Flux.2 Klein Mix Workflow
Workflow with main focus on Image Editing, with control over Image Sizes. Trying to only use the more common Custom Nodes or ones that do not have requirements for a easy and safe installation.
Aiming for consistency and support edits for up to 4 images and Text to Image, 3 different Controlnets, different Inpainting methods and Outpainting. Nodes for selecting image sizes and Aspect Ratio. Uses Flux2 Klein Enhancer nodes, for consistency and prompt impact. Color matching and different sampler with ETA setting.
The workflow has many notes that explain things inside, which you can read or delete.
There is a second Version inside for people that use Lora Manager so you can pick what you want.
The workflow only support a Positive Prompt except for the 2-pass mode in V5 for T2I that uses both the Base and the Normal model and a negative prompt.
All custom nodes used:
Essentials
Rgthree
Easy-Use
Impact Pack
KjNodes
GGUF
Controlnet_aux
LayerStyle
ArtVenture
Below Does not have any requirements making for an easier and safer install:
Inpaint-CropAndStitch
ModdedNodes
Comfyroll
ControlAltAI
Flux2Klein-Enhancer
CRZnodes
mxToolkit
ZImageTurbo-FlowSampler
Note that V2 and V1 do not use the same custom nodes, and are harder to use because of that, and I do not recommend using them!
Feel free to criticize, comment or ask questions.
Description
FAQ
Comments (22)
Interesting workflow, I even went down the route of installing llama.cpp to run your WF.. from my understanding, I should have llama.cpp running at the same time i'm running comfyui in order for your workflow to work?
No you do not need that. I do not have that myself, so can't really help you with that. I only use Prompt Manager to save good prompts for easy access. But it has more features if you want like the llm_input but not something I have used yet. The workflow works by using the prompt manager node just like a regular positive prompt node.
@Whistler42 really? LOL, i've spent the last 10 hours configuring llama.cpp and Open webui because I thought the prompt manager required it, i mean, it said so in their page.. oh man, ahh well, i learned a lot in the process regardless.. I'll try your WF and see how everything goes.
not sure why I have to select a gguf model when I'm using the safetensors model?
also under Make Image subgraph clip text node is empty and red.
I added my prompt at Prompt manager node.
It will not use the GGUF model when you pick false on "use GGUF" but yes if will not work if you do not have one selected, you can pick any GGUF even for an other model as it is not used then, or you can go inside the Models sub graph and bypass both the "Unet Loader (GGUF)" and "use GGUF" nodes.
It should not be red. That node has two inputs Clip to Clip and Prompt to text. The Clip is broadcasted from the Loras node, so maybe that is your problem? Right click the Loras node to see if it is broadcasting. You can also connect the Clip from "Loars" to the Clip in "Make Image" and the Model from "Loars" to the Model in "Make Image" and it should work.
@Whistler42
conditioning from the prompt manager was hooked to text, hooked it to clip and it worked.
What annoys me is that I can't get the original resolution of the photo (width and height)
Also no negative prompt?
@drfaker911219 I had it that way before so it only down-scaled if bigger then input Mpx. But you are correct that would be good to have, I will update my workflow to add that.
You need to use the Base model of flux2 Klein to have a negative prompt and this workflow is not made for that. That also use many more steps but have seen there are ways to make it faster now so might update my workflow to work with the base model of Klein also.
comfyui manager cant find all of these extra nodes suite involved, why is it so common nowadays NOT to list all these exotic Nodes-suites used :
Probably because you think the Manager should handle it, and it is hard to know / remember what does not work without trying on a fresh install. I added a list of all custom nodes in the description now anyway. The light version might be easier to run also.
In OP's defense, I always install all custom nodes manually and then pip install the requirements, because Comfyui Manager has failed me many times, so I stopped using it.
"Mode Select " Node - Option is showing "disconnected"
How to fix it ?
Where to download the v2.1 ?
Inside the Mode Select Subgraph there is a node called int that is just a number. If it says disconnected you have deleted that node.
2.1 has replaced 2.0 as it had one broken function so I didn't leave it up. So just download v2 again.
Yeah, I have the same issue showing disconnected for both versions. I did not delete any nodes. Oh well.
Then the problem is that you do not have the custom node for the integer value inside the sub graph. I will try and make a new version with less needed custom nodes. The workflow has way too many now, should be alternatives that I can use.
This workflow is perfect, I really LOVE IT. So first BIG THANKS TO YOU. I can use it with comfy 3.16. But... I've updated today (2026 3rd march) and It doesn't work anymore. By chance I still have my 3.16. I manually reinstalled every customs node but nothing happens. I mean, the workflow return no error and runs super fast and gives nothing. I'm investigated.
Thanks, I think everything works for me still on the latest version but my whole comfy installation is messed up from updates with errors everywhere. So I'm working on making several smaller comfy installations that can do specific things. Also want to make this workflow require less custom nodes so will try and get things better then and easier to run then.
Well. Something very strange happen. When I run the workflow for the first time, it runs well, and next times it doesnt. When I save it it doesnt want to load. It's very suspicious
@quanticalcapybara886 Hard to say what is wrong if you do not have a error message. If you installed other custom nodes later maybe they messed things up? You can type pip check to inside your ComfyUI python environment to check for broken requirements. You could also try downloading the workflow again I guess. I have seen sub-graphs acting weird with newer ComfyUI version but I deliberately made changes for that so they at least seem to work for me now. I'm using ComfyUI v0.18.2 with Python 3.12.12
@Whistler42 Thanks for your reply, it works fine with 0.18.2.
неработающая херня! версии 2.1 в архиве на 28.03. нет ) Не тратте время