) mismatch input_type(['simple', 'sgm_uniform', 'karras', 'exponential', 'ddim_uniform', 'beta', 'normal', 'linear_quadratic', 'kl_optimal', 'bong_tangent', 'beta57'])
How to fix this?
EDIT - found a solution. You need to delete node Scheduler Selector (Comfy) (Image Saver) from workflow and it will generate images normally as it should. On newest version of ComfyUI-Image-Saver it is not avalible anymore
But how do we vary the scheduler then?
@oliOil Well.. we don't untill the author of orginal workflow fix this in other way.
@equinoxe_c64 This updates that break stuff are annoying. T_T
I'll see what I can do. Maybe a new version (v.4) is in order? I will look into it.
@DigitalPastel Thank you for your reply, and hope you will find the way to make this workflow in fullest glory :)
Just downloaded and latest version of the WF, and it don't work. Error:
Prompt outputs failed validation: KSampler: - Return type mismatch between linked nodes: scheduler, received_type(['simple', 'sgm_uniform', 'karras', 'exponential', 'ddim_uniform', 'beta', 'normal', 'linear_quadratic', 'kl_optimal', 'bong_tangent']) mismatch input_type(['simple', 'sgm_uniform', 'karras', 'exponential', 'ddim_uniform', 'beta', 'normal', 'linear_quadratic', 'kl_optimal', 'bong_tangent', 'beta57'])
Dude... just look on my comment next to yours... if you click "Show more" you will see a temponary fix.
@equinoxe_c64 yup, thank you. I found a solution mysalf, but thank you pointing me out your comment. And i also read the comment of the WF author, 2 months ago, he was telling it will update the WF. But two months later, the WF has not been updated lol
for some reason my version does'n allow anything in the set size box.. it's just grey space in there
I am having a problem on using this Workflow. Everything is very new to me. When this Workflow is opened, everything is blank that I have the Convert a lot of the Inputs to Widget.
Problem is, the image sample seen above is outdated and no longer looks the same as the one provided on the latest file.
The Image Result node doesn't seem to display anything after the image is generated... Is anyone else having this issue?
yes true
The Faacee Detail seems to stop working after the newest ComfyUI update
Thanks for commenting, had to Google why the workflow wasn't working and figured out, I could disable the face detailer to make it work. Elsewhere, they said to move the model fusion or something before the detailer but I don't see any node like that. If anyone can provide a fix, it would be appreciated.
@allofdarkness https://github.com/ltdrdata/ComfyUI-Impact-Pack/issues/1113#issuecomment-3368356145 that should be a fix for it works for me
great
top
Good to learn off of
Thank you!This workflow is very helpful for me to learn!
I keep getting this even though i have the sam file in there... Prompt outputs failed validation: SAMLoader: - Value not in list: model_name: 'sam_vit_b_01ec64.pth' not in []
i cant even find the sam loader to double checkit
Hey tried to transition over to ComfyUI from Forge but with your workflow i keep running into a consistent issue, the hands keep merging with parts of the body. Using the same SmoothIllustriousV5 checkpoint on Forge, I never had this issue, in fact, the outputs turn out amazing. But with ComfyUI using this workflow, I can't seem to stomp out this issue and i tried using the same prompts, the same loras as with the forge version but that hasnt helped. Think you can lend me a hand?
oh? Usually crazy body deformation/positioning is related to the image size used. Here is a quick post I made with a few images using illustriousv5.
Funny thing is whats happening to you usually happened to me but the other way around. lol Forge is the one that would mess things up.
There is a possibility that in Comfy UI you need to set the values of steps and cfg lower than in Forge. Also, do not forget that in Comfy UI clipSkip
-2 is equal to clipSkip 2 in Forge
Is there a reason why in the workflow you use the sdxlVAE ? Can we use the VAE inside your models or you suggest to use the default sdxl?
Excuse me. Could you please tell me how to link ControlNet on this workflows? I'm not sure where to connect the positive and negative parts.
Given groups=1, weight of size [64, 3, 3, 3], expected input[1, 4, 156, 90] to have 3 channels, but got 4 channels instead It says this and I can't proceed. What should I do?
Check your VAE or CLIP
This WF can only create 1 image, I can't find the batch size
It is underneath the Width x Height node. Just unpin it and move it out of the way to get to the batch size.
thanks for the workflow pretty awesome, love u
Nice workflow. Was looking for something fairly straight forward. It does what it says on the box without installing a slew of weird, one-off nodes.
Thanks!
[FIXED for me, see edit. Workflow is great!!!]
Hi, I keep getting the error in the FaceDetailer node " AttributeError: 'DifferentialDiffusion' object has no attribute 'execute' " any idea why? Everything else seems to work just fine
EDIT: https://github.com/ltdrdata/ComfyUI-Impact-Pack/issues/1113#issuecomment-3378720742
update your comfyui, check to see if it works, if it still doesn't then follow instructions in this link
thanks to squarepandahead for pointing me in the right direction.
why is it impossible to find workflow that just works out of the box without breaking whole comfyUi install, or running into random errors, and thats issue with EVERY SINGLE OUT OF 50 SDXL/ILL workflows I have tried. probably time to give up on SDXL/ILL models and move to 22 century, with QWEN, Zimg etc.
It would be no problem if the issue was just missing nodes. but for this one i cant even install the neccesary nodes for some reason...
Because you have to install all the nodes and all the dependencies (Python libraries) manually, carefully checking for version conflicts between the dependencies of each node. Because each node developer does what they want, and many nodes are outdated with old dependencies. Never use the ComfyUI Manager, ever. And also, when you load a third-party workflow, it loads it with the node versions that the developer has, and sometimes you have to manually re-upload some nodes to update them to the version you have installed. But these things have always existed.
"Failed to validate prompt for output 51:
* SAMLoader 30:
- Value not in list: model_name: 'sam_vit_b_01ec64.pth' not in []
Output will be ignored
Failed to validate prompt for output 75:
Output will be ignored"
Someone could help me??
The default workflow is likely calling to a model you don't have in your models folder (either by path or specific name). Find the node with 'sam_vit...' - click the model, select a model you have, then re-run.