Forum
>
Site Discussion
>
Thread
Log in |
Register
For everyone saying "please no", this would of course be a sister site to GWM, not a part of the main GWM gallery. Or, at the least, it would be separated so you don't see both real and AI images at the same time.
Hmm, maybe. Although right now (kind of goes with what Chipperpip said) I'm not sure there's enough people making quality content mostly because the barrier to entry is so high. I think one problem the site would face is that it's too easy to pump out really mediocre gens and the site would get flooded with these, relative to the good ones at least.
GWM doesn't even have a great tag system and if I had this figured out I would implement it for GWM first before doing it for a different site.
This wouldn't be people generating images on the site itself (which would be a very difficult technical feat to pull off), but rather uploading their images they generated elsewhere.
I think the point of doing it is so that the best images can be upvoted, so we can start to actually give more focus to the good artists. and we can return to artists we like rather than them being lost in a thread somewhere. it might even push innovation and improvement. i know there is deviant art, but i actually think the functionality of gwm works better for this sort of thing. deviant art covers all sorts of topics and isnt focused enough. i agree, you should not mix the ai art with real humans. i also dont think that ai will mean we stop looking at real women. i mean c'mon. and i'm not sure what the pint of ai skepticism is. i mean it's not going to go away, no matter how much you complain, so what is the point of complaining exactly. but let's put it somewhere else, separate from the real girls.
chainer, i think it would be sufficient to replicate the gwm site as it is, call it gwmai, empty it of models, change model name to artist name as the upvoting system and boooooooom done. dont need to over complicate it.
Mmh now that chainer said he is not 100% against this idea you know what would also be cool. what if he or someone else would make a site where people can upload female muscle art like a deviant art alternative. I know a lot of people are frustrated with deviant art now. and I personally would like to have a site similar to that but without all the annoying futa and furry and art based on nickelodeon and cartoon network cartoons. and other random fetishes that are unrelated to female muscle
> > So, when do you expect to make your first million once the lawsuits have died down to a mere trickle. > > AI imagery is nothing more than extreme plagiarism.
Who said a website has to be for profit? Also lmao who is going to sue over AI images and win...
LOL! I think that you will be shocked at what is coming soon over so-called AI Art. As the age-old proverb says: Pride goes before a fall.
My idea, I don't know what you guys think but "girlswithfigure" or "girlswithshape" , I hope most of you like.
I'm grateful you guys viewed this criticality and considered this. The feasibility and ethical hurdles may be too high now but AI is evolving at exponential rates and starting this conversation now will hopefully keep our community at pace so we can be ahead when the time is right
On the tag side of things, Stable Diffusion/Automatic1111 stores the prompt, negative_prompt and seed in the image metadata so you could just require that to be present to upload an image (although this is often overwritten with upscaling) or check for the SD watermark to verify it is an AI generated image. You probably want some way to reward uploaders that provide their full workflow and custom LoRA’s in the spirit of crowd sourcing better images.
The other (computationally expensive way) is to just throw the image back through CLIP. Those tags are probably not the best for search, but should be good enough to cluster similar images for a next image recommender system. You could even have some folding@home style api/script where users donate compute time in exchange for fake internet points to save some cash here.
I’d suggest bringing back the 0-10 rating system, and clearing out images that don’t meet some threshold after a set number of days to avoid being overwhelmed by mediocre images.
Legal side you should be OK as long as people aren’t uploading deepfakes of Taylor Swift, since AI images can’t be copyrighted.
Intriguing but a few things come to mind. 1) Since the images will be idealized girls that check every box, most would probably have numerous tags. Search results could be very similar regardless of what filters you choose. But AI images would certainly have their own set of tags to filter on. 2) We know that several girls request that their photos/videos not appear on the site. If someone uses them as the basis for an AI image (and it's obviously them), will they still be able to have the AI image removed from this new site? 3) While I've seen some fantastic AI images, overall I can't get excited until the technology advances enough for people to produce high-quality and realistic video clips/animations. I generally only browse the videos on this site now.
So far AI imagery of muscular women (and men) has been disappointing. All that I have seen so far are highly derivative and very much the same. Quite dull and boring actually.
Fans don't need images that are more detached from reality than what we've already got.
I know you're just expressing a layman's opinion, but I have to disagree with you on that. The market - aka demand - determines what the fans want. The market's - aka supply - job is to address that want.
It's my opinion that there is a growing demand for this type of content. Now, will that translate to a viable and sustainable business? Time will tell on that part, more and more people are adopting the creator tools every single day and helping the developers to fine-tune the machines.
It's my opinion that there is a growing demand for this type of content. Now, will that translate to a viable and sustainable business? Time will tell on that part, more and more people are adopting the creator tools every single day and helping the developers to fine-tune the machines.
And now, it turns out that teams of very sophisticated Chinese hackers have managed to get involved in the development as volunteers under pseudonyms and have been having a field day inserting malicious code into just about every generator machine. That's lovely, isn't it.
I have just thrown away my Rose Coloured Glasses.
And now, it turns out that teams of very sophisticated Chinese hackers have managed to get involved in the development as volunteers under pseudonyms and have been having a field day inserting malicious code into just about every generator machine. That's lovely, isn't it.
That sound wildly overblown, I assume you're talking about shit like this?
That's a known issue with using pickle model files (.ckpt for Stable Diffusion ones)- they can potentially contain executable code. That's why most people switched to using the .safetensors format a while back, since it doesn't have that problem.
And now, it turns out that teams of very sophisticated Chinese hackers have managed to get involved in the development as volunteers under pseudonyms and have been having a field day inserting malicious code into just about every generator machine. That's lovely, isn't it.
Are you talking about the the xz Utils backdoor? https://arstechnica.com/security/2024/04/what...
> I made a website of my own to just host my AI generated art. > It lacks tags and filters yet but this could become a real idea.
So, when do you expect to make your first million once the lawsuits have died down to a mere trickle.
AI imagery is nothing more than extreme plagiarism.
That's 100% an assumption. You can draw your own image and have AI turn it into reality, you can upload your own images and create a model. Also training off thousands of pictures online is hardly plagiarism if it isn't reproducing the image. Its no different to a photographer recreating the Marilyn Monroe skirt image.
@norbertwinston If you want to be cryptic while not actually saying anything, go elsewhere.
While I'm not a fan of AI generated images, unless they are inspirations for interiors and architecture, I can certainly see this finding an audience. I say this based on the number of Facebook pages popping up with AI generated images, that are getting attention.
While I could see the site sharing some of the features of this one, it could be also form a test bed for ideas that may not necessarily be ready for a site showing pictures of real people.
One of the things that mods of such a site would need to watch out for is non-AI content, that looks AIish. Think of artists who make genuinely make art that could be confused with AI or photo manipulations, that aren't really AI. Also, given that a lot of these AI generation tools can churn out unique variants at a rapid rate, what is the upload rate likely to be and could the site just be generating its own content?
For everyone saying "please no", this would of course be a sister site to GWM, not a part of the main GWM gallery. Or, at the least, it would be separated so you don't see both real and AI images at the same time.
Hmm, maybe. Although right now (kind of goes with what Chipperpip said) I'm not sure there's enough people making quality content mostly because the barrier to entry is so high. I think one problem the site would face is that it's too easy to pump out really mediocre gens and the site would get flooded with these, relative to the good ones at least.
GWM doesn't even have a great tag system and if I had this figured out I would implement it for GWM first before doing it for a different site.
This wouldn't be people generating images on the site itself (which would be a very difficult technical feat to pull off), but rather uploading their images they generated elsewhere.