stability diffusion vore

Keep our community informed! This forum is for discussing and sharing vore-related information. Post any relevant material and/or links here, and engage in conversations!
Forum rules
This is for general discussion, if you found something you want to post, please use one of the upload forum, if you made something and want to share them, please use the work to be shared forum!

stability diffusion vore

Postby philip92dk » Tue Sep 06, 2022 1:57 pm

i wonder if it's possible to make vore art with it
philip92dk
New to the forum
 
Posts: 7
Joined: Mon Sep 14, 2020 2:41 pm

Re: stability diffusion vore

Postby MyHeartInAcid » Tue Sep 06, 2022 8:31 pm

Elaborate what you mean by that?
MyHeartInAcid
Somewhat familiar
 
Posts: 109
Joined: Thu Sep 09, 2021 11:18 pm

Re: stability diffusion vore

Postby TestAccountPleaseIgnore » Tue Sep 06, 2022 9:54 pm

MyHeartInAcid wrote:Elaborate what you mean by that?


Looks like Stable Diffusion is a text-to-image AI, sort of like DALL-E.
User avatar
TestAccountPleaseIgnore
New to the forum
 
Posts: 15
Joined: Sun Dec 12, 2021 4:56 am

Re: stability diffusion vore

Postby MyHeartInAcid » Tue Sep 06, 2022 10:10 pm

TestAccountPleaseIgnore wrote:
MyHeartInAcid wrote:Elaborate what you mean by that?


Looks like Stable Diffusion is a text-to-image AI, sort of like DALL-E.


Oh, I see! Thank ya
So many of these things now
MyHeartInAcid
Somewhat familiar
 
Posts: 109
Joined: Thu Sep 09, 2021 11:18 pm

Re: stability diffusion vore

Postby AlluringPredation » Thu Sep 08, 2022 9:43 am

Discounting finetuning, with img2img and textual inversion, very likely. Will it be good? Not for a while. Even in the best of outcomes there are glaring anatomical issues that make AI art easy to spot, notably mangled hands. I know a lot of artists bemoan inability to draw hands, AI is worse.

I'm loathe to the eventuality of 'AI art' being uploaded here, being that it can not be given US copyright as authorship can only be granted to a human. I feel it has no place in traditional art galleries, and all it will take is a few people to start flooding the front page with their generations to drown out handcrafted work entirely. Plugging some text into a prompt does not make one an artist.

AI assisted art would be a different story where the AI assists with completion rather than doing all the work, leaving a lot more room for individual control beyond rolling and re-rolling until a desired output occurs. Photoshop and Krita and working on plugins to incorporate AI assistance into their programs, used proper can save a lot of time for rendering. Improperly, we're going to end up with a lot of art that looks traced/overpainted/photobashed to hell.
Discord for artists and those who commission them! https://discord.com/invite/F3UDqqAx53
User avatar
AlluringPredation
Participator
 
Posts: 293
Joined: Fri Mar 27, 2015 11:10 am

Re: stability diffusion vore

Postby pendingdelet205h2n9 » Sat Sep 10, 2022 12:35 pm

I agree with AlluringPredation. Currently it's pretty terrible at explicitly erotic stuff unless you re-render it and edit a lot afterwards.
It can do some basic nudity, but that's about it.

Probably within the next few months there's going to be a huge wave of AI generated art on this platform, and every other.
Eventually it will be high quality enough that we can just enjoy infinite great content (several years down the line), but until then we're going to be drowned in janky-ass renderings for awhile, which I am not looking forward to...

Maybe we'll have a "no AI generated images" rule...

For context here's a few of the better erotic-ish images I was able to get:

Spoiler: show
00003.png

00004.png

00002.png

00002octb.png

00000loba.png


You don't want to see the lovecraft-esque images it came up with when I tried to do anything vore-ish...
User avatar
pendingdelet205h2n9
Been posting for a bit
 
Posts: 50
Joined: Tue Feb 14, 2017 8:44 pm

Re: stability diffusion vore

Postby LightDragon » Sat Sep 10, 2022 5:09 pm

Yeah I wouldn't worry too much if I were you...

