• Sign up or login, and you'll have full access to opportunities of forum.

Discussion about A.I.

Go to CruxDreams.com
Heads up for anyone interested in AI...

My 3D art board, Visions of Darkness, is going to allow AI art starting April 3rd.

I strongly dismissed AI after first learning about it last March. Since then, I've learned how it works and that there are valid arguments for and against its use. The ethical debate about its development is what I've struggled with most. Artists I've talked to who are experimenting with it have told me that's a separate issue for them. As fellow hobbyists, they just want to have fun and test its limits, whether enhancing their original work or creating new images. I must adapt if I want VoD to stay relevant and be successful.

Here's a a screenshot of our upcoming new gallery layout taken from VoD's test board:

new-galleries.png

https://visionsofdarkness.net - Original Dark Art & Fiction
 
Last edited:
Heads up for anyone interested in AI...

My 3D art board, Visions of Darkness, is going to allow AI art starting April 3rd.

I strongly dismissed AI after first learning about it last March. Since then, I've learned how it works and that there are valid arguments for and against its use. The ethical debate about its development is what I've struggled with most. Artists I've talked to who are experimenting with it have told me that's a separate issue for them. As fellow hobbyists, they just want to have fun and test its limits, whether enhancing their original work or creating new images. I must adapt if I want VoD to stay relevant and be successful.

Here's a a screenshot of our upcoming new gallery layout taken from VoD's test board:

View attachment 1447394

https://visionsofdarkness.net - Original Dark Art & Fiction


Wise move
 
Maybe you've already seen some of these colourised / upscaled films using AI, but probably not with such quality:


As AI technology gets better every day, I'm glad we now have a clearer window through which we can peek at everyday life in past centuries. Personally, I can't wait to see similar enhanced videos restored from WW2-era films.
 
I want to start April with the new gallery structure already in place, so I've moved the launch date up to March 31st.

I also want viewers to know whether they're looking at generated or enhanced AI art, so I'm going to create two forums for AI art. Since that will result in a total of 8 forums between the Main Hall and Back Room, I'll be turning both galleries into a dual column display to minimize vertical scrolling on the forum index.

Comically, doing that on VoD's test board triggered my OCD, so now every category except the Writers Nook and Guest Lounge will have dual columns.

Finally, the current name for the 3D art gallery is Digital Art. Per my staff's suggestion, I'm leaving the name alone to avoid leaving artists with the impression that we only welcome Daz and Poser art.

Here's a preview of the new layout for the Member Lobby and Galleries.
new-galleries.png
 
...
As AI technology gets better every day, I'm glad we now have a clearer window through which we can peek at everyday life in past centuries. Personally, I can't wait to see similar enhanced videos restored from WW2-era films.
In line with such a case where AI technology is helping us obtain a more vivid glance into the past, here's another video that recently popped up on my YouTube list. It's a Hitler's speech reconstructed in English:


The channel also contains various other reconstructed audio/video clips, including the famous Goebbel's "Total War" speech.

I don't like how they vilify Hitler - sometimes almost to a caricatural level - when they depict him in films or other media. It's not that I think he was a good person but because I believe it will only give people a false impression that only evil people supported such an obvious villain or that the people were more stupid back then as to be fooled by him.

But I believe they should teach how evil and cruelty of such a scale can be born out of what many people would regard as "patriotism" or even "common sense". It should show how easily they can be persuaded to believe such things in the right circumstances with a bit of help from prejudices and propaganda.

In that regard, I think videos can even be a good educational tool, showing how much of his speech still feels relevant and persuasive today. This means that the danger was never quelled but just lay dormant, waiting for the moment when things like prejudice against minority groups, political enmity, economic crisis, etc., all fall into place once again, gathering around a catalyst like a highly charismatic political leader, for example.
 
There has been a bit of drama going on concerning Stability, the company behind Stable Diffusion.

I won’t delve into details because few would be interested in. But it might become a very bad news for those use SD for creating kinky art, so I will briefly elaborate on it.

As I occasionally mentioned here, AI is becoming a battlefield for a renewed war between the open-source camp and those who oppose it. Ideology aside, what concerns us who either create or enjoy AI-generated NSFW contents is that all closed-source AI services are heavily censored and thus, not suitable for making such images.

And when it comes to image-generating AI, Stable Diffusion has been the only open-source model that started a growing community which has been contributing all sorts of NSFW finetunes and LORAs based on it.

Recently, Stability has been promoting the next version of their model, Stable Diffusion 3, which is believed to have drastically improved many limitations of Stable Diffusion XL. However, they announced today the resignation of their CEO, Emad Mostaque who has been a proponent of the open-source ideals.

Allegedly, the main reason for his resignation was because the investors weren’t happy about the company’s financial situation and they wanted to monetise more of their models. It means now we don’t know when, or if Stable Diffusion 3 will ever be released as an open model like its predecessors. The worst case scenario would be they decide to release it as a commercial product under a closed-source license which doesn’t allow generating NSFW contents or community finetunes.

We’ll likely see how it will unfold soon.
 
There has been a bit of drama going on concerning Stability, the company behind Stable Diffusion.

I won’t delve into details because few would be interested in. But it might become a very bad news for those use SD for creating kinky art, so I will briefly elaborate on it.

