Bias disclaimer: I am personally largely unaffected by deletions. If my webUI ran on Steam it would show a total playtime of like 2000 hours. I also have a decent GPU and therefore have both the knowledge and capability to more easily do quality AI gens. I also know several of the Janitors and am likely to get implicit preferential treatment. I'm definitely in the 1%.

I think the best example of the current state of the site was #90021. The user announces their departure of the site, naming emotional stress over deletions.
Submissions getting deleted felt bad to them. The quality standards are haphazard, arbitrary. The user says with irony "This will be my last post, unless deleted of course".
What's the point of improving after all, if there is no space to improve in, to show progress.
The upload was later deleted for having 2 mouths. And it did have 2 mouths.

_____________________

AI is 'complicated'. I don't necessarily mean doing AI art, I mean the entire 'ecosystem'. You are able to produce industrial quantities of good looking but ultimately uninteresting slop, not because individual pics are bad in isolation, but because they become uninteresting next to each other. Therefore there is a reasonable quality standard towards not posting variation seeds. Dystopian examples of how this can go bad can be found on the CivitAI image tab or rule34. E6ai (correctly in my opinion) does not want to end up like a dumping ground like rule34 has. I even think e6ai has a unique opportunity here. Hardly anyone I know would argue against this standard.

AI has a non-trivial learning curve to it in order to get the output that you like. This learning curve however has been getting smaller and easier as time has progressed. On the contrary, by relation the hardware requirements have also risen. This presents an annoying conundrum, where users with older graphics cards that are stuck on SD1.5 models, may be getting gated out of participating. SDXL models are smarter and the higher resolution works to several benefits, I'm sure the overall shift with what you are seeing on the frontpage these days is pretty evident towards the overall advancements in AI. But in general you have to go with the times, even then you have lower hardware requirement models with strong capabilities like ChromaXL. You also get remote stuff like CivitAI, but everyone knows these are highly limiting. Therefore I cannot blame the administration for their standards towards higher resolution, reasonable resolutions achieved through GAN upscaling, 'overall' quality standards, and being somewhat picky about anatomical errors.
Info on how to properly use tools like inpainting and the 'meta' for current models is still obtuse. From my experience there isn't too much knowledge gatekeeping, unless it's people doing a heavy grift. You can DM people on here. Join the Discord or something.

A lot of problems exist in the theoretical. Non-standard poses or compositions (ie; more creative stuff), tend to be under harsher scrutiny. The safest thing to post is still good ol (1girl, standing, rear view). There needs to be a bit more leniency given for uploads that are more complex in nature, I think this is already the case though. On paper there is a worry that quality standards towards technical perfection limit expression, I don't think this has been the case either.
How AI is judged is also just a side effect of how AI functions. AI itself just isn't very expressive. It tries to solve and complete. AI has to be judged on it's very nature. However as human involvement increases, so needs that leniency. This isn't always visible or evident. Janitors may make mistakes. From what I've seen, this has also not been an issue.

Now, I don't want to regurgitate the grievances other people have had with deletions over small shit. I don't think I'm in the position to have a say on that anyway. I do want to emphasize though, that oftentimes the sentiments by those who are very vocal, is that of entitlement. E6ai does not owe you serverspace. I see a profound disinterest in those who upload with fat attempted signatures, that could be edited out in a minute. I've seen people run the wrong hires.fix settings so their fingers melded together (understandable, you give them advice, they fix it, the world improved. This is a positive example). And I've seen ones that outright refused to fix a hand for 3 minutes in an image editor because "It's just AI". In any case, You are the one hitting the upload button. You do not have to upload. Maybe you need to have higher standards for yourself. The recent uploads of my stuff by technical-grid are all things I did not upload because they don't pass my own quality standard. And no, just because your Loona got a bazillion upvotes doesn't put it above scrutiny.

AI is weird socially. I don't mean that in the wider art space, I mean that in isolation for the AI 'community'. People treat it very impersonal. I remember the first time I had a "yea this looks kinda shit" on a pic I actually sat down several hours for. People generally wouldn't say this on a traditional art pic, because the premise of amateurism gives you the social cue not to blurt that out, I don't think I need to explain why this is asshole behaviour.
Ironically the more you do yourself, the more errors occur. AI is good enough to not make major mistakes anymore. If I sketch something first, I basically hinder the AI. It craves (1girl, standing), I force it to do (1girl, handstand). AI can be deeply unsatisfying if you need external gratification (just like traditional art btw!). The stuff that 'does best' is predictable. AI is predisposed towards the stuff that did best in traditional art after all. But know there are people who will see and appreciate effort. I mention this because (just like traditional artists) you have to shield your ego a little. A deletion should never be a reason for emotional distress. i feel like people treat AI like some competitive game sometimes. They see the deletion and go straight to the janitors own uploads. "If you can't do better, who are you to tell me." There is some truth to this yes, but it's not like you can solve this issue. Imagine the dystopia of only the 0.1% being Jannies. Take a deep breath.

