#me @ my whole deviantart

[florshedworf]

im sorry to say but im a scratch defender. yes people got banned for stupid ass reasons but i’ll tell you that shit kept me from getting actually harassed

[galactaknightyaoi]

honse…,

Honse indeed ��️

That's Judith. Or just Jude, for short.

She was my first ever OC, and the only one out of my OCs with a November birthday like me. So I figured she was a good choice for a birthday profile picture lol.

That's her. Isn't she sweet? :)

[flamboyant-king]

Wow fellas, I was going through my blog for something and realized I have almost 500 posts tagged #comic. Almost been here 10 years posting shitpost nonsense. I've got over 5600+ posts on this here blog. Maybe 80% is all drawing.

Look at me. Still going. One day, it will all amount to something big. But for now, stick around and enjoy the show.

Love yall.

[gorillaxyz]

THE THUNG SBE DREW ME. GRACE. SHE FTEEW ME NIXON L.

[shevr]

thinkin back to that post a while ago i'd tagged w/ "sorry for not being hornier" actually; like i know this stacks on top of my general issue with failing to draw more in general but sometimes i somehow feel bad or inadequate for not indulging in drawing nsfw things. never been a thing i feel like ive ever rly naturally left like doing so much

idk if that's me wishing i could feel like it, or wanting to do what the other cool artists do, or just another extension of wishing to do anything

but. monday 2am rambles, nothin serious. i can draw dicks whenever, whatever, no rush

[destinywillowleaf]

i'm working on a thing and trying to finish it before next week (before riley in masters is REAL real instead of just "in the game data set to be launched real" and i am losing my shit over the microwave send help

@jennajayfeather genuinely i still don't know how i didn't notice how similar the lines were but very much thank you for that

[verawhisk]

injury body horror kidnapping tw all that stuff sorry god!!

[ophexis]

trying not to think too hard about how quickly a certain art site that shall not be named went downhill the second they started doing their opt-in/out bullshit

[simonpetrilove]

2014 me would have LOVED all of this simon content

[the-acid-pear]

I still find it hilarious how I can say I got into Madness Combat Before It Was Cool but by a saving grace of like. A fucking week. Like the timing was legit uncanny.

[jkflay]

I LOVE BEING DISTURBING!!!!!!!! I LOVE BEING THE ONLY GOOD JK ONLINE!!!!!!!

I've rarely seen a more validating sentence in my entire life.

[englishsucks]

Do you wanna know how long it took me to answer 11 questions?

It took 30 Damn minutes! Do you have any idea how much of a time that is? It's around 3 minutes per question. My avarage time on a question is 30 seconds! I am a fast and efficient reader, if I answer 10 questions with 30 seconds per question there's at most only 1 question wrong! But here I am with 4 wrong answers out of 11 questions with 3 minutes per every stupid question when there are way better questions to ask.

But no AP lang just decides that they are so Fucking better and makes these weird ass questions so that I can just waste my ridiculously precious and limited time on switching between 2 pages because AP lang decided to be so fucking different and original that they make 2 options literally copy and paste except for one word and both those words are the words that even Shakespeare wouldn't use or think of ever creating. And the worst part is that 3 options out of 5 are correct and the AP langs defence is that one is "more correct" or "Truer"

Are you shitting me? I don't have time for AP lang staff not being able to decide between themselves about a stupid question so that they put both statements on the test as an option and set correct option according to the status on which one that was less picked by the students.

I don't have time for you stupid Americans ask me for the umpteenth time about the freaking American Dream! I will die against American Dream even if you keep deducting points from me over and over again. If you don't want politics in your goddammed exams then don't put your goddammed politics into the exam.

I have a Biology project waiting for me to finish it but here I am reading a ridiculously small fronted text in a ridiculously long amount of time about a ridiculous subject just because AP lang has nothing better to do except torment.

You think I am done? You wish! This is what AP lang has done to me I am now a master at extenting my ridiculous sentences. Just like how AP lang chooses their articles to ask students with long ass sentences that is longer than my actual height.

Even with all these shits that they keep pulling AP lang also never changes. "Oh Yea just ask them about the Animal farm, The Great Gatsby" and if they do this for 2 years theyll just pull another shit that's called "American Dream + Slavery + Freedom + Some Random Writer that no one in the history ever heard of but let's just put one of their fans essay about them that from 18th century" and I am sick and tired of this them keep pulling this shit. I don't give fuck about Thomas Carlyle and his fanboy writing an essay on how we should worship this man.

