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

Discussion about A.I.

Go to CruxDreams.com
Ulfen said:
I sincerely hope that AI engines out there have filters of some kind. At least if someone would ask the AI engine to make images of children and such. And also if someone would ask to create a bomb and how to poison your boss or school teacher or whatever. etc etc

And sure. If a person look too young I guess it's down to the eye of the beholder of someting like that.
Click to expand...
It's a slippery slope that may lead to a dangerous place, even if we set aside the problem of determining the actual age of a randomly generated character.

Honestly, I'm not interested in NSFW content featuring underage characters at all. But if we're to censor fantasy depictions by applying real-life morality I don't see what's so special about underage content either.

Yes, sexually abusing children is evil, even when depicted as a fantasy art. But what makes things like burning women alive or plunging a wooden stake into their genitalia through the mouth any less so?

I won't try to argue if any of you somehow assert that it's much worse to see a child sexually molested than to see an adult woman burned alive. But can you persuade people outside our little kinky community that we do need censorship on fantasy art, but torturing and murdering innocent women should be exempt from that because that's totally fine by real-life morality?

Besides, if we need a filter to prevent people from generating images of imaginary underage children in the nude, what about other tools like Photoshop or Daz3D? Should Photoshop reject the paste action when a user tries to copy an underage girl's face over a porn image? Or should Daz3D refuse to render when the scene contains underage models without clothing props?

Making and publishing fake photos of real kids in a sexual context is a crime, regardless of the exact method used to produce them. As such, I believe all we need to do is make sure the perpetrator gets caught and pays for his or her crime, not make new restrictions on what fantasy people can create to prevent crimes most of them have no intention to commit.
 
fallenmystic said:
By the way, I just published a custom-trained Lora for depicting slave restraints on Civitai, which I used for rendering the images I posted above.

I'm still new to Stable Diffusion so I'm unsure how well it'd work for others. But if you struggle to depict a collar or shackles for your BDSM renders, you can try it and see if it can help.
Click to expand...
I have installed Easy diffucion to my computer and tested some custom models. However, I have never get it working properly, except with standard stuff. Usually I can’t replicate example images with given parameters and descriptions and get out some random garbage, what is not oven close what I ask. Usually UI don’t regognize some parts of the prompt and highlites failed parts.

I assume this is because I have missed some important details. I haven’t find a simple instructions for complete working setup, which would be good starting point for impatient people like me.

I feel curiosity towards this technology but in it’s current form and capabilities, I don’t think I’d use it much.

I have lot’s of different python stuff in my computer, so it’s possible that it’s just not working as it should. I also have a blender AI render pluging, but that one haven’t been able to put out anything meaningful.

I haven’t use lots of time on this, but in a past year or so I have few times checked if something has changed.
 
fallenmystic said:
It's a slippery slope that may lead to a dangerous place, even if we set aside the problem of determining the actual age of a randomly generated character.

Honestly, I'm not interested in NSFW content featuring underage characters at all. But if we're to censor fantasy depictions by applying real-life morality I don't see what's so special about underage content either.

Yes, sexually abusing children is evil, even when depicted as a fantasy art. But what makes things like burning women alive or plunging a wooden stake into their genitalia through the mouth any less so?

I won't try to argue if any of you somehow assert that it's much worse to see a child sexually molested than to see an adult woman burned alive. But can you persuade people outside our little kinky community that we do need censorship on fantasy art, but torturing and murdering innocent women should be exempt from that because that's totally fine by real-life morality?

Besides, if we need a filter to prevent people from generating images of imaginary underage children in the nude, what about other tools like Photoshop or Daz3D? Should Photoshop reject the paste action when a user tries to copy an underage girl's face over a porn image? Or should Daz3D refuse to render when the scene contains underage models without clothing props?

Making and publishing fake photos of real kids in a sexual context is a crime, regardless of the exact method used to produce them. As such, I believe all we need to do is to make sure the perpetrator gets caught and pays for his or her crime, not making new restrictions on what fantasy people can create to prevent crimes most of them have no intention to commit.
Click to expand...
Sure. A valid point you have there.

Since we are adults in here and understand this to be a thing between adults. BDSM and related fantasy. We would then not look specifically for underage content either because of it. That's why I don't how big of a problem it would be here. At least the way I think of it.
 