I encourage everyone to be a little nicer. Yes, some people hit generate and then press upload. Not everyone though. The same goes for the treatment of the Jannies. Do they often speak from authority even though their own works suffer from the same problems? Yes. Do some act nice in public then gloat behind doors? Some maybe. Do they do it for free? Absolutely. Have these things gotten better? Yes. Except the last one. They still do it for free.
Scrutiny should be where it counts.

The next is just some food for thought.
Ai has advanced at breakneck speed. A lot of the issues are that of communication and rapid changes. But some are also more complicated.
If I go back and look at how #90021's story shaped out, they returned to Civit. Lots of variation seed posting. In retrospect I find many of their pics uploaded here impressive, by merit of having been done on Civit. Civit gives you far fewer tools for high quality. They did cherrypick the best of the bunch for e6ai.
People are getting gated out of quality if they don't have a strong GPU to local gen. You can't just 'pick up that pen' with AI. This stuff can be complex. I think there are problems we just don't have good answers for. Do we want to have high quality standards even if it means gating people out? Do we want to allow everything and flood the site like rule34 did? I would like users to be aware that no one's acting in bad faith here. E6ai is compromising. But if I were to judge their volume of posts on Civit, as what they would ideally all upload? I want to say this user is ultimately better served on Civit.

TL:DR
Understand the position of the Moderation and the Janitors. Mistakes happen. But you the uploader also make mistakes. This is not a personal attack. These are efforts with good intentions.
Have higher standards for your own uploads. View this as an opportunity to participate in a space that does not treat everything as disposable slop.
If you are incapable of taking criticism then please continue to write rants, they are VERY funny.
And everyone, be just a little nicer to each other.

Got two of my pictures deleted. Didn’t care that much, although I feel like it definitely made me more paranoid about what I upload. Since then, I’ve decided not to upload some pics because I wasn’t sure if they would stay up or not (i can't draw and things that are looking alright for me might simply be incorrect, especially perspective and proportions).
I’ve also uploaded pictures that, by the classic Does not meet minimum quality standards (Anatomical) rule, should have been removed. Yet, they’re still up. Sometimes I noticed the issue too late; sometimes, I just didn’t feel like fixing it.

In my opinion, people (coomers) come here to get a fix for their kinks. Take realistic pictures, for example—not a lot of people draw them, but plenty generate them. And when people are starved for decent content, they lower their standards. I don’t blame them; I walked that path myself 5–10 years ago with a niche kink. There’s just not enough content, and at some point, you’ve seen it all.

Why do I mention coomers and lowered bars? Because they are the core audience for AI-generated images. There aren’t many people producing high-quality AI art (I’m probably not one of them). Even fewer are doing something unique and developing their own style (I’m definitely not one of them). Most of the time, you can guess the model just by looking at the picture (notice how i've replaced art with picture here).

One time, I spent way too much effort fixing the background and perspective. I was proud of my fixes and posted the edited image below the original. I got one comment that still sticks with me: "Who cares? No one is looking there."
Fine. Later on, I spent time setting up filters, adding chromatic aberrations, color-correcting images with LUTs, and playing with backgrounds. I posted two comparison shots. One person commented: "What’s changed? I can’t see a difference."

That kinda changed my perspective. I still can’t imagine posting a raw image without manual edits, but now I understand that most people who view my images simply do not care. Effort is not a guarantee of success. If you spend five hours cooking dinner and it turns out shitty, no one is going to eat it just because you worked hard on it. And if a raw Loona with a third eye gets more likes than my Loona that I spent five hours on, why should the raw one be removed?

I feel like people collectively should decide whether a picture stays or not. Let unapproved pictures "marinate" for a period of time—say, 12–24 hours. If an image reaches some arbitrary score in that period, it stays, regardless of quality or anatomical issues. If it doesn’t, then the mod team can decide whether it stays or not (if it has the mentioned issues).

I’ve had my share of images that were pretty much ignored compared to others: post #91034 post #89825 post #62285 post #47411. So, it’s not like I’m some elitist whose images always get high scores. But if people like an image and upvote it, why should the mod team be able to simply take it away from those who enjoyed it? It’s getting ridiculous even from a viewer’s standpoint—you basically have to save every image you like, just in case it gets deleted later, even if it was previously approved.