I haven't seen a single AI-generated vore art that was semi-decent.
LightDragon
Participator
 
Posts: 209
Joined: Mon Aug 24, 2015 3:16 am

Re: stability diffusion vore

Postby MementoMori » Sat Sep 10, 2022 5:40 pm

Yeah, I second the sentiments here. FurAffinity just released a rule banning the posting of AI-generated images, and I do hope all artist platforms follow suit. Once the technology picks up, people don't need to sift through spam posts, if they want to use the AI they can easily just google it and utilize it themselves. But yes, the AI art is definitely not the prompter's creation.
It's technically the bot's creation, but even then, it's an amalgamation derived from countless real artists' works, of that particular prompted tag or theme, that came before (getting pretty Matrix-esque around here).
So I don't think someone should be allowed to post it - when AI art advances enough and becomes popular, the art platforms should be a safe haven where people go to still see real art made by.. well, actual artists. Sad to say that art may soon become a dying art! :? Didn't see that one coming, at least not in my lifetime..
But that been said, I believe the rules should adapt in time to leave AI art as more of a novelty, and as a device used to easily create concept art/assist a creator with larger projects.
I've thought a lot about AI art when it first debuted, and to me, the biggest factor saving the real artists is their styles and consistency. If someone used AI art, once it advanced enough, to try to pretend they were a real artist, the illusion would fail from those two factors, at least. Even if they always chose works that were, for example, tagged as "digital painting" or "graphite pencil drawing", they would not retain a cohesive look, style, or tone between the artworks. I noticed too that AI art, even the highest most impressive models, seems to be lacking in some of the 8 elements of art, such as symbolism and composition, which makes it have a sort of uncanny valley. You know it's art, and it could even be really good - but it's missing this certain something, and it has this sort of emptiness. I suppose in the end, these elements will become the "checkmark if you're a human test" for artwork.
[ ʟɪᴠᴇ ʟɪᴋᴇ ʏᴏᴜ'ʀᴇ ᴅʏɪɴɢ ]
Image
User avatar
MementoMori
Advanced Vorarephile
 
Posts: 708
Joined: Thu May 31, 2007 11:00 pm
Location: US

Re: stability diffusion vore

Postby jaggedjagd » Tue Sep 13, 2022 7:49 am

AlisaBlueBird wrote:You don't want to see the lovecraft-esque images it came up with when I tried to do anything vore-ish...

Actually Lovecraft-esque vore is my fetish, so... POST IT.
User avatar
jaggedjagd
Advanced Vorarephile
 
Posts: 755
Joined: Sun Dec 08, 2013 12:45 pm

Re: stability diffusion vore

Postby jaggedjagd » Tue Sep 13, 2022 8:02 am

MementoMori wrote: I noticed too that AI art, even the highest most impressive models, seems to be lacking in some of the 8 elements of art, such as symbolism and composition, which makes it have a sort of uncanny valley. You know it's art, and it could even be really good - but it's missing this certain something, and it has this sort of emptiness. I suppose in the end, these elements will become the "checkmark if you're a human test" for artwork.

I disagree on the composition, you can't tell me art like this:



Doesn't look breathtakingly beautiful. Sure there's a certain samey-ness to it, since it's all done by the same AI (Midjourney), but that's called an art style. Every artist has a level of repetition in their art, or is favoring certain symbols and themes over others. Midjourney seems to have a thing for very colorful clouds in sort of a impressionist style. Think Claude Monet. Maybe i'm biased becaue i love impressionist paintings, or maybe i'm not human. I failed the human test cause i like algorithm generated junk. As for meaning, meaning is always derived from the observer, what's a touching piece of art to one may be a pile of non-sensical junk to another.

Or maybe it's really just cause i have a fetish for lovecraftian shit, and a lot of AI art is some of the most lovecraftian shit i've ever seen. Considering it's quite literally made by an inhuman, unthinking entity... this is as close to the face of Cthulhu as we'll get.
User avatar
jaggedjagd
Advanced Vorarephile
 
Posts: 755
Joined: Sun Dec 08, 2013 12:45 pm

Re: stability diffusion vore

Postby masterofmanyforms » Tue Sep 13, 2022 6:04 pm

I am aware that my opinion on the internet is meaningless and that this website, like all other websites with galleries, depends on ancillary sales and business to keep content flowing. But I beseech that at worst AI art needs a special tag or be sent to its own forum (we do that for photoshop pieces).

Something endearing about this site is that it is not a fine art society, but rather a place to share sensual experiences through various mediums. Not all of the art on this website will appeal to everyone's sense of quality, but that's fine. Artistry, like everything else, is subjective. For years, people have debated at what level of granularity collages become one person's new art. Instead of prohibiting it, placing it in a separate location will allow for greater freedom of expression and the sharing of experiences from those who have difficulty expressing themselves.
masterofmanyforms
New to the forum
 
Posts: 7
Joined: Wed Sep 01, 2021 2:08 pm

Re: stability diffusion vore

Postby Aleph-Null » Thu Sep 15, 2022 1:11 am

AlluringPredation wrote:I'm loathe to the eventuality of 'AI art' being uploaded here, being that it can not be given US copyright as authorship can only be granted to a human. I feel it has no place in traditional art galleries, and all it will take is a few people to start flooding the front page with their generations to drown out handcrafted work entirely. Plugging some text into a prompt does not make one an artist.

