AI???

[cisco47]

I have been dabbling with AI over the past few months. I am still a novice but I find it fun to create fantasy stories and images.

I submitted a couple on this website here:

viewtopic.php?f=16&t=1014

and here:

viewtopic.php?f=16&t=1146

If folks find some entertainment in these, I will upload another AI story.

[dragontwo2]

I gave it a try

I also gave it a try dragontwo2, with some success. However, the restraints and details in my images are not as good as in yours. So your prompts much be of a superior quality. Can you post your prompt for these so that I can compare?
I will, of course, understand if you consider them intellectual property which you cannot share.

public whipping 5.jpg

Hi, williej
I send private message to you.
Please check it!

[dragontwo2]

new

[dragontwo2]

more

[GraekrasBlight]

I was using getimg.ai and getting pretty decent results, which you can check out here: https://www.deviantart.com/graekrasblight/gallery
My plan is to learn comfy ui which is a node based software to generate AI images locally. I loaded it on my computer, but it quickly crashed it so I need to upgrade my hardware first.
getimg.ai is restricted, but it is inconsistent - I generated fully nude scenes for several months before I got a policy-violation pop up. Even then, sometimes I would re-run the render with no problem. And I learned to game the algo. Topless imagery rarely triggered the warning, but full nudity would. This all worked until recently when after a string of warning pop-ups I was officially banned. I could easily re-register with a different email, but I'm using it as an excuse to take a break.
I've been impressed with the results that people are getting with text-to-image AI, but I prefer doing image-to-image AI, and I build the initial input using DAZ + Blender which is an entirely different workflow.

[bakerboy]

The last set is nice, but it does NOT look as a "live" shot. Also, pictures in AI are many a most are great, but video is still in its infancy.... I am waiting

In the last post ...Who is censoring/banning AI ???" I thought once you had the software and knew how to use it, you can do what you want?"

[GraekrasBlight]

The last set is nice, but it does NOT look as a "live" shot. Also, pictures in AI are many a most are great, but video is still in its infancy.... I am waiting

In the last post ...Who is censoring/banning AI ???" I thought once you had the software and knew how to use it, you can do what you want?"

Most web based AI tools are restricted. Software to generate AI imagery locally like Comfy UI requires a fast CPU + GPU plus a lot of RAM and SSD. Unrestricted locally generated AI is the dream. I use a gaming laptop with a good CPU/GPU, but I don't have the RAM + SSD, so until I upgrade I am using online tools only. I installed ComfyUI and it crashed my computer after two images. Curious from anyone reading this thread who is using a software like ComfyUI what hardware they are running. For web based, Midjourney and ChatGPT are probably the most well known, but they are highly restricted. Ideogram is really good, but also very restricted. I've tried a handful of online tools that are not restricted, but I haven't found any that I've liked. getimg.ai is the one tool that offers a lot of control over parameters and I can kind of push the envelope with it - bondage is okay, topless is okay, but nudity gets flagged.

[Kelderek]

Curious from anyone reading this thread who is using a software like ComfyUI what hardware they are running.

I run Flux.1 Dev on a local ComfyUI server with no problem, but I have some serious hardware for it:

CPU: Intel i9-13900K
64 Gb RAM
GPU: Nvidia RTX 4090 , 24 Gb VRAM

For AI, it's the GPU that matters. Ideally, you want to load the entire AI model into the GPU VRAM. For models based on Flux.1 Dev, that's pretty much always 10 Gb or more... If you can't get it all into VRAM, it will be split between VRAM and RAM, with serious performance degradation and unstability as a result.
Conclusion: get a monster GPU with as much VRAM as you can get!

I use the Krita AI Diffusion plug-in, which includes a ComfyUI server, but takes away the hazzle of building ComfUI node trees from scratch... You work in the Krita UI and all the ComfyUI stuff happens under the hood. This made it easy for me to test several different Flux-based AI models and find the ones that worked on my hardware and gave me the result I want (check my gallery on DeviantArt !)

If VRAM is the issue, you can easily find "pruned" AI models that are smaller in size and works on your hardware. There are plenty of (uncensored!) Flux models available to download on CivitAI , find one that is smaller in size and test drive it on your machine!

[bakerboy]

In the last post ...Who is censoring/banning AI ???" I thought once you had the software and knew how to use it, you can do what you want?"

Most web based AI tools are restricted. ........ - bondage is okay, topless is okay, but nudity gets flagged.

Wow I did not know that. So AI adult videos I am waiting to develop ( see the Paris clip on p 2) are already censored. Good riddance, I will wait until the technology is developed enough to bypass these PC idiots. I guess I need to stick to a live production for now. Saaaaaad

[Kelderek]

In the last post ...Who is censoring/banning AI ???" I thought once you had the software and knew how to use it, you can do what you want?"

Most web based AI tools are restricted. ........ - bondage is okay, topless is okay, but nudity gets flagged.

Wow I did not know that. So AI adult videos I am waiting to develop ( see the Paris clip on p 2) are already censored. Good riddance, I will wait until the technology is developed enough to bypass these PC idiots. I guess I need to stick to a live production for now. Saaaaaad

The Paris clip is not made with a web based (and potentially censored/restricted) AI tool. It's most likely made with a local AI server installation, for capacity and performance reasons. No web based AI tool can handle the workload needed for such a large scale and professional production.

It's not the AI tool (software) itself that is restricted/censored, it's the AI model the tool uses:

Example: ComfyUI is the tool, the software running on your machine. It can't do anything unless you load it with an AI model.
The AI model is the "brain" that actually generates the content (pictures or movies). The AI tool (e.g. ComfyUI) is the software the runs the AI model.

There are plenty of AI models out there: Stable DIffusion, Flux, MidJourney etc... some are restricted, some are not.

I use Flux.1 Dev, which in the base version is restricted: it does not allow nudity.
However, since Flux is open source, it can be "re-trained" to allow pretty much anything... so there are Flux variants available that allows nudity and sex!
Consequently, I can do nude stuff with Flux, as you can see in my DeviantArt gallery I linked to above!

But: you're unlikely to find these Flux versions in the web based AI tools... you need to run them on your own AI tool. And for the kind of productions you want to do, you will need to run your own AI tool anyway, for capacity and performance reasons!