Viewers' opinions should take priority over what the mod team thinks—unless there are clear rule violations (e.g., CP).

I assure you, the mod team are the only ones counting how many claws Loona has on her hand while there's a huge ass taking up 75% of the screen.
Either we give the mod team full power and stop these discussions (making this a website for mods), or we give power back to the viewers (making this a website for coomers). I don’t think there’s an in-between.

Also, I’ve heard the common argument that "Loonas will always get a high score."
First of all, that’s not true. My first removal was also my first takedown—a picture of Loona (which was approved at the time) that got so few likes I couldn’t bear to keep it up and deleted it myself. (https://e6ai.net/takedowns/56)
Second, as I said before, effort is not a guarantee of success. If your catgirl can’t compete with Loona in terms of score, that’s a you problem. It just means people aren’t interested in what you made (post #47411 again, took a lot of time actually). And it doesn’t matter if your catgirl is a perfect combination of anatomy and upscaling techniques while Loona has a third eye.

This website already has a reputation as an "AI slop dump." Overly harsh rules aren’t changing that because that reputation isn’t just an organic opinion—it’s being pushed by interested parties. Some traditional artists wouldn’t even be able to pass E6AI’s anatomy and quality rules. I’ve already mentioned milesdf once as an example. There’s simply no way most of his recent images would stay up here. That’s a fact.
So, how is it that a "slop dump" has stricter rules than the original site?

The arbitrary approval system certainly causes some serious issues, but is it widely disliked by the userbase, or is it just a vocal minority? If there was an alternative website with much more lenient content policy (like Furbooru for AI images), would a lot of users choose to emigrate from e6ai?

ayokeito said:
[...]
Viewers' opinions should take priority over what the mod team thinks—unless there are clear rule violations (e.g., CP).
[...]

I disagree. The rules should apply to all posts in the same way no matter the score, that’s the only fair way to handle it.
The voting system favors very few characters, concepts and styles from the furry mainstream. The mainstream posts already have a big advantage as it is now and if the approval would be linked to the votes we would lose even more diversity.
How do you explain to somebody who worked for hours on an image that shows a rare kink or species getting deleted because “too low score”, while 30 seconds effort rawgen Loona getting approved?
If that person would be angry I would understand that 100%. On the other hand I completely fail to understand that there are users who spend more time arguing over deletion than they actually spent on making the image...
Also the argument “I have to save every image before it gets deleted”, doesn’t seem valid to me. It would be if e6ai would the one and only site for hosting AI art, but it isn’t.

You can use Civitai for example as an archive for everything, every rawgen, every experiment. Civitai has zero quality control.
And you can even upload 20 images in one batch there, you don’t have to tag, a vision-LLM is doing it for you. That should be enough even for the most active genners.
And personally I consider e6ai rather as some kind of showcase for the best gens.

But I can understand that the quality standards are a problem for some people and keep some users from being able to participate.
My idea to resolve this: Be a little more lenient with quality, but implement a hard upload limit of 3 images/day per user. That would motivate people to only upload their best gens and keep the showcase character of the site.
The equation with the increasing/decreasing limit may make sense for e621 but imo it’s not good for e6ai. If I post 10 rawgens everyday and 8 get approved my limit will increase and increase. So the limit as it is now favors “quantity over quality”.

ayokeito said:
Some traditional artists wouldn’t even be able to pass E6AI’s anatomy and quality rules.

That's what I meant by "Ai is being judged by it's own merit". Most artists wouldn't pass. I'm not saying this is right, it's just one of the few semi-objective criteria in which you can sift out AI submissions. The quality control standards towards anatomy are primarily a filter. I don't think it's bold to claim that striking a pic over a 6th finger on which you can see 3 pixels of is beyond nitpicky. It's them going 'look over your shit'. You could mandate that there shouldn't be attempted signatures. It would accomplish the same effect. I just don't think the janitors have really internalized that this is what it's actually about yet, so removals are often arbitrary and feel like stop and frisk. Occasionally it feels like they are going out of their way to find errors because "that's their job". Good cops don't seek out crime, that sorta thing.
It sucks. It's not ideal. I have far more issues with default model throwaway output that is 'correct', than I do the reverse. But it's also not like you can or should interfere there, far too subjective. My only actual fear is that this would nonetheless get encouraged, given the current standards. Even safer and more boring content. But like I mentioned, the people who actually bother to do fun shit tend to pass just fine.
On the topic of pics doing numbers, you should stop caring or actually go full force into it. This is the exact same problem traditional artists face by the way. People LOVE 1girl standing. People love porny stuff more than SFW. Some of the most successful artists are producing the definition of content slop. It's up to you if you wanna chase numbers or do your thing, but you can't have both and expect to succeed. Just how it is.
I value the feedback by people who I care about in return. Why do you care what xXLoonaGoonaXx thinks. If this site wasn't just a "gooner content-dump" then we would still see NSFW stuff succeed more, it's just how people are. My SFW stuff does by far the lowest numbers and I'm not surprised, because I shouldn't be.
I totally get your issues but ngl, I feel they mostly have little to do with the site and more to do with the woes that artists also have. If you wanna be a doomer about what you do, and submit to the fact that your ultimate mission is to serve material for the coomers and e6ai is your chapel? Then sure. You care way too much about numbers.
>Loona
Your 10 highest rated pics have Loona 5 times. I have 0 clue what you mean here. The Loona rule never fails.

