AI is getting real crazy, y'all

[Vigilante8814]

That's pretty cool. However, elon musk's cimpany can burn to the ground before i will ever patronize it and send him any amount of money. I'll train my own models. I'm also never calling twitter that because musk has no right to resemble the hands down best iteration of Mega Man. If somebody wants to steal and release the code, though.... just um... just saying.

Dude... you can use Grok for free. I don't even have an X account myself, I didn't even know that AI program is affiliated. I got it through the Googke Play Store.

[read the comics]

That's pretty cool. However, elon musk's cimpany can burn to the ground before i will ever patronize it and send him any amount of money. I'll train my own models. I'm also never calling twitter that because musk has no right to resemble the hands down best iteration of Mega Man. If somebody wants to steal and release the code, though.... just um... just saying.

Dude... you can use Grok for free. I don't even have an X account myself, I didn't even know that AI program is affiliated. I got it through the Googke Play Store.

Dawg, the developer is literally them. (Still not fucking calling it that.) Letting them run ads on my phone still earns them money. I would like few things more than the free market to strangle the life out of elon musk's entire entrepreneurial universe, and anyone else who thinks election interference and amplifying literal nazi ideology online are swell passtimes for the wealthy. Plenty of people are working on good models. I'm not giving that piece of literal human filth a penny whether it's from my own pocket or otherwise.

[Totallibertarianism]

For the AI arts technologies, as I said it in the DD forum, not only the fetish producers will be hit hard but also the biggest audiovisual, cinematographic industries, multinationals such as Hollywood and others multimedia outlets. And because of this, it is all about making money and profits especially in the western countries, so the powerfull economic "lobbies" and also the political ones in the US and in the UE will not let this to happen.

There are already many laws and regulations protocols that are voted in the US and in the UE, the recent one by the new US administration "Take it Down" act... So in the coming years, there are high risk that the AI technologies will be completely censored or available only in high pay mode. So in this case, the traditional audiovisual producers will be able to continue their works, but maybe in different way, for example mixing the AI with the regular materials to create a very realistic movies and reducing slightly its price of selling.

I already used VEO 3 ( https://www.femmefatalities.com/phpBB3/ ... hp?t=24264 ), and it is effectively not very interesting for my point of view, it is already harshly censored and in pay mode... I prefer using my 100 % free and uncensored locally installed AI software, the results are better for a death fetish fan like me.

And by the way, this topic shouldn't be in the AI section of the forum, and not in the regular videos releasing section ?

[read the comics]

They're not stopping this. I will be able to train my own private models by the end of the decade. It's not even farfetched to imagine that anyone might be able to have a conversation with a model, with zero understanding of the way this tech works, and that model would train whatever they want without even much work to clean whatever data set they use, if they even still need one. I can't imagine they wouldn't but veo3 gets very close to what these perils actually look like in official content. Chatgpt has demonstrated a remarkable ability to predict the sorts of connections that excite me the most and it learned most of that from reading online fan fictions. And I'm not talking about the wider trips of the genre. I mean that within a few exchanges of me describing the content i aim to create, it was predicting details specific to the way that I, personally, enjoy the genre. At any rate, it took me a few hours and some private messaging on deviant art to train the LoRAs that I'm running right now and this is still very early stages of viability beginning to appear. With 55 images, one of my models turned out overfit. It's laughable to suggest they're going to have any more control over this tech than they do torrents. Once i get a good model with the data i have, i can basically train a second with data that's produced with the first and meticulously cleaned. I would need some new data to feed the third generation in order to prevent uncanny drift, but I have time to figure that out.

[otischloro]

AI is all hype. I have played with some of these AI video and image generation technologies and I don't see it working "right" anytime soon. Good luck trying to train a video or image generation technology to perform any kind of action correctly. You see, even the smallest oddity in these videos kills the vibe (try to enjoy an scene with a character carrying another that suddenly grows a third leg, second torso, etc.). So far absolutely nothing has come even close to give me something satisfying, at least in terms of OTS carries, and I have REALLY tried. It's pretty much impossible to create consistent characters to make more than 2 scenes or images. Maybe in something like 10 years we might get something good, but anytime soon? No way in hell.