Stungiskhan said:
I have installed Easy diffucion to my computer and tested some custom models. However, I have never get it working properly, except with standard stuff. Usually I can’t replicate example images with given parameters and descriptions and get out some random garbage, what is not oven close what I ask. Usually UI don’t regognize some parts of the prompt and highlites failed parts.

I assume this is because I have missed some important details. I haven’t find a simple instructions for complete working setup, which would be good starting point for impatient people like me.

I feel curiosity towards this technology but in it’s current form and capabilities, I don’t think I’d use it much.

I have lot’s of different python stuff in my computer, so it’s possible that it’s just not working as it should. I also have a blender AI render pluging, but that one haven’t been able to put out anything meaningful.

I haven’t use lots of time on this, but in a past year or so I have few times checked if something has changed.
Click to expand...
I can't say much about Easy Diffusion because that's not the program I'm familiar with. But I might be able to help if you could elaborate on the problem you had.

As for the potential Python dependency conflicts, most of such programs use a virtual environment to prevent them, although cleaning up the global site package directory could still be a good idea.

In any case, I'd recommend trying A11 or SD.Next in case you want to try anything serious with AI. Those are by far the most popular choices among AI enthusiasts, and they provide the most number of features unless you want to go with ComfyUI, which is for more advanced users.

As for the Blender plugin, I haven't yet found a usable one either. The last one I tried was Dream Textures but it lacked too many functionalities and was too outdated to be really useful for me. Some of the possibilities they show are certainly intriguing, like texturing the whole scene by prompting alone or using SD as a render engine. But the last time I checked they looked far too immature to have much practical value.

Edit: This is from a test I did with Dream Textures, using it as a render engine within Blender:

1695648026192.png

By the way, I found that Blender + Stable Diffusion can be a great combination even without a proper plugin. I've been using Blender to make a base image for AI render, and it's quite easy to generate all the needed ControlMap images with it automatically:

1695648194321.png
 
Last edited:
fallenmystic said:
It's a slippery slope that may lead to a dangerous place, even if we set aside the problem of determining the actual age of a randomly generated character.

Honestly, I'm not interested in NSFW content featuring underage characters at all. But if we're to censor fantasy depictions by applying real-life morality I don't see what's so special about underage content either.

Yes, sexually abusing children is evil, even when depicted as a fantasy art. But what makes things like burning women alive or plunging a wooden stake into their genitalia through the mouth any less so?

I won't try to argue if any of you somehow assert that it's much worse to see a child sexually molested than to see an adult woman burned alive. But can you persuade people outside our little kinky community that we do need censorship on fantasy art, but torturing and murdering innocent women should be exempt from that because that's totally fine by real-life morality?

Besides, if we need a filter to prevent people from generating images of imaginary underage children in the nude, what about other tools like Photoshop or Daz3D? Should Photoshop reject the paste action when a user tries to copy an underage girl's face over a porn image? Or should Daz3D refuse to render when the scene contains underage models without clothing props?

Making and publishing fake photos of real kids in a sexual context is a crime, regardless of the exact method used to produce them. As such, I believe all we need to do is make sure the perpetrator gets caught and pays for his or her crime, not make new restrictions on what fantasy people can create to prevent crimes most of them have no intention to commit.
Click to expand...
You have many valid points, but I see difference between torture fantasies and child/underage fantasies.

Later one is real problem in our society but there’s much less people who go and impale someone else just for fun.
 
Stungiskhan said:
You have many valid points, but I see difference between torture fantasies and child/underage fantasies.

Later one is real problem in our society but there’s much less people who go and impale someone else just for fun.
Click to expand...
What about rape then? I bet it's at least as widespread as - if not far more so than - sexual abuse of children. Would it be a sufficient reason to censor fantasy depictions of nonconsensual sex?
 
fallenmystic said:
I can't say much about Easy Diffusion because that's not the program I'm familiar with. But I might be able to help if you could elaborate on the problem you had.

As for the potential Python dependency conflicts, most of such programs use a virtual environment to prevent them, although cleaning up the global site package directory could still be a good idea.

In any case, I'd recommend trying A11 or SD.Next in case you want to try anything serious with AI. Those are by far the most popular choices among AI enthusiasts, and they provide the most number of features unless you want to go with ComfyUI, which is for more advanced users.