silvicultor said:
My idea to resolve this: Be a little more lenient with quality, but implement a hard upload limit of 3 images/day per user.

I have suggested this a couple of times in the past but it doesn't really solve anything tangible. You would prevent variation seed spam, which no one gets upset about if removed anyway.

tyto4tme4l said:
The arbitrary approval system certainly causes some serious issues, but is it widely disliked by the userbase, or is it just a vocal minority? If there was an alternative website with much more lenient content policy (like Furbooru for AI images), would a lot of users choose to emigrate from e6ai?

Communication has gotten a little better so it's not as arbitrary anymore, which is a good thing. A lot of the dislike for the system stemmed from people who uploaded prior to the quality changes too, so they suddenly got stuff deleted that would not get deleted before. And of course there's also just people are just kinda mad. How some of those haven't been banned yet is beyond me.
If an alternative site with more lenient upload rules existed, a lot of this would solve itself by people naturally drifting there if they got issues imo. I would certainly rather be here than be buried under 20 variation seeds. Rule34 found this out the hard way, they celebrated their 1 millionth AI pic submission by finally blacklisting it for users by default.

fluffscaler said:
[...]
I have suggested this a couple of times in the past but it doesn't really solve anything tangible. You would prevent variation seed spam, which no one gets upset about if removed anyway.
[...]

Well, the work that staff members have to do (removing the seed spam) seems very tangible.
But probably even more important would be the psychological effect of a hard upload limit:
You can’t throw in 10 rawgens everyday and get rewarded for it. (increased limit + potential upvote)
Instead they will start to think: “Which gen represents my work best?”
And that mindset is what we need if e6ai is supposed to stay in the “showcase mode”. The alternative is the Civitai/R34 mode...
Of course there could be exceptions from the hard limit for users who want to bulk upload a bunch of high quality works from an other site. But that should always require staff permission.

silvicultor said:
Well, the work that staff members have to do (removing the seed spam) seems very tangible.
But probably even more important would be the psychological effect of a hard upload limit:
You can’t throw in 10 rawgens everyday and get rewarded for it. (increased limit + potential upvote)
Instead they will start to think: “Which gen represents my work best?”
And that mindset is what we need if e6ai is supposed to stay in the “showcase mode”. The alternative is the Civitai/R34 mode...
Of course there could be exceptions from the hard limit for users who want to bulk upload a bunch of high quality works from an other site. But that should always require staff permission.

I think a hard limit and slighly relaxed quality standards would be a good compromise to the current, very strict guidelines. I wanted to try making videos using the recently published public models for that, because it's an interesting challenge, but there's always some very small defect appearing during the few seconds of video, making showing results here is virtually impossible, and it can't really be corrected. But since it's videos, fingers being fused for half a second is barely noticeable at all...

We all want to avoid the situation of the gallery being flooded with broken and low-quality pictures, but the situation with videos is different, it takes 100x more time to generate one, meaning you'd get one or two worth being posted per day. A hard upload number limit instead of 100% absence of defects would allow this new form of generative creativity to exist instead of being suffocated under standards of perfection which were designed for editable and inpaintable pictures, I think it's really worth considering it.

fluffscaler said:
>Loona
Your 10 highest rated pics have Loona 5 times. I have 0 clue what you mean here. The Loona rule never fails.

Well, it's 9 if you sample 33 pics, and the ratio worsens for Loona the more pictures you add. It's not about the top-rated pics; it's about the overall score. A Loona post is not guaranteed to succeed.

fluffscaler said:
I totally get your issues but ngl, I feel they mostly have little to do with the site and more to do with the woes that artists also have. If you wanna be a doomer about what you do, and submit to the fact that your ultimate mission is to serve material for the coomers and e6ai is your chapel? Then sure. You care way too much about numbers.

