This is an advanced workflow designed for people familar with Comfyui and who want pro grade results. I encourage you to learn it, but if it is overwhelming, the simple version can be found HERE
A complete user and troubleshooting guide inclusing a list of common error messages can be found here https://civarchive.com/articles/24678/how-to-use-my-workflows-a-comprehensive-breakdown
Each Flux model has it's own combination of text encoders. Read the notes carefully. You will get MATT errors of it is incorrect.
也有中文说明
Ver 6.0
Controlnet/ Image reference area, gut & rebuild:
you can now set the image ratio based on your original image. Huge timesaver for Controlnet & I2I
Base image 🖼️ controls everything, so no need to set multiple images
You can even use it as a color reference in post processing
Better Prompt generator (Florence)
Huge update to detailers. I basically yanked them out and replaced them with my V5 units.
You an use Flux or use an SDXL based model to detail. Just switch it.
Updated 🎨 post production suite that includes realism enhancers and an inline crop tool
Updated the main area
Huge update to 🗒️notes.
Ver 5.0
Added controlnet
Added a sentence remover to prevent CivitAi Flags
Ver 4.0
Total overhaul
Optimized for Klein_9b
Options to control the scheduler
Upgraded detailers
Upgraded Post Production Suite
📷 Optical Realism upgrade
Arcane Bloom FX
LensFX
ClarityFX
也有中文说明
Flux 2D I2I, T2I, Upscaler, Detailers, and a hefty Post Production Suite
Optimized for Klein as well as Flux 2D
Include the Klein Enhancer (Similar to a detail daemon) to improve the quality of the Quant images
Prompt Assist
Allows up to 10 images to be loaded for reference
Ultimate Upscaler package
Seed VR2 Upscaler
Face, Hand, Eye, and Expression Detailers
Add lighting, color correction , and specicl effects after generation in the Post Production Suite
Flux 2D Base Model
Flux 2D Model (You have to agree to their terms first):https://huggingface.co/black-forest-labs/FLUX.2-dev/tree/main
Flux 2D GGUF Model https://huggingface.co/gguf-org/flux2-dev-gguf/tree/main
Flux 2D Mistral text endoders https://huggingface.co/Comfy-Org/flux2-dev/tree/main/split_files/text_encoders
Flux 2Dev Turbo Lora https://civarchive.com/models/2266313/flux2-dev-turbo-lora?modelVersionId=2550958
Flux 2 Klein Model & Text encoder
Klein 4b
Text Encoder qwen_3_4b.safetensors
Flux 2 Klein diffusion_model Flux-2-klein-base-4b-fp8.safetensors
Klein 9b
Text Encoder Qwen_3_8b.safetensors
Flux 2 Klein 9b FLUX.2-klein-9B
VAE (for both)
Instagram: https://www.instagram.com/synth.studio.models/
Buy me a☕ https://ko-fi.com/lonecatone
This represents hundreds of hours of work that I give away for free. If you enjoy it, Please make sure to 👍like and post to my workflows. That is how you support me. Also, feel free to ⚡tip 😉
Description
added controlnet
added sentenmce remover to prevent CivitAi flags
FAQ
Comments (30)
Everything works perfectly. Special thanks for adding ControlNet to the workflow.
This is really great job! I have a quastion. Do you know the setup if i have trainded character LoRa for flux.2-klein-9b and i would like to keep LoRa consistency and use reference photos for changes? Is there a way to control the controlnet strengh?
You can place th eimage in the first image slot. If you have a LoRA, place it in th lora loader.
I also have a consitent character workflow.
iam having trouble in using controlnet in Flux 2 klein 9b
loaded image , kept AIO as depth /hed, but not working
Wierd. Do yo have it turned on?
message me. I'm not sure what the issue could be. I got it to work 100% of the time in test runs. I will say it requires a good image though
@lonecatone23
i just imported your workflow , and enabled only below using bypasser
Load Image
Remove Background
Controlnet
Klein Conditioning
#Create Subfolder
#Save Dra ft
#Save w/Metadata
added an image in load image and generated a image , but pose in not as per loaded image
https://ibb.co/MDR347H9
cant get control net to work either - did you happen to work it out?
@isit I've never really had a problem with it.
Understand that this is not a controlnet. it takes controlnet and uses it as a reference image. I haven't looked at this workflow in a while, but I'm pretty sure I left good notes on how to use it.
i actually found a significantly better way to do it. I just haven't released it yet as it's not finished
@lonecatone23 true, with more playing and not trying to use it like a other model controlnets ive worked out some of its limitations, moving towards a more i2i workflow and replying more on prompting vs image+lora+cn thinking
@isit Yeah, I found a node group that absolutrluy nails it. I'l probably make a workflow this weekend.
we want nsfw version :)
So go to .red
@lonecatone23 i cant see NSFW version, Fix that image not same :)
@denshk93805 It works for both
FLUX.2 Klein Enhancer mb will be updated and this workflow now broken
Did you update? It's fixed. Try downloading it now
Really disappointed that I couldn't get this to work. It looks really promising. Comfy says the workflow is corrupted, several of the nodes fail to install (the FX nodes), I updated comfy and installed the last updates, even tested earlier versions of the workflow, nothing works. 😪
Yeah, this is two updates back, so I'm not surprised. I'll look at this though. FX nodes typically aren't problematic unles you are running on desktop or haven't updated. I'll hit you up when I figure it out.
Edit: I just checked. There is one node that doesnlt work anymore and that's the Sentence filter node. The crt nodes (FX) were updated two weeks ago and sem to be fine.
I'm going to update the worflow this weekend though.
@lonecatone23 I look forward to your results! At a glance I can see the potential, Klein is already a ridiculously powerful tool, with all these features you are adding.... I can only imagine the possibilities! 🙏❤️
@RLS_Animation On my to do list after Ideogram build
@RLS_Animation ask and ye shall recieve 🙏
@lonecatone23 TY!! , I will test it out and let you know the results!👏👏
If it works as planned, trust me I will put it to good use, and show you the results!
@RLS_Animation make sure to update. ComfyUI messed up a lot of thing the last month
@lonecatone23 DING! 🥳
I got it up and running! Thank you so much for this! It's certainly going to level up my videos. I currently have something in the works using your workflow, and I will add it to your gallery when it's done!
@RLS_Animation Yay! Tell me what works and doesn't. This is built for power users like you.
@lonecatone23 Off the bat, this is more than I could possibly ask for. Between allowing for multiple reference images, detailers, and ControlNet processor, this covers pretty much everything. Luts are just the cherry on top. If I was to ask for anything else, it would be a strength dial for the controlnet to lock in the most stubborn poses and dynamic angles, but even then, I have other methods to make that happen. More so out of curiosity than necessity, how many reference images can Klein handle? I know that OpenAI and Nano Banana Pro are sitting at 7 reference images. 7 images is overkill, currently 3 makes sense, 2 characters and a background. However, if a fourth reference image could be achieved, then it would be possible to have 2 different character reference sheets, a background image, and then a pose guide image. That would allow for most interactions between two characters with interchangeable backgrounds and poses.
@RLS_Animation 10 image references (9 if you subtract the cnet one) Just daisy-chain them. I think I left notes for it)
I've tried more than 4 with mixed success.
Controlnet needs a very high quality image to work. It does not work well with images of low contrast or blurry ones. Still, it's kind of a cool feature, and I've goteen it to work fairly well when I use it with a good prompt
@lonecatone23 Here is the results thanks to your workflow
@RLS_Animation Oh wow..... 😍
Thanks for sharing!
@lonecatone23 Thank you so much for the tip! Your workflows are simply amazing! 🙏