Nice!!!! But I have noticed that you switched back to Stable Diffusion. Does that mean SD is more better then Flux? Or is it that SD is more reliable then Flux?

Thank you! Flux is better than SD in many ways, BUT... the emphasis on censorship has compromised the training dataset, resulting into being a model great for anything but people. Being a pinup artist, that is a problem to me. Flux can do people, but it has trouble with poses and human anatomy. People have been experimenting with mixing Flux with SD models, trying to get around this, but so far it remains a compromise. In addition, Flux still lacks full support for ControlNet and other important extensions, so I rather wait until all that is worked out. It's not bad - it's still new, incomplete, and experimental, where censorship remains a centerpiece. They should resolve this over time.

SD...Flux.. I miss crayons. Beautiful render, though.

Thank you! Flux, SD and crayons are all tools to a common goal - to express yourself. ^^

Fantastic

Thank you! ^____^=b

The master at work creating images for the pleasure of us viewers...this vampire theme image looks fantastic! I am always amazed by how well your app can transfer the pose so well into Stable Diffusion, the results speak for themselves.

It's a great idea of yours to continue to place the small poser graphic in your image corners to illustrate this point across more clearly. Keep up the great work! :-)

Thank you for the thoughtful comment! 💘

It's a powerful workflow. Let's me do way more than I could before. ^^

Noooooo! Where's my crucifix?? She's waaay too beautiful for this nasty dude!

Really amazing work here, my friend!

Thank you! Dracula's hypnotic gaze is quite effective. ^___^

I'm suprise to hear Flux is censored? When Flux came out people were excited and said that it was uncensored compared to Midjourney. Then again that comparing it to Midjourney, that heavily censored. And are you sure that their no controlnet. I see videos on youtube about flux controlnet and tutorials and how to use it.

https://www.youtube.com/watch?v=HVYXM9bPFTs

As far as I know, I have followed Flux from the very beginning, and it has never claimed to be uncensored. On the contrary - even the training dataset was extremely censored. By now, after several model mixes and modding from the community, people managed to get way less censored versions, but not completely. You can get away with some things, like ground poses, but it's still censored. The way the community has been dealing with this was to merge Flux with SDXL models, but only to a lesser degree, like 70-30. If we do more than that, all the advantages of the Flux model are lost. In a way, that's like saying we can only remove 30% of the built-in censorship before losing the model's quality.

As for ControlNet, you can see in your video link that only 3 of the 10+ modes are available, and OpenPose used in this image is not there. It's too early, and too limited, not to mention I am not a fan of ComfyUI. There is a webUI for Flux, but it has the same limitations. Don't get me wrong, I am not saying Flux us bad - it's just not quite usable to be as it stands now. May take some time until it is.

Ah I see. Thanks for breaking it down for me. now that I think about it it was YouTubers I was saying flux was uncensored. I forgot that YouTubers tend to exaggerate AI tools.

In a way, it makes sense that Flux is this much censored. The company behind it (Black Forest Labs) is made of Stability AI dissidents who formed their own company. This is how they could come up with a model so fast - that is actually a version of SD3, with the same built-in censorship infrastructure as the original. The idea behind censored models is liability - if a child generates something they shouldn't see, that can turn into a lawsuit. Therefore, they have spent considerable resources to censor the model so that won't happen. The problem is that this compromises the model for everybody else, which is not very democratic, but liability takes precedence.