As for the Blender plugin, I haven't yet found a usable one either. The last one I tried was Dream Textures but it lacked too many functionalities and was too outdated to be really useful for me. Some of the possibilities they show are certainly intriguing, like texturing the whole scene by prompting alone or using SD as a render engine. But the last time I checked they looked far too immature to have much practical value.

Edit: This is from a test I did with Dream Textures, using it as a render engine within Blender:

View attachment 1362154

By the way, I found that Blender + Stable Diffusion can be a great combination even without a proper plugin. I've been using Blender to make a base image for AI render, and it's quite easy to generate all the needed ControlMap images with it automatically:

View attachment 1362155
Click to expand...
Thank’s for the answer. I may ask again when I decide to take another look.

I have most of my python stuff in virtual environments, but there are always some buggy installers, which do something they are not supposed to do and mess things up.
 
fallenmystic said:
What about rape then? I bet it's at least as widespread as - if not far more so than - sexual abuse of children. Would it be a sufficient reason to censor fantasy depictions of nonconsensual sex?
Click to expand...
Rape is a problem, for sure. It’s one of the few red flags for me, when I stop reading a story.

I don’t like most abusive words like cunt, slut, whore and such. I can tolerate mutilation and violence by reading between the lines or changing things in my mind, if the story setup is otherwise interesting.

Rape is also a problem in real life, so that’s why I feel so strongly against it. My fantasies are about imaginary pleasure for a woman and I can’t or don’t want to associate it for a rape.

There’s an edge case in a fantasy, where women herself want it to happen, and then it wouldn’t be rape at all. However that’s very dangerous fantasy because propably most rapists want to think it that way. In real world, it practically never be like that. Some women can have a rape fantasies, but it’s totally different than wanted to be raped.

What come to censorship, that’s really difficult path and hardly never works. Therefore the self censorship and awareness would be much better choice.
 
Well, as far as CruxForums are concerned, our site owner, and our service provider, are quite clear: we want no content that appears to portray under-age subjects. XenForo is a pretty broad-minded company, but they quite understandably don't want trouble, and nor do we. So, yes, we have to make judgements - not on whether a figure is or isn't 'actually' 18+, but does the image give that impression? Is it likely to attract unwanted attention? As our Code of Conduct for Moderators says, we depend on the co-operation of members: we don't want to be patrolling the site arbitrarily 'censoring' what's posted, but we do have a duty to keep the site safe, and hope that members will understand and help us with that by taking care in creating 'models'.
 
A tip for preventing AI from generating inappropriate underage content:

Put “kid” or “child” in the negative prompt. You can also use “woman” instead of “girl” in the positive prompt but it may affect other aspects of the output.

I usually do that for most of my renders. A notable exception was a selection of images I posted in this thread a few pages back. It was because those were test renders and I still haven’t decided on the negative prompt for SDXL. It made the faces of the characters look dangerously young. But I thought it ok since they have mature bodies and those faces aren’t something impossible to have for adult Asian women.

Anyway, it’s still better to stay away from a dangerous territory, so I recommend the method mentioned above to all who use AI to generate NSFW content.
 
fallenmystic said:
A tip for preventing AI from generating inappropriate underage content:

Put “kid” in the negative prompt. You can also use “woman” instead of “girl” in the positive prompt but it may affect other aspects of the output.

I usually do that for most of my renders. A notable exception was a selection of images I posted in this thread a few pages back. It was because those were test renders and I still haven’t decided on the negative prompt for SDXL. It made the faces of the characters look dangerously young. But I thought it ok since they have mature bodies and those faces aren’t something impossible to have for adult Asian women.

Anyway, it’s still better to stay away from a dangerous territory, so I recommend the method mentioned above to all who use AI to generate NSFW content.
Click to expand...
I’m guilty of calling all the women as a girls. I’m trying to get rid of that habit, but it’s deep.

In my own language and with friends, we used to call all the women as girls, even when they are 45, married and have adult children.

In English, girl is also much easier and better sounding word.
 
I haven't read all the posts here but I'm guessing some love AI (me) and many think poorly of it
I have only spent a few hours and am amazed what a powerful resource it is for producing fabulous images
Many of us can't draw artistically at all but stll have artistic sense.
AI offers us the way out so less gifted can express their thoughts beautifully withour going to art school!