MementoMori wrote:It's technically the bot's creation, but even then, it's an amalgamation derived from countless real artists' works, of that particular prompted tag or theme, that came before (getting pretty Matrix-esque around here).
So I don't think someone should be allowed to post it - when AI art advances enough and becomes popular, the art platforms should be a safe haven where people go to still see real art made by.. well, actual artists.


There is no legal precedent showing that all text to image AI art, or image to image AI art are the creations of the algorithm. So, no it is not technically the bot's creation. Additionally, these systems do not work by amalgamating other artistic works.

It isn't clear at this time if the output of text to image systems is copyrightable. Precedent for copyrighting AI generated artwork has been based on academic efforts to argue that machines should be able to copyright images (to date all efforts have failed.) But it must be noted that the failure was explicitly because the copyright owner was listed as a machine. Similar efforts to copyright religious text that are attributed to angels have failed. It isn't that the texts were not copyrightable, it was that the copyright holder was listed as an angel (not a human).

There is no precedent or ruling that gives guidance on where a certain level of algorithmic involvement in the creation of an otherwise copyrightable work nullifies a copyright.

I think it would be a mistake to ban AI generated art at this point. I don't think we are at a point where we know what the use will be to artists. The only real negative that I have heard from any artist community is spam. Spam is a problem in most artist communities, including this one. I think the solution to this problem should be to rework the gallery so that new works are not pushed off the latest updates page by mass uploads. Additionally, better tools for filtering what you see in latest updates could help out.

For example, group all of an artist's uploads over a rolling 24 hour period into one entry on latest artwork, and only have one bump to the top of the list per hour.
Aleph-Null
Advanced Vorarephile
 
Posts: 798
Joined: Mon Aug 07, 2006 11:00 pm

Re: stability diffusion vore

Postby LightDragon » Thu Sep 15, 2022 8:27 pm

If the images are ever copyrighted, the copyright will go to developers of the AI, not the final users.

The only thing the end user does is to write "male magician in a purple dress equipped with a star-shaped magic wand and a golden grimoire". Anyone who inputs this phrase into the AI will be given similar results, and that sentence is not copyrightable.

Aleph-Null wrote:Additionally, these systems do not work by amalgamating other artistic works.

They literally do. They reproduce the work of millions of pre-existing artworks, assembling them randomly, and prioritising the reproduction of the elements mentioned in the input text.
LightDragon
Participator
 
Posts: 209
Joined: Mon Aug 24, 2015 3:16 am

Re: stability diffusion vore

Postby masterofmanyforms » Fri Sep 16, 2022 9:37 am

I would like to be transparent in my argument by explaining that AI stable diffusion is more than just text prompts. he scope of the ability to control the algorithm includes the ability to feed the program new information via an image in order to make it serve one's own creation. As an example, I used an image I checked that was not in the database for Stabile diffusion I made a corporate art thing with a boat at the bottom.
Image

I then badly drew some poor details, swapped colors, and add a cursed blimp shape.
Image

Then I ran it through stable diffusion tweaking the sliders and running a few attempts and this is what the software created. prompt was A blimp hovering above the ocean,((boat)),(Clouds). Note that the software allows increasing the weight of some parts of the word input for more robust control.

Image

All I am trying to show here is that one of the talented artists here can use this software as a viable tool and keep everything to their own expectations of composition. I in my own research have managed to shape so many different art styles that unlike with mid-journey stable diffusion if the artist is willing to master it will not lead to overly similar styled art spammed.
masterofmanyforms
New to the forum
 
Posts: 7
Joined: Wed Sep 01, 2021 2:08 pm

Re: stability diffusion vore

Postby RoTheHuman » Wed Sep 21, 2022 9:09 am

I've been playing around with stable diffusion and trying to find prompts that produce anything particularly vorish. So far no real success! The closest I came (and I can't remember my specific prompt) was for the inside of a monsters stomach full of slime, but it of course doesn't look particularly good. I haven't managed to get it to produce an image of any kind of creature eating another.

I think something like vore is too esoteric for its training. Now if someone took one of these AIs and built a model based on a ton of varied vore artwork, then maybe we'd get something interesting.
User avatar
RoTheHuman
Intermediate Vorarephile
 
Posts: 363
Joined: Sun Nov 13, 2005 12:00 am

Re: stability diffusion vore

Postby masterofmanyforms » Wed Sep 21, 2022 5:58 pm

RoTheHuman wrote: Now if someone took one of these AIs and built a model based on a ton of varied vore artwork, then maybe we'd get something interesting.


the process requires some very good hardware but it is both documented and doable. In the meantime, it is best used with images you mockup and then edit in parts with IMG2IMG.
masterofmanyforms
New to the forum
 
Posts: 7
Joined: Wed Sep 01, 2021 2:08 pm

Re: stability diffusion vore

Postby RoTheHuman » Thu Sep 22, 2022 10:24 am

masterofmanyforms wrote:the process requires some very good hardware but it is both documented and doable.