I don't really have issues. I seemingly have no problems passing the quality checks. And I can stop posting to E6AI, and nothing will change for me. I treat it as a storage space for original-quality pics without compression for the few people who care about that (almost no one does, but I do). I could easily self-host, but I don't see the need. I doubt anyone would notice or complain if I stopped posting here. That's just the nature of this website and the AI gen scene in general—no one cares about individual uploaders (or "creators," if you will). In traditional art spaces, people rush to upload pictures made by their favorite artists. On E6AI, I’d guess 90% of users don’t even know the names of the uploaders. They’re just doomscrolling the feed for something they like. One of my AI-bros, lumidetsu, makes images of (subjectively) very high quality with a lot of effort. He has 13.5K followers on X—more than most of us. Yet, he’s not uploading his pictures here, and no one cares. I see realistic cookies on feral animals (in all honesty, i couldn't tell if they are raws or not) routinely getting more likes on this website compared to his pics when they are uploaded here. But I’m not judging—who am I to eww someone’s yum?

I once had my post deleted because it supposedly had an extra finger. It didn't in the picture, but mod felt like it did. https://x.com/artifluff/status/1871813129254146417
That honestly said everything that i needed to know. If we apply these rules to E621, 90% of the posts are now gone. But i'm not complaining, i'm just saying that i completely understand people who are complaining. Because some of these things you just won't see in advance unless you know how to actually draw (perspective, anatomy, etc.)

As for the numbers, they’re the only objective way to judge things. Otherwise, we end up evaluating images on a subjective scale, trying to determine if someone put in "enough" effort. Neither of these factors matter to the audience of this website. Just look at the E621 forums—people openly describe the site as a dump. Strict rules won't change that.

silvicultor said:
The voting system favors very few characters, concepts and styles from the furry mainstream. The mainstream posts already have a big advantage as it is now and if the approval would be linked to the votes we would lose even more diversity.

Circling back to what I said before: who are you to decide whether popular concepts, characters, and styles are worth less than obscure ones? Why the "popular = bad" mentality? This isn’t a website dedicated to obscure characters doing handstands. If people want 1girl, standing, ass focus, why should your opinion matter more than theirs? We need to consider who this website is actually for.

silvicultor said:
How do you explain to somebody who worked for hours on an image that shows a rare kink or species getting deleted because “too low score”, while 30 seconds effort rawgen Loona getting approved?

I feel like you didn’t read my post. I've worked on images featuring more than just 1girl, standing plenty of times. The fact that I spent 3+ hours on a picture doesn’t mean it deserves a high score. Your work isn’t graded based on effort or time spent. Expecting others to judge images that way is unrealistic. I'll even say that effort doesn't matter ANYWHERE, in any field of life. It's the result that matters. There's a reason why people doing literally any paid work are paid for the result, not for how much effort they've put in. If someone's raws are better (= more popular) than my images that I've spent 3 hours on, that's my problem, not theirs.
In my experience, the time spent on an image doesn’t matter. Maybe a couple of AI nerds (not ART nerds!) will appreciate the details, but they’re a tiny, quiet minority. In three years of making SD pictures, I’ve only been "correctly" complimented (by another creator) once on a piece I spent a lot of time on.
If you want this to be a gallery for ultra-high-effort work, then open the platform to more genres and make quality rules even stricter. It’d basically be Civtai in reverse. But don’t expect a big audience—most people consuming AI images aren’t here for "quality" or effort.
Once, again, I'll reiterate saying that this is a dump, a ghetto for pictures that will never make it to "normal" E621 regardless of how much time or effort you've spent. No one cares. Because in the eyes of people who are making decisions, your effort is simply not good enough and amounts to nothing.
On a side note, nothing in AI space entertains me more than a post that "took off", with comments full of people realizing they actually liked AI art and getting angry about it. It's priceless.

silvicultor said:
How do you explain to somebody who worked for hours on an image that shows a rare kink or species getting deleted because “too low score”, while 30 seconds effort rawgen Loona getting approved?

You misread my suggestion. I’m not saying low-score posts should be automatically deleted. I’m suggesting that posts with a high enough score should be automatically approved unless they break objective rules—not based on subjective quality judgments. Which is the problem for people that are currently complaining on the forums.

silvicultor said:
My idea to resolve this: Be a little more lenient with quality, but implement a hard upload limit of 3 images/day per user.