I hate AP Lang

[foone]

Your posts are in an AI model

and then Tumblr decided to sell them to AI models.

Now, don't get me wrong, tumblr selling out the users to AI companies is bad, yes, they shouldn't do that. It sucks.

but don't lets get this confused: your posts were already in there. Tumblr selling them is about tumblr making some money and about the AI models having more exhaustive post collections. It's not about your posts being in an AI model, vs not being in one. That battle has already been lost.

Can you find your post on google? Then it's almost certainly in an AI model already. Think about it: These AI sites showed up before all the sites were making deals to sell their users' content, right? How do you think they built them in the first place?

They scraped the posts. Just like google and bing and such do when they build their search indexes.

It's a fundamental part of how the open web works: you want your posts on tumblr to be visible to users, right? You want them to be readable?* Like, look how much stuff broke when twitter changed their whole read-while-not-logged-in policy, ruining a bunch of thread links/NSFW links. And if it's visible, it's scrapable. That's what the AI models were built on.

I've done website scraping before (not for AI models, of course. I was doing search engines and website archival), this is just how it works. You hire a few relatively smart CS graduates and tell them "build me a scraper that'll give us a bunch of tumblr posts" and they go off for a month or two and come back with a database of a few billion posts, and you stuff that into your AI model. That's how they got all the deviantart and flickr and twitter and pinterest and so on posts. They didn't pay for them: they just took them.

They only ever pay for this shit because either:

they fucked up in such a way that the site might be able to sue them for taking rather than paying

They can buy them cheaper than they can finish taking them. Maybe they'd need to pay the CS grads for an extra month? well, that might be more expensive than just throwing the site a couple hundred thousand bucks.

ANYWAY: my point is, don't treat this "oh no tumblr is selling our posts to AI" like it's a big thing that might happen and it would be bad to happen. Yes, it's bad, tumblr shouldn't do this, this'll let AI models get continual updates of content for far easier than just scraping them would be, tumblr betrayed user trust, and so on...

but realistically, this is not a black and white matter of "if only tumblr didn't do this, then we'd be safe from AI models!"

Nope. We already lost that battle. I'm sorry, and it does suck, but that's just how it is. The avalanche has already started, it's too late for the pebbles to vote. * I'm assuming here that you don't run a private blog that's set to only followers or something. You'd be safer then, of course, but you're not really my target audience for this rant

[ms-demeanor]

Why reblog machine-generated art?

When I was ten years old I took a photography class where we developed black and white photos by projecting light on papers bathed in chemicals. If we wanted to change something in the image, we had to go through a gradual, arduous process called dodging and burning.

When I was fifteen years old I used photoshop for the first time, and I remember clicking on the clone tool or the blur tool and feeling like I was cheating.

When I was twenty eight I got my first smartphone. The phone could edit photos. A few taps with my thumb were enough to apply filters and change contrast and even spot correct. I was holding in my hand something more powerful than the huge light machines I'd first used to edit images.

When I was thirty six, just a few weeks ago, I took a photo class that used Lightroom Classic and again, it felt like cheating. It made me really understand how much the color profiles of popular web images I'd been seeing for years had been pumped and tweaked and layered with local edits to make something that, to my eyes, didn't much resemble photography. To me, photography is light on paper. It's what you capture in the lens. It's not automatic skin smoothing and a local filter to boost the sky. This reminded me a lot more of the photomanipulations my friend used to make on deviantart; layered things with unnatural colors that put wings on buildings or turned an eye into a swimming pool. It didn't remake the images to that extent, obviously, but it tipped into the uncanny valley. More real than real, more saturated more sharp and more present than the actual world my lens saw. And that was before I found the AI assisted filters and the tool that would identify the whole sky for you, picking pieces of it out from between leaves.

You know, it's funny, when people talk about artists who might lose their jobs to AI they don't talk about the people who have already had to move on from their photo editing work because of technology. You used to be able to get paid for basic photo manipulation, you know? If you were quick with a lasso or skilled with masks you could get a pretty decent chunk of change by pulling subjects out of backgrounds for family holiday cards or isolating the pies on the menu for a mom and pop. Not a lot, but enough to help. But, of course, you can just do that on your phone now. There's no need to pay a human for it, even if they might do a better job or be more considerate toward the aesthetic of an image.