Imagine feeding it all of the images on this site! I certainly don't have the hardware or time to do that though...
User avatar
RoTheHuman
Intermediate Vorarephile
 
Posts: 363
Joined: Sun Nov 13, 2005 12:00 am

Re: stability diffusion vore

Postby mrgrpm » Sat Oct 01, 2022 9:29 am

I tried Stable diffusion with Yiffy - Epoch 18 model.
Its mainly for furry but you can generate some semi-good non-furry results.
Image
More examples - https://imgur.com/a/HsxgFOf
User avatar
mrgrpm
New to the forum
 
Posts: 3
Joined: Sun Jan 13, 2013 4:01 pm

Re: stability diffusion vore

Postby RoTheHuman » Mon Oct 03, 2022 3:05 pm

mrgrpm wrote:I tried Stable diffusion with Yiffy - Epoch 18 model.
Its mainly for furry but you can generate some semi-good non-furry results.
[ https://i.imgur.com/wM4GhP4.png ]
More examples - https://imgur.com/a/HsxgFOf

Oh cool! I didn't realize you could do more training for the models.

Could you share a link to this furry-specific model?

Also, I found a link to someone that explains how to do it: https://github.com/LambdaLabsML/example ... finetuning

I gotta give this a try! I think the artwork here is ripe for trying to add training to a model, because all of our artwork is tagged, sometimes extensively! I might try seeing if I can take a small selection of art from here along with their tags and doing just that. Maybe some of the art from my favorites... *devious maniacal cackling*
User avatar
RoTheHuman
Intermediate Vorarephile
 
Posts: 363
Joined: Sun Nov 13, 2005 12:00 am

Re: stability diffusion vore

Postby mrgrpm » Tue Oct 04, 2022 2:54 pm

This is the model I used.

Yiffy - Epoch 18

IMPORTANT NOTE: during training explicit was misspelled as explict. Using the misspelled version results in better cocks.

This will likely be the last epoch for quite some time, because until further improvements to other parts of the network (such as a custom VAE, variable res, etc), we think that further training is not necesary because the model has converged on a decent style for now.
Direct download: https://sexy.canine.wf/file/yiffy-ckpt/yiffy-e18.ckpt
Magnet:
Code: Select all
magnet:?xt=urn:btih:b177dd04ae7062b541c82ad26f897e0a9fa514f4&dn=yiffy-e18.ckpt&tr=udp%3a%2f%2ftracker.opentrackr.org%3a1337%2fannounce&tr=udp%3a%2f%2ftracker.openbittorrent.com%3a6969%2fannounce&tr=http%3a%2f%2ftracker.openbittorrent.com%3a80%2fannounce&tr=udp%3a%2f%2ftracker.torrent.eu.org%3a451%2fannounce&tr=https%3a%2f%2fopentracker.i2p.rocks%3a443%2fannounce&tr=udp%3a%2f%2fopen.stealth.si%3a80%2fannounce&tr=udp%3a%2f%2ftracker.tiny-vps.com%3a6969%2fannounce&tr=udp%3a%2f%2ftracker.moeking.me%3a6969%2fannounce&ws=https%3a%2f%2fsexy.canine.wf%2ffile%2fyiffy-ckpt%2fyiffy-e18.ckpt


SHA256: 64e242ae67cb1fc85110a68242c936fa6f3e33077f2bfed48c4769959dff6c61
MD5: dbe25794e24af183565dc45e9ec99713

-----------------------
Trained for another two epoches using a dataset of 210k images from e621.
Direct download: https://sexy.canine.wf/file/yiffy-ckpt/yiffy-e15.ckpt
Pixeldrain mirror: https://pixeldrain.com/u/qkRKKpqg
Mega mirror: https://mega.nz/file/zVY2habK#o8qyxHuob ... mheD_kw6oA
MD5: cc354964edbcbb7758cb359743aa1791
The blur is somewhat fixed but some prompts still make it misbehave (EDIT: the prompt with issues was identified and there's no solution because the artist likes to draw blurry images)


You don't need to train it yourself. Just replace your current model with another one.
Training your model is really hard both for dataset(100K+ tagged images) and hardware (20GB+ of VRAM)

Here are some another examples https://imgur.com/a/dlxI99i
Image
User avatar
mrgrpm
New to the forum
 
Posts: 3
Joined: Sun Jan 13, 2013 4:01 pm

Next

Return to General Vore Discussion

Who is online

Users browsing this forum: AquaticCitizen, ClarAya, randomguy15, thechimpo