[read the comics]

AI is all hype. I have played with some of these AI video and image generation technologies and I don't see it working "right" anytime soon. Good luck trying to train a video or image generation technology to perform any kind of action correctly. You see, even the smallest oddity in these videos kills the vibe (try to enjoy an scene with a character carrying another that suddenly grows a third leg, second torso, etc.). So far absolutely nothing has come even close to give me something satisfying, at least in terms of OTS carries, and I have REALLY tried. It's pretty much impossible to create consistent characters to make more than 2 scenes or images. Maybe in something like 10 years we might get something good, but anytime soon? No way in hell.

Have you even seen what's happening recently? It has popped off in the last month. We're inside of two years for working creators and those with patience, 5 for everyone else.

[Totallibertarianism]

AI is all hype. I have played with some of these AI video and image generation technologies and I don't see it working "right" anytime soon. Good luck trying to train a video or image generation technology to perform any kind of action correctly. You see, even the smallest oddity in these videos kills the vibe (try to enjoy an scene with a character carrying another that suddenly grows a third leg, second torso, etc.). So far absolutely nothing has come even close to give me something satisfying, at least in terms of OTS carries, and I have REALLY tried. It's pretty much impossible to create consistent characters to make more than 2 scenes or images. Maybe in something like 10 years we might get something good, but anytime soon? No way in hell.

Yes, for now you are quite right about the AI arts technologies for generating strangling, wrestling and 1 VS 1 fight videos but I have found the technique to do so and I talked about it months ago, to do it you have to mixing video games pictures and videos with your AI software, especially using the img/img and the inpaint mode. In this way the results will be very oustanding !

This is my recent works with my AI software of generating F VS F fights/wrestling/strangulations techniques =>

2 (1).png (1.12 MiB) Viewed 686 times

2 (2).png (1.11 MiB) Viewed 686 times

2 (3).png (1.18 MiB) Viewed 686 times

Now I can generate any F VS F fights scenes with my AI software, and I recommend using an uncensored one to generate this.

Of course, the AI is continuing to improve and maybe in the coming years or maybe months, you will be able to do it in straight way, in one step !

[read the comics]

AI is all hype. I have played with some of these AI video and image generation technologies and I don't see it working "right" anytime soon. Good luck trying to train a video or image generation technology to perform any kind of action correctly. You see, even the smallest oddity in these videos kills the vibe (try to enjoy an scene with a character carrying another that suddenly grows a third leg, second torso, etc.). So far absolutely nothing has come even close to give me something satisfying, at least in terms of OTS carries, and I have REALLY tried. It's pretty much impossible to create consistent characters to make more than 2 scenes or images. Maybe in something like 10 years we might get something good, but anytime soon? No way in hell.

Yes, for now you are quite right about the AI arts technologies for generating strangling, wrestling and 1 VS 1 fight videos but I have found the technique to do so and I talked about it months ago, to do it you have to mixing video games pictures and videos with your AI software, especially using the img/img and the inpaint mode. In this way the results will be very oustanding !

This is my recent works with my AI software of generating F VS F fights/wrestling/strangulations techniques =>


2 (1).png

2 (2).png

2 (3).png










Now I can generate any F VS F fights scenes with my AI software, and I recommend using an uncensored one to generate this.

Of course, the AI is continuing to improve and maybe in the coming years or maybe months, you will be able to do it in straight way, in one step !

Have you tried training a LoRA? That's how I'm getting closer to the imagery i want.

What are you using to animate? I haven't found anything that can do anything that complex yet.

[Totallibertarianism]

Have you tried training a LoRA? That's how I'm getting closer to the imagery i want.

What are you using to animate? I haven't found anything that can do anything that complex yet.

I stopped using Loras files, I used them to generate celebrities in pictures and videos, but now there is no need to use Loras files to get your best wanted model, there is an extension called Face Swap Lab that allows you to generate any existing or non-existing women in this world, with this extension you need just the face picture of any person even your sister or yourself or a random non existing face picture to generate her or him in the AI arts, in pictures and in videos.

=> https://glucauze.github.io/sd-webui-faceswaplab/
=>

This the version for the ComfyUI (ReActor) => https://github.com/Gourieff/ComfyUI-ReActor

But as you know about it, the new law "Take it down act" in the US will prevent you to upload pictures and videos of any existing person in the AI arts in fetish/porn way, if you do so, you could be penally prosecuted.