And they certainly don't talk about all the development labs that went away, or the way that you could have trained to be a studio photographer if you wanted to take good photos of your family to hang on the walls and that digital photography allowed in a parade of amateurs who can make dozens of iterations of the same bad photo until they hit on a good one by sheer volume and luck; if you want to be a good photographer everyone can do that why didn't you train for it and spend a long time taking photos on film and being okay with bad photography don't you know that digital photography drove thousands of people out of their jobs.

My dad told me that he plays with AI the other day. He hosts a movie podcast and he puts up thumbnails for the downloads. In the past, he'd just take a screengrab from the film. Now he tells the Bing AI to make him little vignettes. A cowboy running away from a rhino, a dragon arm-wrestling a teddy bear. That kind of thing. Usually based on a joke that was made on the show, or about the subject of the film and an interest of the guest.

People talk about "well AI art doesn't allow people to create things, people were already able to create things, if they wanted to create things they should learn to create things." Not everyone wants to make good art that's creative. Even fewer people want to put the effort into making bad art for something that they aren't passionate about. Some people want filler to go on the cover of their youtube video. My dad isn't going to learn to draw, and as the person who he used to ask to photoshop him as Ant-Man because he certainly couldn't pay anyone for that kind of thing, I think this is a great use case for AI art. This senior citizen isn't going to start cartooning and at two recordings a week with a one-day editing turnaround he doesn't even really have the time for something like a Fiverr commission. This is a great use of AI art, actually.

I also know an artist who is going Hog Fucking Wild creating AI art of their blorbos. They're genuinely an incredibly talented artist who happens to want to see their niche interest represented visually without having to draw it all themself. They're posting the funny and good results to a small circle of mutuals on socials with clear information about the source of the images; they aren't trying to sell any of the images, they're basically using them as inserts for custom memes. Who is harmed by this person saying "i would like to see my blorbo lasciviously eating an ice cream cone in the is this a pigeon meme"?

The way I use machine-generated art, as an artist, is to proof things. Can I get an explosion to look like this. What would a wall of dead computer monitors look like. Would a ballerina leaping over the grand canyon look cool? Sometimes I use AI art to generate copyright free objects that I can snip for a collage. A lot of the time I use it to generate ideas. I start naming random things and seeing what it shows me and I start getting inspired. I can ask CrAIon for pose reference, I can ask it to show me the interior of spaces from a specific angle.

I profoundly dislike the antipathy that tumblr has for AI art. I understand if people don't want their art used in training pools. I understand if people don't want AI trained on their art to mimic their style. You should absolutely use those tools that poison datasets if you don't want your art included in AI training. I think that's an incredibly appropriate action to take as an artist who doesn't want AI learning from your work.

However I'm pretty fucking aggressively opposed to copyright and most of the "solid" arguments against AI art come down to "the AIs viewed and learned from people's copyrighted artwork and therefore AI is theft rather than fair use" and that's a losing argument for me. In. Like. A lot of ways. Primarily because it is saying that not only is copying someone's art theft, it is saying that looking at and learning from someone's art can be defined as theft rather than fair use.

Also because it's just patently untrue.

But that doesn't really answer your question. Why reblog machine-generated art? Because I liked that piece of art.

It was made by a machine that had looked at billions of images - some copyrighted, some not, some new, some old, some interesting, many boring - and guided by a human and I liked it. It was pretty. It communicated something to me. I looked at an image a machine made - an artificial picture, a total construct, something with no intrinsic meaning - and I felt a sense of quiet and loss and nostalgia. I looked at a collection of automatically arranged pixels and tasted salt and smelled the humidity in the air.

I liked it.

I don't think that all AI art is ugly. I don't think that AI art is all soulless (i actually think that 'having soul' is a bizarre descriptor for art and that lacking soul is an equally bizarre criticism). I don't think that AI art is bad for artists. I think the problem that people have with AI art is capitalism and I don't think that's a problem that can really be laid at the feet of people curating an aesthetic AI art blog on tumblr.