Make it one post per day but allow unused slots to accumulate (up to a cap) for users who don’t post daily. Remove all subjective takedowns and "quality rules." That could work.

b1techienne said:
I wanted to try making videos using the recently published public models for that, because it's an interesting challenge, but there's always some very small defect appearing during the few seconds of video, making showing results here is virtually impossible, and it can't really be corrected. But since it's videos, fingers being fused for half a second is barely noticeable at all...

Videos are much more easily approved and don't have to be perfect. It's not in the rules AFAIK, but it's been confirmed on forums. Still, as per the main issue with E6AI, it's 100% subjective if your post will be deemed good enough or not.

ayokeito said:
[...]
Circling back to what I said before: who are you to decide whether popular concepts, characters, and styles are worth less than obscure ones? Why the "popular = bad" mentality? This isn’t a website dedicated to obscure characters doing handstands. If people want 1girl, standing, ass focus, why should your opinion matter more than theirs? We need to consider who this website is actually for.

I’m not saying popular is bad, only that stuff that is already popular doesn’t need a system that will favor it even more. Give diversity a chance, that my opinion!
But you’re right, it’s not for me to decide what is the purpose of the site. But it’s not for you to decide either.
The people who pay the server have the right to decide. But I have the right to decide if I use the site or not. And if a system is implemented that actively discriminates rare furry species, I’m out of here, very simple.

No, I didn’t misread your suggestion. But I don’t see any reason why the very few standard concepts that already are flooding the site, should gain an unfair advantage over other less known concepts.

And the logic “only the result counts, not the effort” is maybe true for manufacturing real life products. But only to some extent. There are people who care about HOW things were made. And for art this aspect was always important.

We have to consider what the site is really for, you say? So what is it for? Goon slop?!

silvicultor said:
I’m not saying popular is bad, only that stuff that is already popular doesn’t need a system that will favor it even more. Give diversity a chance, that my opinion!

Diversity still has a chance. If goon slop isn't removed, that doesn't mean obscure content has to be deleted.

silvicultor said:
But you’re right, it’s not for me to decide what is the purpose of the site. But it’s not for you to decide either.

I'm not deciding anything. But I suggest that someone in charge finally does.

silvicultor said:
No, I didn’t misread your suggestion. But I don’t see any reason why the very few standard concepts that already are flooding the site, should gain an unfair advantage over other less known concepts.

There is no unfair advantage over lesser-known concepts. No one is suppressing them.

silvicultor said:
And the logic “only the result counts, not the effort” is maybe true for manufacturing real life products. But only to some extent. There are people who care about HOW things were made. And for art this aspect was always important.

Well, then I have bad news for you—because everything on this website is made with AI. That already removes any sense of effort for most viewers. No one is going to believe or care that I put effort into generating an image. I know I did, you might know I did, but almost no one else will care.

Speaking of which, there have been news reports about people winning contests with AI-generated images. I can't recall any of those images being posted on E6AI.

And don't even think about posting your image—the one you spent 10 hours inpainting—on a mainstream website like E621 or FA. Best case? You'll be ridiculed. So much for effort.

silvicultor said:
We have to consider what the site is really for, you say? So what is it for? Goon slop?!

Oh, don’t ask me. Let’s just look at the website’s stats ( https://e6ai.net/stats ):

Safe posts 8,441 9%
Questionable posts 11,651 13%
Explicit posts 72,150 78%

Yeah. Yeah, I’d say it’s for goon slop.

ayokeito said:
[...]
There is no unfair advantage over lesser-known concepts. No one is suppressing them.

Of course they are being suppressed by such a rule. If Loonas get approved no matter the quality, you will see nothing else on the site. Other concepts/characters/species will then have no realistic chance to ever get more popular. Maybe there are many people out there who would love the (yet) unknown concept. But they will never realize that because they will never see anything else but Loona rawgen.

The rule you are suggesting would be like:
New Law: Every millionaire gets 1000 € for nothing from the government. You’re not a millionaire yet? Too bad, you get nothing!
That’s not unfair, no of course not!

I don’t expect thousands of likes or followers or whatever for my work. I only gen what I like personally, so it’s not a real loss if people ignore it. But I expect one thing: That we all are being treated equally here and that the same rules apply to everyone. I don’t think that’s too much to ask.

Well, then I have bad news for you—because everything on this website is made with AI. That already removes any sense of effort for most viewers.