Of course, me too I have stopped to upload pictures and videos of existing women, but only I do with non existing random generated women.
=> https://www.femmefatalities.com/phpBB3/ ... 55#p188555
=> https://www.femmefatalities.com/phpBB3/ ... hp?t=24115

I still generate celebrities and existing women with the Face Swap Lab but keep them only in my hard drive for personal collections. I generated for example beautiful models of the bluestone studio, like Jenny Cherry or Lynda or famous french females reporters, the results were quite good... But of course I will never upload them !

[read the comics]

Have you tried training a LoRA? That's how I'm getting closer to the imagery i want.

What are you using to animate? I haven't found anything that can do anything that complex yet.

I stopped using Loras files, I used them to generate celebrities in pictures and videos, but now there is no need to use Loras files to get your best wanted model, there is an extension called Face Swap Lab that allows you to generate any existing or non-existing women in this world, with this extension you need just the face picture of any person even your sister or yourself or a random non existing face picture to generate her or him in the AI arts, in pictures and in videos.

=> https://glucauze.github.io/sd-webui-faceswaplab/
=>

This the version for the ComfyUI (ReActor) => https://github.com/Gourieff/ComfyUI-ReActor

But as you know about it, the new law "Take it down act" in the US will prevent you to upload pictures and videos of any existing person in the AI arts in fetish/porn way, if you do so, you could be penally prosecuted.

Of course, me too I have stopped to upload pictures and videos of existing women, but only I do with non existing random generated women.
=> https://www.femmefatalities.com/phpBB3/ ... 55#p188555
=> https://www.femmefatalities.com/phpBB3/ ... hp?t=24115

I still generate celebrities and existing women with the Face Swap Lab but keep them only in my hard drive for personal collections. I generated for example beautiful models of the bluestone studio, like Jenny Cherry or Lynda or famous french females reporters, the results were quite good... But of course I will never upload them !

I don't train my LoRAs on people, i train them on concepts so they can do the facial expressions and poses that i want without a ton of coaxing. I don't need a law to know that generating imagery of real people in sexual content is unethical.

At any rate, thanks for sharing the link, I'm gonna try it out. I've been using vidu to animate stills that i generate with my LoRAs and I've gotten promising results. Photorealistic is much more difficult than just on the upper side of the uncanny valley for my work flow so far. It's almost like pulling the realism back makes the flaws stand out less.

I could honestly get a voice actor on fiver to affect this and I would be able to ignore the flaws for a couple rounds at a time here and there on a whim. Its biggest issue for me is that it's too short to enjoy for much longer than a minute or two at a time without burning itself out.

But more importantly, if i can feed this to models in the future that are smarter and they correct the flaws and extrapolate a full scene, that would be fucking dope.

[Director B]

Interesting.

[read the comics]

Interesting.

Interesting indeed. And very accurate. I've had to learn a lot about these things to get anywhere at all with them because it really is like communicating ideas to an alien mind. You never know how it's actually going to interpret your instructions.

For me, the best results so far have been from training LoRAs on specific peril elements so that i have an actual language to communicate to the model with and then i can pair those with character model LoRAs to generate stills depicting those elements. Then i can take those stills to vidu, tell vidu what is happening in 5 second increments and with a little luck, i get rough content. Rough by today's standards, though. The 18 second clip i stitched together today would've blown me away in 2003 (didn't know about uksg back then, agent blonde would've blown me away more in 2003) and that's the biggest thing that i keep thinking about. It's flawed for different reasons but it's arguably better than probably most of the superheroine content that was floating around in the early days of the internet when Danielle Trixie was being chloroform and groped simultaneously by a man who doesn't know how to wear a ski mask and definitely listens to Rush Limbaugh on his way to a job that he sucks at yet somehow still has.

[Director B]

Have you tried training a LoRA? That's how I'm getting closer to the imagery i want.

What are you using to animate? I haven't found anything that can do anything that complex yet.

I stopped using Loras files, I used them to generate celebrities in pictures and videos, but now there is no need to use Loras files to get your best wanted model, there is an extension called Face Swap Lab that allows you to generate any existing or non-existing women in this world, with this extension you need just the face picture of any person even your sister or yourself or a random non existing face picture to generate her or him in the AI arts, in pictures and in videos.