Machine learning isn't the fucking problem the problem is massive corporations have been trying hard not to pay artists for as long as massive corporations have existed (isn't that a b-plot in the shape of water? the neighbor who draws ads gets pushed out of his job by product photography? did you know that as recently as ten years ago NewEgg had in-house photographers who would take pictures of the products so users wouldn't have to rely on the manufacturer photos? I want you to guess what killed that job and I'll give you a hint: it wasn't AI)

Am I putting a human out of a job because I reblogged an AI-generated "photo" of curtains waving in the pale green waters of an imaginary beach? Who would have taken this photo of a place that doesn't exist? Who would have painted this hypersurrealistic image? What meaning would it have had if they had painted it or would it have just been for the aesthetic? Would someone have paid for it or would it be like so many of the things that artists on this site have spent dozens of hours on only to get no attention or value for their work?

My worst ratio of hours to notes is an 8-page hand-drawn detailed ink comic about getting assaulted at a concert and the complicated feelings that evoked that took me weeks of daily drawing after work with something like 54 notes after 8 years; should I be offended if something generated from a prompt has more notes than me? What does that actually get the blogger? Clout? I believe someone said that popularity on tumblr gets you one thing and that is yelled at.

What do you get out of this? Are you helping artists right now? You're helping me, and I'm an artist. I've wanted to unload this opinion for a while because I'm sick of the argument that all Real Artists think AI is bullshit. I'm a Real Artist. I've been paid for Real Art. I've been commissioned as an artist.

And I find a hell of a lot of AI art a lot more interesting than I find human-generated corporate art or Thomas Kincaid (but then, I repeat myself).

There are plenty of people who don't like AI art and don't want to interact with it. I am not one of those people. I thought the gay sex cats were funny and looked good and that shitposting is the ideal use of a machine image generation: to make uncopyrightable images to laugh at.

I think that tumblr has decided to take a principled stand against something that most people making the argument don't understand. I think tumblr's loathing for AI has, generally speaking, thrown weight behind a bunch of ideas that I think are going to be incredibly harmful *to artists specifically* in the long run.

Anyway. If you hate AI art and you don't want to interact with people who interact with it, block me.

[tricksterlatte]

ngl this is just a silly post I made while goofing around, but finding out the two sides to this newfound issue are either "Akechi is a Norse mythology nerd and goes online to write essays on how badly Marvel messed everything up" or "Akechi actually kins MCU Loki due to his own relationship with Shido and his mother versus MCU Loki with Odin and Frigga" is actually hilarious.

If Persona 5 begins during 2016, this would mean Akechi previously awakened to Loki during the prime time of MCU Loki being a tumblr sexyman. Do whatever you’d like with that information

[destinationtoast]

1/3 - Hi there! Three (I think) part ask incoming. You're the main person I know of who compiles tons of interesting fandom stats, so I wanted to ask you about it if you have the time to answer. :) I think a lot about how AO3 works great as a fan*fic* archive, but for other fanworks, like images, audio, video, etc., it's only as good as wherever the media is being hosted. With the way hosting sites come and go, or change their TOS to nuke nsfw or queer content, etc., it makes me wonder

how many broken image links litter AO3 at this point. I know it's not considered the primary place to find fanart, but a lot of folks do post images there—for events like Big Bangs, as standalone art, and even as decorative section breaks, etc. My question is: do you think there's a way to look at, say, works tagged with #fanart (of which there are 99,504 atm) and determine what percentage of those are broken links? From what little I understand, one would have to (perhaps with the use of a simple bot?) try to open any link bordered by the <img src> html, and see what portion of those return an error versus what ones actually load? I suppose it could even be something like looking at fanart posted in 2007, 2012, 2017, and 2022 to compare how many older links are broken versus newer links. Anyway, this may be completely unfeasible, but I figured I'd ask about your thoughts! Thanks!

Ooh, thanks for the great question! I took a while to answer because I wasn't initially sure what to recommend and ended up gathering some data to investigate. (If anyone else also has relevant data, please share in the notes!)

I liked your idea of looking at samples different years going back, and I decided to look through 100 AO3 works tagged "Fanart" (or a subtag) that were posted 10 years ago -- as a very fast starting point, I didn't even take a random sample of works, I instead looked at the first 100 multimedia fanworks posted in July 2014. (And August, when necessary; see more notes on methodology at the end.) Please keep in mind that this sample that may not be very representative of AO3 more broadly; to get better estimates, more sampling would be needed. Based on this initial data gathering (and the fact that most fanworks on AO3 were posted within the past 10 years), I would tentatively guess that that most fanart, fanvids, and podfic on AO3 still have accessible multimedia.