Here's two I chopped out today in YODAYO (and many more besides)
Not perfect but superb for the small amount of effort involved


AI Pic Tutorial TEST_18_pic.jpg AI Pic Tutorial TEST_19_pic.jpg

I use YODAYO because it's a pay as you go system so you don't have to sign up to monthly subscription.
(I am afraid that subsciptions are sometimes hard to back out of, so prefer pay as you go)
You can try it for free (and I don't get commission!)
 
stretchedgirl said:
I haven't read all the posts here but I'm guessing some love AI (me) and many think poorly of it
I have only spent a few hours and am amazed what a powerful resource it is for producing fabulous images
Many of us can't draw artistically at all but stll have artistic sense.
AI offers us the way out so less gifted can express their thoughts beautifully withour going to art school!

Here's two I chopped out today in YODAYO (and many more besides)
Not perfect but superb for the small amount of effort involved


View attachment 1363203 View attachment 1363205

I use YODAYO because it's a pay as you go system so you don't have to sign up to monthly subscription.
(I am afraid that subsciptions are sometimes hard to back out of, so prefer pay as you go)
You can try it for free (and I don't get commission!)
Click to expand...
It’s a good addition if it don’t replace other methods, at least before it’s really ready.

In many areas, AI can do things way much faster but not better than traditional methods. In arts, the quality of a single image may not be bad, but so far biggest bottleneck is the lack of variation.

Single model tends to create similar pictures, espicially when trained to make detailed pictures like these.

In commercial world, time is money and money is too important. There’s a risk that transformation will be faster than it should.
 
Stungiskhan said:
In arts, the quality of a single image may not be bad, but so far biggest bottleneck is the lack of variation.
Click to expand...
Personally, I don't feel AI limits the variety of what I can generate with it much. We have so many base models and external networks freely available now, which can be used to express different styles.

If AI renders felt lacking in variety, I suspect it could be because most of them are created by people who have a casual interest in AI using one of the free websites which only provides a few select models and minimal functionalities. To see what more dedicated AI users generate with the tool, I'd suggest visiting Civitai and browsing its posts section.

I think one of the most significant issues Stable Diffusion has now is the difficulty of making complex scenes. To be specific, it's not easy to make a scene involving many different characters even with tools like ControlNet or Regional Prompting extension. While these tools made it possible to compose a complex scene as intended, sacrifice in quality is often unavoidable and it still takes a disproportionate amount of effort to fix problems (e.g. malformed toes) in such renders. There's a reason why a vast majority of those posts in Civitai involve only a single character, despite the variety in style they show.

But considering it's been barely a year since the first public release of Stable Diffusion, I'm still optimistic that those problems will get resolved soon enough.
 
A brief update/tip on AI roleplaying setup:

I managed to load a decent model (Mythalion-13b) on my local PC (RTX3080 / 10Gb VRAM) with the following settings:

1696039851454.png

Not sure if these are optimal settings, or if there's a better way to do it. But it's the first working setup I found with which I can run the popular Mythalion 13b model locally with decent performance. It means I don't have to burn my Runpod credits for RP any more.

So, if you have a better graphics card than RTX3080, chances are high that you can also play decent roleplaying sessions (both SFW and NSFW) with AI partners for free.

For detailed hardware requirements, you can see this post.

As I'm much less familiar with text generation AIs than with Stable Diffusion, meaning and proper usage of the model parameters and presets have remained a mystery to me. But I just stumbled upon this page, which offers detailed explanations of the topic with examples.


P.S. : I noticed how more people started using Stable Diffusion to generate great images on CF recently, and the general atmosphere regarding AI became much less hostile than a few months ago. However, I haven't seen much regarding AI-assisted roleplaying or story-writing so far.

It feels strange considering how many people in this community are interested in RP in general. I assume it could be because of the aversion to the idea of playing with a bot (which I understand) or because of the technical hurdles involved. If the latter is the case, please let me know so I can help you make a similar setup as mine.

I'm having consistently great RP experiences with Mythalion 13B, so I'd be happy to help others so that they can enjoy it too.
 
fallenmystic said:
As I'm much less familiar with text generation AIs than with Stable Diffusion, meaning and proper usage of the model parameters and presets have remained a mystery to me. But I just stumbled upon this page, which offers detailed explanations of the topic with examples.
Click to expand...
I just realised I forgot to add the actual link. Since nobody complained about it, I suppose it shows almost no one is interested in the subject yet. :D

Anyway, here's the link to the website mentioned in the previous post: The Local LLM Settings Guide/Rant
 
You must log in or register to reply here.
Back
Top Bottom