Yeah, true. There are many people who hate AI art and these people of course also refuse to believe that AI (assisted) art can be art at all and that real humans put real effort into it.
But just because it is like that now it doesn’t mean it will stay like that. The use of AI tools will become more and more common.
In 5 years every image editor will have AI features.
Then the “real artist” will have to say: “Hey, support me I guarantee I made this 100% without AI! Please appreciate it!”

Yeah. Yeah, I’d say it’s for goon slop.

You are making it too easy. Not every explicit post is goon slop.
Because yes, even pornography can be art.

silvicultor said:
Of course they are being suppressed by such a rule. If Loonas get approved no matter the quality, you will see nothing else on the site.

Look, there are plenty of Loonas with low scores. You can just search for loona_(helluva_boss) order:score_asc and see for yourself. It's not like there's some Loona shadow lobby that upvotes every Loona picture. I've personally made pictures of her that flopped. Hard.

It's easier to score higher with well-known characters—this is completely normal. It doesn’t suppress more niche characters. Look at Fluffscaler, who consistently pulls higher scores on unnamed kobolds, lizards, etc., compared to my Loonas. He has only 10 pictures below 100 favorites, whereas I have 126—seven of them being Loonas. Being a popular character didn’t save them, and they (deservedly) still rank lower than literally anything that Fluffscaler uploads.
People are fine with any character on screen, as long as the quality is good enough.

If you want a truly unbiased voting system, just remove public favorite and score numbers. Make them private for the uploader. That would level the playing field.

If anything, allowing obscure content to remain up when it reaches high scores will improve its visibility. Why? Because it's difficult to make a handstand, to re-create a new character with no LORAs, or to generate a Velari without a LORA—it’s easy to mess up. So, any minor issue, like extra claws or a slightly off pose due to some proportion rule, could lead to the picture being deleted. But if enough people like the result, the system I propose would actually keep the post up, even if it has a watermark somewhere.

silvicultor said:
I don’t expect thousands of likes or followers or whatever for my work. I only gen what I like personally, so it’s not a real loss if people ignore it.

That doesn’t make a lot of sense because you're sharing your work. If you truly didn’t care whether people liked or ignored it, you wouldn’t share it at all. We’re all affected by the numbers we get in some way or another.

silvicultor said:
In 5 years every image editor will have AI features.
Then the “real artist” will have to say: “Hey, support me I guarantee I made this 100% without AI! Please appreciate it!”

Furry art, in general, suffers from a cult of personality. If an artist says AI is bad, their fans will parrot that argument. And they won’t care how much effort you've actually put in.

ayokeito said:
Look, there are plenty of Loonas with low scores. You can just search for loona_(helluva_boss) order:score_asc and see for yourself. It's not like there's some Loona shadow lobby that upvotes every Loona picture. I've personally made pictures of her that flopped. Hard.

Loona is just an example we all know.
But why you don’t want to apply the same rules to all the posts. Real life laws also work that way (ideally), they should be applied to everybody no matter who they are or if they are popular or not. That is the one and only fair method.
But I can acknowledge that your suggestion would be a little bit less unfair if the votes would be invisible.

That doesn’t make a lot of sense because you're sharing your work. If you truly didn’t care whether people liked or ignored it, you wouldn’t share it at all. We’re all affected by the numbers we get in some way or another.

I won’t claim that the numbers don’t affect me at all. But maybe they are affecting you a little bit too much if you can’t think of another reason to share something.
Like I said I only gen what I like. And I share (some) of it in good faith that others might enjoy it, too. I don’t really want anything in return.
Except maybe that one thing that the same rules apply to me as for the people here who just post here to shill Patreon or Twitter for their financial benefit. (Those would benefit most from your new rule!)

Most of my stuff that's been deleted was honestly not that great or really did need to be fixed, but I have had a few deletions that felt silly.
I'd be lying if I said I'm not feeling a bit deterred from posting more here though. That's a me problem though, I guess.
On the bright side I guess it did give me a push to make my own little site to share my stuff.

When I see a new upload that is great but has an obvious mistake in it I feel sad and then confused when it gets approved anyway like half of the time.

silvicultor said:
So what is it for? Goon slop?!

'rating:safe' gives approximately 8k results
'~rating:explicit ~rating:questionable' gives approximately 78k results
8/(78+8) = 93% of images are sexual in nature (ignoring gore which is a tiny percentage)
🤔

silvicultor said:
You can use Civitai for example as an archive for everything, every rawgen, every experiment. Civitai has zero quality control.

The one time I tried to use Civitai for feral porn they blocked my post within like 30 minutes. Granted I haven't tried uploading there since then.

silvicultor said:
Loona is just an example we all know.