Given how many broken links and embeds there are on older webpages, I assumed that a ton of the links from 10 years ago would be broken. But I was pleasantly surprised by the results:

Wow -- 10 years on roughly 90% of the multimedia still works! I was honestly floored; I'd been originally planning to also look at 5 years ago to see how much better that was, but if ~90% are still working 10 years on, 5 years ago doesn't have room to be dramatically better. (However, I'd love to see more follow up sampling across different years to find out.)

There were a lot of AO3 users in this sample who posted multiple works -- some posted as many as a dozen multimedia works in July 2014. I didn't want the results to be overly skewed by any one fanwork creator, so I also redid the analysis with just one work from each unique creator:

Okay, cool, those results are pretty similar. I also did some further breakdowns on this smaller set of works to look at which hosts creators were using, and how many of the hosts were still working:

The most common fanart host used in this sample was Tumblr, then wixmp -- which I think from some very quick googling might be because Deviantart switched to using Wix for image hosting at some point? (i.e., I think most of those artists may have posted their art on Deviantart, then linked to/embedded the image on AO3, and the image's direct URL was was wixmp.) There were a few other hosts at the time that were used by 5+ different artists in the sample, and then there were a whole lot of hosts were used by just one or a few artists.

Most of the 10-year-old fanart is still up for all of these hosting categories! Photobucket is the least reliable of the most commonly used hosts. In the Other category, 25% of the links are broken, but that's still better than I expected (see full host list here).

This is getting long, so I'm moving the breakdowns for fanvids and podfic beneath the cut:

Fanvids were almost all hosted on YouTube, Vimeo, or both (the above categories are not mutually exclusive). All the Vimeo links still worked, whether they required a password to view or not. Most YouTube links were working, and the few missing ones had almost all been taken down by YouTube for copyright reasons (according to the errors I got -- I'm not rendering judgment about whether they were actually fair use), rather than by the vidder who posted it. And almost a third of vidders also linked to other hosts besides the big two, but many of those links were broken; 59% still worked. (see full host list here)

For podfic, mediafire was a popular solution 10 years ago, though many podficcers used it as a backup rather than the main link that they shared. A lot of podficcers made use of a fandom hosting site that specialized in podfic -- either parakaproductions.com or audiofic.jinjurly.com. Four podficcers used soundcloud (often as a backup). And once again there were a lot of less-frequently used hosts, often used as backup links; 69% of those still worked. (see full host list here)

Some methodology notes and further thoughts:

For fanvids and podfic (but mostly not from fanart), the fanwork creators tended to provide multiple links, and in those cases, I counted the multimedia as working if at least one of the links was still working.

I counted embedded media and links to other sites that host the media all the same way.

I counted the media as broken if I got a 404 when I tried to visit it, or if a site like YouTube had taken it down due to copyright issues, or if I got an Access Denied message for a site like Google Drive.

I counted the media as working if it required a password that was given on the page (common with Vimeo), or if an embed was broken but there were working links to other sites.

How representative is this data? Well, these samples contained most/all of the multimedia fanworks posted in July 2014; that month, there were 70 fanvids, 135 podfic, and 186 pieces of fanart posted that haven't been deleted since. So it's pretty representative of July 2014 specifically. :) But there could have been, say, a fanwork challenge going on in July 2014 that caused unusual uploading patterns then.

The above data gathering and analysis took me several hours over several days. If you want to follow up, you could do more data gathering similar to what I did (I'm happy to elaborate on my process as needed). Or you could write a bot to do something similar; you could have it fetch more AO3 fanworks and try following the links within each work. However, that would be slightly tricky; I ran across more kinds of errors and complicated situations than I expected (e.g., if a YouTube video has been taken down due to copyright, it still has a working YouTube page; sometimes an embed is broken, but if you open the link within the embed in a separate window, it still works fine; many Vimeo links require a password to test, and it could be hard for the bot to reliably find the password in the surrounding text). So you'd have to program your bot to be able to handle a bunch of different special cases.

Regardless of which path you are considering, if you or anyone else does any follow up work here, I encourage you to start by looking through a bunch of fanworks yourself and deciding which scenarios you want count as "working" vs. "not working," and any other things you want to pay attention to.

Hope that helps, and please feel free to DM me with follow up questions. And if you follow up, please share anything else you figure out in this space!