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
Added:
*New Layout with new nodes for settings.
*New default sampler called midpoint, easy to change back to euler if you want.
*Options to set ETA and S_noise.
*Inpainting menu with segmentation, box and crop options with previews.
*New Ootpainting and Controlnet menu-layouts with preview.
*4th Image input.
Changes:
*Better handling of image sizes and changed to pixel height instead of Mpx of the final image. Added a preview node that will show the final image resolution. Plus some changes for image size calculations.
*Rerouted the Enhancer node to get a somewhat better effect.
*New default vaules for the Enhancer nodes with notes for them and the dev defaults.
*All notes rewritten and new ones added.
Fixes:
*If you set output size lower then image 1 the input will no longer be cropped.
*Fixed a bug when using gguf version of the klein model.
FAQ
Comments (6)
Hi, sir,
The v3 workflow is bravo!
But when I use the v4 workflow, the following error will occur. How should I fix it? thanks!
WebUI:
can't access property "output", res is undefined
Console:
Failed to validate prompt for output 1203:
* SamplerCustomAdvanced 2827:2723:
- Return type mismatch between linked nodes: sampler, received_type(IMAGE) mismatch input_type(SAMPLER)
* SamplerCustomAdvanced 2827:2716:
- Return type mismatch between linked nodes: sampler, received_type(IMAGE) mismatch input_type(SAMPLER)
* SamplerCustomAdvanced 2827:2734:
- Return type mismatch between linked nodes: sampler, received_type(IMAGE) mismatch input_type(SAMPLER)
* SamplerCustomAdvanced 2827:2729:
- Return type mismatch between linked nodes: sampler, received_type(IMAGE) mismatch input_type(SAMPLER)
Output will be ignored
Failed to validate prompt for output 1172:
Output will be ignored
V4 has 3 new custom nodes they are listed at the bottom of the list both inside the workflow and here on the Civitai page. The error you are seeing is because you do not have the "Z-lmage Flow Sampler" node that is inside the "Generation Settings" Sub graph. So you need to get the ZlmageTurbo-FlowSampler custom node.
@Whistler42 Yeah, I am getting this same issue, "can't access property "output", res is undefined" on both V3 and V4.. I also tried a fresh install of all the nodes to rule out conflicts but no joy. I have safetensors disabled and Q8 versions of both the model and text encoder loaded so not sure if that could be the issue.
@PlayBoi_AI For V3 there is an error with GGUF but you should get an other error. You can solve it on V3 inside the models sub-graph by manually enable GGUF node and connect model to model directly as my method somehow only worked for Safetensor.
But on V4 GGUF should work, I tried 2 different GGUF for both text encoder and Klein and it worked for me anyway. For V4 you need to put the option below to 2 and make sure GGUF is not disable in the node below the model load node. There is a note that explain that also.
And my solution above for the error above error should work, I tested and bypassed the Z-image flow sampler node inside the Generation Settings sub-graph and I got the same error. You could also go into the Generation Settings Sub-graph and replace the Z-image flow sampler for a regular KSamplerSelect node and connect it in the same way to test if that works better.
I'm using newest ComfyUI version 0.18.2 with Python 3.12.12
very interested to try it, unfortunately all it return is just black image for all modes, looks like it doesn't even load the model to the vram, it returns a black image in like 2 seconds, 4090
Can you post what the log says. Click the Console bottom at the lower left of ComfyUI. I can guess that you do not have all Custom Nodes, but I can not be more precise without knowing what the error is.