As I occasionally mentioned here, AI is becoming a battlefield for a renewed war between the open-source camp and those who oppose it. Ideology aside, what concerns us who either create or enjoy AI-generated NSFW contents is that all closed-source AI services are heavily censored and thus, not suitable for making such images.

And when it comes to image-generating AI, Stable Diffusion has been the only open-source model that started a growing community which has been contributing all sorts of NSFW finetunes and LORAs based on it.

Recently, Stability has been promoting the next version of their model, Stable Diffusion 3, which is believed to have drastically improved many limitations of Stable Diffusion XL. However, they announced today the resignation of their CEO, Emad Mostaque who has been a proponent of the open-source ideals.

Allegedly, the main reason for his resignation was because the investors weren’t happy about the company’s financial situation and they wanted to monetise more of their models. It means now we don’t know when, or if Stable Diffusion 3 will ever be released as an open model like its predecessors. The worst case scenario would be they decide to release it as a commercial product under a closed-source license which doesn’t allow generating NSFW contents or community finetunes.

We’ll likely see how it will unfold soon.
This is why we will need to be running it locally not inline - then we can do whatever the fuck we want with it. The source code is already out there so even if they do decide to make it closed at some point, the genie is already out of the bottle. It won't be the first open source software to be closed off by new owners, but the open version will always be out there to be forked off into something else that will fill the gap. Whether we like it or not, AI technology is too important to allow it to be controlled by a single entity but the fact that SD is open source means that it will always be around in some form or other and as computing hardware gets ever more powerful, it will just become more accessible.
 
This is why we will need to be running it locally not inline - then we can do whatever the fuck we want with it. The source code is already out there so even if they do decide to make it closed at some point, the genie is already out of the bottle. It won't be the first open source software to be closed off by new owners, but the open version will always be out there to be forked off into something else that will fill the gap. Whether we like it or not, AI technology is too important to allow it to be controlled by a single entity but the fact that SD is open source means that it will always be around in some form or other and as computing hardware gets ever more powerful, it will just become more accessible.
I wholeheartedly agree about the importance of locally runnable open-source models. However, there is a reason why the war is about to rekindle after the remarkable victory of the open-source camp in past decades.

What you said is true of traditional software but there’s a peculiarity with AI models that makes things complicated. Long story short, you are highly unlikely to be able to reproduce what the original author did with an AI model without having a gigantic computing power and an access to additional information like training methods used.

That’s why there’s a considerable danger of the immense power of AI technology may end up being monopolised by a handful of companies which can amass sufficient human and computing resources to stay competitive.

So, unfortunately, if Stability decides not to release Stable Diffusion 3 as an open source model, it’s likely that the kinky art community will get stuck with Stable Diffusion XL for foreseeable future.
 
So, unfortunately, if Stability decides not to release Stable Diffusion 3 as an open source model, it’s likely that the kinky art community will get stuck with Stable Diffusion XL for foreseeable future.
The hackers will always find a way. They always have, and they always will
 
Good news: the CTO and interim co-CEO of Stability just confirmed that they'll release Stable Diffusion 3 as an open-source model as originally planned. :)


It looks like Stability may stay on the open-source side in this war despite the leadership change. Meanwhile, the CEO of OpenAI reportedly has been lobbying in Washington, urging the government to regulate open-source AI models.

At any rate, I think there's a decent chance we'll start to see much better-quality AI art before the end of this year.
 
For those who don't know why a kink art community might be excited about Stable Diffusion 3:

You may have noticed that generating images with a nontrivial composition in Stable Diffusion requires a lot of effort. We've all seen AI images of characters with missing limbs, mangled faces, and other similar deformities. Also, there are much fewer images with a complex composition, like one would be needed to depict an auction or torture scene.

This is because the current version of Stable Diffusion requires extensive experience with various tools to fix such deformities or lay out a scene exactly as you imagined.

Of course, some artistic tools require a lot of skill to learn and use properly. But if one has to spend much time learning how to create art, I think it's far better to learn different ways to express one's creativity than to learn workarounds and tools to circumvent the technical limitations of the chosen medium.

In short, Stable Diffusion 3 will likely allow anyone to create decent kinky images (after the community catches up with NSFW models) just by describing what they want in plain language. While lowering the hurdle, however, it will still provide greater control for more dedicated AI creators.

For example, when you give it this prompt:
a realistic anthropomorphic hedgehog in a painted gold robe, standing over a bubbling cauldron, an alchemical circle, steam and haze flowing from the cauldron to the floor, glow from the cauldron, electrical discharges on the floor, Gothic

It will generate something like this. See how closely it can follow the instruction, which is something SDXL is notably incapable of:

스크린샷_20240326_093250.jpeg

The above prompt and image came from a Reddit thread linked below. I recommend taking a look at the comment section as there are many more interesting examples that showcase SD 3's capabilities:

 
Yesterday, OpenAI released AI videos created by artists and filmmakers using their flagship video-generating model, Sora.

This is one of the examples, a short film called "Air Head":


Can't wait to see what we'd be able to create when an uncensored open-source alternative will eventually be made available in future.
 
Back
Top Bottom