วีดีโอ

FLUX is far superior and the most realistic, I think SD 3.5 has a chance in the future, but FLUX for right now is on top of every base model.

Your link This page isn’t working

Interesting model and results, just want to put this major caveat out there: although this model creates interesting results, the workflow takes a long time to execute. Also, the base image produced by flux, in my workflow is supposed to be an uncensored image because I have certain stacks of Lora's that unlocks/uncensors Flux. The itercomp process not only translates certain issues/inconsistencies into the original photo generated by flux, the later steps proceed to censor the images all over again! Also it cannot fix deformed hand positions/wrong anatomy or poorly rendered content if flux just goes awry with the original image. That is probably the most annoying thing about the model. Also the itercomp process introduces NEW inconsistencies and issues with every pass, while it does add detail, it also sometimes ends up screwing up things in the image that is perfectly fine. I experimented with adding those uncensored Lora's in loaders back into the 2nd step of the workflow (your Itercomp flow and model) and that simply does not work, itercomp does the censorship again. I create a lot of content using certain artistic Lora's and all of those have issues in this workflow. This is definitely more suited to enhancing usage of vanilla Flux - and just to put this out there - with vanilla Flux and no Lora's this model does an excellent job. I would say this model is suitable for a lot of interesting use cases: however, uncensored images and complex shots of subjects holding stuff (Dreaded hands problem) isn't resolved. You are back to using mask/in stitch techniques to fix those. Overall, an interesting model, but will cause issues with certain stacks of Lora's. I would separate the Upscaling from the model completely, and reserve that step only after repairing the best content with institch/masking techniques in a separate workflow. A lot of trial and error. This isn't really the one stop shop and click technique to produce any kind of content one needs, but it can enhance what you can do with base Flux that is for sure. For my own needs anyways, I will go back to simpler step techniques - because this workflow is too large, and often will push through images that should be manually repaired before going onto later steps. And the censorship - makes this a no go for producing any of that kind of content. Good job again though, definitely learnt some things trying out this model!

Excellent workflows and explanations!

Thank you! Good job!

So many hoops just to test the workflow. MY GOD...

It’s the prompt I like in flux not image quality

the screwed up site doesn't like my email

detail daemon has been existed in a1111 extension for a long while.... haha comfy is late

When you have a website that requires login to download something for free they are 100% providing your email and other details to 3rd parties they have been contacted by. My spam folder is full of crap because of sites like this. When possible always use temporary email providers. This shit is annoying AF.

🤣

flux is better so far

great. Thanks

DUDE! Learn how to implement a proper CAPTCHA. Dunno why you would go through the trouble of creating this video, to then lock it behind a broken CAPTCHA.

Very cool!

What is your specs you use ComfyUI in

Nice one! I've been saying it for a while. Flux is unbelievable, that's why I use SD!

Nice checkpoint 😁

also not sure how you got those full body poses head to toe type shots from this, mine are always ribs up range.

seems like the only thing missing is the bewb size :) add that in and it's perfect

Perfect. However, I changed two things, as the generated images were blurry. Lora: I used " Hyper Realism - Aidma "; Sampler: euler

The preview video is not shown after running. any advise?

I got through allll the other errors... But when I put 2 images in to combine them into 1 photo, I get this error "Omnigen Node -if you want to make sure the output image have the same size as the input image, please only input one image instead of multiple input images" Any idea why? Thanks!

hello, It works, but how to fix if the text is not clear as the sample?

Ultrarealistic lora is bad with character loras...Dont use

i installed everything like you: Missing Node Types When loading the graph, the following node types were not found PromptTextNode OmniGenNode when i click install missing nodes, there is a import error... i hate comfy ui -.-

that's great,you are amazing~

I ran it, and it showed the following error. Please help me: 'Missing Node Types When loading the graph, the following node types were not found Text to Prompt"

It looks like you are using a lot of extras that is not in comfyui, could you share where to find these ? one minute in and already stuck lol :) Sorry, but this guide was no good.

when i use comfyUI it doesnt allow me to download anything from the manager every time it says failed and bad request etc etc it never lets me download anything nodes models nothing. I can download it normal using cmd and git clone directly to my folder but not from manager

😉😘gracias por compartir

Site doesn't open to be able to download the workflow

So nice! Thank you. But how to make it without generation? Like if I want jus drop already generated image for upscale

good stuff, and nice to hear desi voice behind the tut

This scheme creates a completely different picture. That is, not at all similar to the original, that is - the place is different, the person is different, the clothes are different, the colors are different. This is the worst thing I have seen.

finally worked ,thanks 😍❤

I don’t have the option to use image to video or manager at 1:08 do I need to download something else?

👍👍👍😍

THANKS YOU KURWA Z CAŁEJ PIZDY MISTRZU !!!!

Thank you for the workspace! However, It'd be very helpful if you teach us how to install OneDiff and the required OneFlow. With all due respect, that's the most important part of the tutorial and the hardest part to get going if you're not experienced in this, I've been trying for 3 days and there are practically no tutorials on TH-cam or Reddit explaning how to install OneDiff. Thank you again!

Thank you for sharing. I think it may not work with anime/painting style images. I tried your workflow and it made my painting images look weird and blurry.

terrible audio quality

Many Thanks, my Friend! 💓 New subscriber here. 🙏🏾 Best Wishes

Thats cool! Can you make a image2video tutorial with it?

This video is Great, when i try to generate image not get the best result always confuse about cfg steps.. but after this i start generating high quality image with SD3.5 thank you so much

Thanks for such a great video! You do a great job of talking about simple and more complex concepts. You are a good teacher!

what is i want make longer video than 5 sec?

hi....I did exacly the same bus some photo's it add stuf or make the person older. I fad a photo of my son a 14years an afther the proces it makes him 20. how can istop that? but overall the most are pretty good

can it do consistent characters?

it made my image worse somehow