About This version:First release tested with Flux.2-Klein-9b, this should work with the 4b model that I don't use.The text encoders are not standard, they are the ones from the suite comfyui-prompt-control which allow for better response to prompting.These are my first shared little works, so be kind :-)This workflow allows the use of 3 different pictures for image-to-image editing.Just like in the Pulid workflow you will find the TextEncodeQwenImageEditPlus aside the text encoder, not the standard one but the one from the suite comfyui-promt-control.I'm switching more workflows that I can to this text encoder because it seems to me that it allows better prompt following, also considering that with flux.2-klein it's better not to raise the cfg over 1 and the Prompt Weighting like (word:weight) can't be used because there is a serious risk of introducing ugly visual artifacts, distorted colors, or missing elements into the generated image.This not not effective as the head swap workflow or a pose controlnet but the exemples like "change the shirt color" are already covered :-)Speaking of it I'll publish shortly my version of an Head Swap for multiple Head Swaps.I've set the parameters for klein 9b distilled so 4 steps and cfg 1.Update! Version 2: In this new version: <ul><li>Power Lora Loader is correctly connected (sorry for that…) </li><li>I changed TextEncodeQwenImageEditPlus with TextEncodeQwenImageEditPlus lrzjason which support loading and processing of 5 images. TImages are automatically scaled to the correct size, so you can add images without having to worry about their dimensions being divisible by 8, 16, 32, etc. About that, keep crop disabled to allow the model to read the full hair and neck structure of your face references. That said, facial resemblance with this type of workflow is relative; you might want to perform a subsequent head-swap on the character itself using the head-swap workflows I will be publishing shortly. </li><li>It is always the resolution of the first image that determines the resolution of the final image. </li><li>Added upscaling which you can keep bypassed. </li><li>Added some promting tricks for Klein in a markdown note inside. </li></ul>Update V3.0I’ve updated several aspects of my Flux.2 Klein I2I and T2I editing workflow. I want to acknowledge and thank <a target="_blank" rel="ugc" href="https://civarchive.com/user/subotaplaya565">https://civarchive.com/user/subotaplaya565</a> for the work upon which this update is partly based. What’s new: • Replaced the TextEncodeQwenImageEditPlus node with the Painter Flux Image Edit node.• Includes 5 image-loading nodes/groups: one each for inpainting, face cropping, outpainting, and background removal.• I considered adding a PuLID group as well, but I’ll be releasing a dedicated "3-PuLID" workflow update in the coming days.• The workflow is highly automated, minimizing the need for repetitive settings adjustments.• I’ve included detailed Markdown notes containing instructions and tips. NEW v 4!! You have to try this one!! NEW MODE T2I or I2I with FLUX.2 Klein IDENTITY GUIDANCE This method replaces the PulId system in a simpler and more efficient way—without the need for a LoRA, simply by using the FLUX.2 Klein Identity Guidance node; now its very simple to maintain identity during editing and new generations. MORE: now you can EDIT an image choosing new dimensions for the resulting image. See inside Markdowns. Note: the three close-up portrait images are generated with the upcoming Krea 2 workflow. • I’m running a card with 16GB of VRAM and a reasonably up-to-date version of ComfyUI; any missing nodes used here should install without major issues.I don’t have much experience adapting workflows for low-memory systems, but for anyone wishing to make the necessary modifications... the workflow is right here :-) Have fun, don't break anything :-)

Flux.2-Klein Image To Image editing

<pre><code>I’ve updated several aspects of my Flux.2 Klein I2I and T2I editing workflow.

I want to acknowledge and thank https://civitai.red/user/subotaplaya565 for the work upon which this update is partly based.

What’s new:

 • Replaced the TextEncodeQwenImageEditPlus node with the Painter Flux Image Edit node.
 • Includes 5 image-loading nodes/groups: one each for inpainting, face cropping, outpainting, and background removal.
 • I considered adding a PuLID group as well, but I’ll be releasing a dedicated "3-PuLID" workflow update in the coming days.
 • The workflow is highly automated, minimizing the need for repetitive settings adjustments.
 • I’ve included detailed Markdown notes containing instructions and tips.
 
 • I’m running a card with 16GB of VRAM and a reasonably up-to-date version of ComfyUI; any missing nodes used here should install without major issues. 
I don’t have much experience adapting workflows for low-memory systems, but for anyone wishing to make the necessary modifications... the workflow is right here.</code></pre>

Files with SHA256: 9255fc366ed6fc726d5b68368ad867e2ef78707f430e3ceda047c58300b07c7c