=> https://glucauze.github.io/sd-webui-faceswaplab/
=>

This the version for the ComfyUI (ReActor) => https://github.com/Gourieff/ComfyUI-ReActor

But as you know about it, the new law "Take it down act" in the US will prevent you to upload pictures and videos of any existing person in the AI arts in fetish/porn way, if you do so, you could be penally prosecuted.

Of course, me too I have stopped to upload pictures and videos of existing women, but only I do with non existing random generated women.
=> https://www.femmefatalities.com/phpBB3/ ... 55#p188555
=> https://www.femmefatalities.com/phpBB3/ ... hp?t=24115

I still generate celebrities and existing women with the Face Swap Lab but keep them only in my hard drive for personal collections. I generated for example beautiful models of the bluestone studio, like Jenny Cherry or Lynda or famous french females reporters, the results were quite good... But of course I will never upload them !

I don't train my LoRAs on people, i train them on concepts so they can do the facial expressions and poses that i want without a ton of coaxing. I don't need a law to know that generating imagery of real people in sexual content is unethical.

At any rate, thanks for sharing the link, I'm gonna try it out. I've been using vidu to animate stills that i generate with my LoRAs and I've gotten promising results. Photorealistic is much more difficult than just on the upper side of the uncanny valley for my work flow so far. It's almost like pulling the realism back makes the flaws stand out less.

I could honestly get a voice actor on fiver to affect this and I would be able to ignore the flaws for a couple rounds at a time here and there on a whim. Its biggest issue for me is that it's too short to enjoy for much longer than a minute or two at a time without burning itself out.

But more importantly, if i can feed this to models in the future that are smarter and they correct the flaws and extrapolate a full scene, that would be fucking dope.

I'm sorry, but that video looks very cartoonish; it doesn't work. But I'm sure in the future it'll be better. There will be gatekeepers on this stuff, and it's only a matter of time before they start policing them. Yes, you may, for now, do things like this, but as it gets more life-like, you can bet the authorities will be clamping down on questionable prompts. Then you'll have to go underground to create your videos. It may end up a crime to create what you really want to see and make. Then again, I could be wrong.

[read the comics]

I stopped using Loras files, I used them to generate celebrities in pictures and videos, but now there is no need to use Loras files to get your best wanted model, there is an extension called Face Swap Lab that allows you to generate any existing or non-existing women in this world, with this extension you need just the face picture of any person even your sister or yourself or a random non existing face picture to generate her or him in the AI arts, in pictures and in videos.

=> https://glucauze.github.io/sd-webui-faceswaplab/
=>

This the version for the ComfyUI (ReActor) => https://github.com/Gourieff/ComfyUI-ReActor

But as you know about it, the new law "Take it down act" in the US will prevent you to upload pictures and videos of any existing person in the AI arts in fetish/porn way, if you do so, you could be penally prosecuted.

Of course, me too I have stopped to upload pictures and videos of existing women, but only I do with non existing random generated women.
=> https://www.femmefatalities.com/phpBB3/ ... 55#p188555
=> https://www.femmefatalities.com/phpBB3/ ... hp?t=24115

I still generate celebrities and existing women with the Face Swap Lab but keep them only in my hard drive for personal collections. I generated for example beautiful models of the bluestone studio, like Jenny Cherry or Lynda or famous french females reporters, the results were quite good... But of course I will never upload them !

I don't train my LoRAs on people, i train them on concepts so they can do the facial expressions and poses that i want without a ton of coaxing. I don't need a law to know that generating imagery of real people in sexual content is unethical.

At any rate, thanks for sharing the link, I'm gonna try it out. I've been using vidu to animate stills that i generate with my LoRAs and I've gotten promising results. Photorealistic is much more difficult than just on the upper side of the uncanny valley for my work flow so far. It's almost like pulling the realism back makes the flaws stand out less.

I could honestly get a voice actor on fiver to affect this and I would be able to ignore the flaws for a couple rounds at a time here and there on a whim. Its biggest issue for me is that it's too short to enjoy for much longer than a minute or two at a time without burning itself out.

