AI Generators - Unrestricted

[Ultegra]

I'm not a big poster here. Mainly I just lurk around are read. Lately, I've been messing around with some basic AI generating tools like text to video and image to video but for the most part, most AI generators are restricted. I'm looking for any recommendations for AI generation tools that are unrestricted. I have been able to create some standard stuff like my wife dancing, hair style changes and background replacement, but she has to be clothed or the generator will not produce the image or video. Mind you, these are 5-10 second vids.

Believe it or not, my wife is thinking of breast augmentation surgery and we would like to generate some images with bigger breasts using her body. Yes, I know a surgeon probably has some software that given a current image then can add different types of breasts, but she really just wants to play around with the prospect of having larger breasts and AI should be able to do this. Finding a generation tool that is unrestricted for content though is the tricky part.

I have searched all the forums the best I could and most of the threads dealing with AI center around story creation. I am not an author. I just need some suggestions for some unrestricted AI generators.

Any help would be greatly appreciated.
Ultegra

 

[yano2mch]

If you're looking for cloud use, Grok or Venice.AI are suppose to be good.

If you want to do it locally using your own hardware, you use a combination of tools and models.

Recommended you have a GPU, and a LOT of hard drive space and ram. I'm not kidding. The models you'd probably be using are in the 6Gb for image (SDXL) while LLM models range in the 3B (Billion) to 1T (Trillion). For simplicity assume the node is the same size as the file, and 70B's or some 24B-50B node MOE (Mixture of experts) are decent and much faster than non MOEs. But unless you have say 64Gb Ram and 8Gb VRam the quality will be considerably lower and slower.

Regardless unless you have tens of thousands in GPU hardware, expect it to take a while. For me it takes 5min per 1 image (Using Geforce 3070), so you might do a batch of files overnight and see if among the 50 you made if 2-3 are actually worth using. Text... 1-15 tokens a second is a good speed in my opinion. 10t/s is fast enough it's faster than you can read it.

As for story generation... TOS here says they don't allow AI uploads. Which is understandable. I tend to use LLM's for roleplaying, and might convert some to stories/books later but right now have not shared any. Having it just 'write a novel' doesn't work that well. You often get half-assed drafts that need heavy editing or tends to have unrelated fluff that literally just takes up space.

Though RPing quality... can be another matter entirely.

I'm fine with talking pros and cons of the tech as a tool and helping if i can, just realize you probably can't put anything you generate here.

 

[Kevinjkson]

Do you recommend a particular image gen model to use at home on our own hardware that delivers good output?

 

[yano2mch]

Do you recommend a particular image gen model to use at home on our own hardware that delivers good output?

'Good Results' is under interpretation. But i'd say so, several actually. Civitai hosts these

JuggernautXL Ragnarok
Diamond XL
Hezi League of Legends
JRD Green Whisper (deleted, but civitai archive has a copy)
Unholy Desires Temptation

I often find 1 in 20 to be immediately usable, and 3-4 need touchups before they'd be worthy of anything.

Though ADetailer to fix face/hands takes 3x longer but may salvage a close image to a workable image.

 

[PeteyPP]

I haven't used it yet but a friend has gotten some great results with this
https://perchance.org/ai-text-to-image-generator