But more importantly, if i can feed this to models in the future that are smarter and they correct the flaws and extrapolate a full scene, that would be fucking dope.

I'm sorry, but that video looks very cartoonish; it doesn't work. But I'm sure in the future it'll be better. There will be gatekeepers on this stuff, and it's only a matter of time before they start policing them. Yes, you may, for now, do things like this, but as it gets more life-like, you can bet the authorities will be clamping down on questionable prompts. Then you'll have to go underground to create your videos. It may end up a crime to create what you really want to see and make. Then again, I could be wrong.

Did that make you feel better about the radiation flooding the room thing? Yes, I'm aware that it looks cartoonish. That's why i said ALMOST. I never said "this is a good clip." Details are important in conversations, particularly words like "almost." When you observe words like "almost," you don't correct people who are already on the page you're trying to direct them to and they don't get annoyed by it.

Those of us who are primarily into superheroines grew up getting excited by cartoons, so that's not going to be a dealbreaker for me when the models get better with motion and continuity. Sorry, you'll have to wait longer than me, i guess?

It's a work in progress, but thanks for informing me that my work in progress is a work in progress. Not sure what I'd do without you.

I've also said multiple times that I'm training my own models so I'm already underground.

[read the comics]

I stopped using Loras files, I used them to generate celebrities in pictures and videos, but now there is no need to use Loras files to get your best wanted model, there is an extension called Face Swap Lab that allows you to generate any existing or non-existing women in this world, with this extension you need just the face picture of any person even your sister or yourself or a random non existing face picture to generate her or him in the AI arts, in pictures and in videos.

=> https://glucauze.github.io/sd-webui-faceswaplab/
=>

This the version for the ComfyUI (ReActor) => https://github.com/Gourieff/ComfyUI-ReActor

But as you know about it, the new law "Take it down act" in the US will prevent you to upload pictures and videos of any existing person in the AI arts in fetish/porn way, if you do so, you could be penally prosecuted.

Of course, me too I have stopped to upload pictures and videos of existing women, but only I do with non existing random generated women.
=> https://www.femmefatalities.com/phpBB3/ ... 55#p188555
=> https://www.femmefatalities.com/phpBB3/ ... hp?t=24115

I still generate celebrities and existing women with the Face Swap Lab but keep them only in my hard drive for personal collections. I generated for example beautiful models of the bluestone studio, like Jenny Cherry or Lynda or famous french females reporters, the results were quite good... But of course I will never upload them !

I don't train my LoRAs on people, i train them on concepts so they can do the facial expressions and poses that i want without a ton of coaxing. I don't need a law to know that generating imagery of real people in sexual content is unethical.

At any rate, thanks for sharing the link, I'm gonna try it out. I've been using vidu to animate stills that i generate with my LoRAs and I've gotten promising results. Photorealistic is much more difficult than just on the upper side of the uncanny valley for my work flow so far. It's almost like pulling the realism back makes the flaws stand out less.

I could honestly get a voice actor on fiver to affect this and I would be able to ignore the flaws for a couple rounds at a time here and there on a whim. Its biggest issue for me is that it's too short to enjoy for much longer than a minute or two at a time without burning itself out.

But more importantly, if i can feed this to models in the future that are smarter and they correct the flaws and extrapolate a full scene, that would be fucking dope.

I'm sorry, but that video looks very cartoonish; it doesn't work. But I'm sure in the future it'll be better. There will be gatekeepers on this stuff, and it's only a matter of time before they start policing them. Yes, you may, for now, do things like this, but as it gets more life-like, you can bet the authorities will be clamping down on questionable prompts. Then you'll have to go underground to create your videos. It may end up a crime to create what you really want to see and make. Then again, I could be wrong.

I'm sorry if i made you feel threatened with my enthusiasm for what the tech will be capable of soon, but digging up videos about how difficult it is to work with AI and insulting me for being on step 3 of 1000 isn't going to prevent what's coming. You will have to adapt the same way everyone does when a new technology shakes up their industry and the more grace you use throughout that process, the better your reputation will be for it when it's all over. But if i see you getting nervous now over a "cartoonish 18 second clip that doesn't work" then I'm not actually sure how much faith i have in your ability to navigate those waters.