...and you can find images with low scores for every character you can imagine.

silvicultor said:
But why you don’t want to apply the same rules to all the posts. Real life laws also work that way (ideally), they should be applied to everybody no matter who they are or if they are popular or not. That is the one and only fair method.

I don't think I understand. I'm suggesting a simple rule for all posts, regardless of content: if it reaches (random number) 50 score in 24 hours, it stays—even if it has a watermark all over the image.

That means the content of the picture is valuable enough that even glaring issues didn’t matter, and people still liked it. It’s so fair that 1/4 of my images wouldn’t be automatically approved under this rule today (not 24 hours in). They would still need to be checked for issues, just like every other picture currently does.

Once again, this doesn’t mean that everything below 50 is removed. It only means that everything above 50 is not.
The only problem this rule aims to solve is preventing the subjective opinion of 1 (make it 5 if it's collective decision of mod team, idk, never bothered to learn how they decide what to remove and what to keep) person from outweighing the opinions of 50. We can tune these numbers until they feel fair. We can make it 75 score, then it would filter 75% of my pictures. But I'm not sure how realistic 75 in 24 hours is. Never looked into score dynamics. Or we can go down to 30, idk.

silvicultor said:
Except maybe that one thing that the same rules apply to me as for the people here who just post here to shill Patreon or Twitter for their financial benefit. (Those would benefit most from your new rule!)

There’s no financial benefit to shilling your Twitter, since it doesn’t allow monetization on porn accounts. You can't even pay to advertise your content, because if it's flagged, the button simply isn’t there.

As for Patreon, I wouldn’t know, but I’d guess that it’s not the best idea to upload something for free if you could put it behind a paywall. In any case, I can’t imagine people making money on Patreon posting frequently on E6AI—it kind of defeats the purpose.

'rating:safe' gives approximately 8k results
'~rating:explicit ~rating:questionable' gives approximately 78k results
8/(78+8) = 93% of images are sexual in nature (ignoring gore which is a tiny percentage

It's a self-feeding loop. The higher the numbers are for nsfw posts' score, the more directors feel discouraged to post swf here and bother diligently tagging everything since they know in advance their post will most likely get ignored. The less available sfw posts are present simply because of "scoring", the more it's gonna get treated like a coom site by most users, and its reputation will precede even more as just a place to look up porn rather than AI art/videos. And it keeps looping until further swf are automatically removed because of scoring and all you have left is nswf.

E6ai is a free service which has to pay for its own maintenance/storage space/self-host. So for me it makes sense to not accept everything considering the sheer volume of what can be produced by a single user in a single day, let alone every director that uploads here. Putting some real effort in a gen and then seeing it basically get ignored simply because a bunch of users decided to batch upload 20 variations of discutable quality of the same seed right after you and your single post just gets buried under the slop is just as disheartening feeling as seeing a post you cared about being taken down. In both cases you feel like you were wasting time taking the time to upload. That's why this site should not be treated as your own personal gallery but should rather compromise to make users as motivated and invested as they can be doing what they like and showing off the impressive stuff they can make with the quickly advancing tech.

For farming likes/favorites/reputation/praise, you can use any other social media sites you wish if that's what you seek for. I'm in favor of e6ai being a special place that differs from a website having next to no moderation effort and not being able to find any content you are searching for because uploaders will barely take the time to tag their posts properly and keep reusing the same pose/seed over and over again because it's "popular".

I guess I have to explain why auto-approval off of score is a terrible idea after all. If you have a large fanbase or are popular then you are immune from quality rules no matter what you do. Given that you yourself view problems with cults of personality in traditional art, I think you can see the pitfall. I get that your idea here is to give power to the people and not the few but this likely instantly backfires and just fosters an equivalent elitism, just from another direction.
The thread has sidetracked a little. Pretend I gave all of you very smart sounding answers to everything you said.
There's plenty of other issues I have voiced in the past as well and most of the responses have been dismissive at best. I'm not super happy with moderation/janitors either. There's reasons why I stopped uploading actively for a while and fucked off to Twitter. I'm mostly just trying to be a little diplomatic and give them a chance because I think the site still has potential. There's definitely an effort to make things more transparent like with https://e6ai.net/wiki_pages/753

There's definitely a lack of focus. But this thread has also lost focus. I've said my piece and I'll leave it at that.

Part of my issue is the amount of nitpickiness and that it is also subjective depending on which person does the Approving/Denying, As shown by the comparison in the first dropbox link.
https://www.dropbox.com/scl/fi/81sq7ylhtid5ncrcwbvtl/denyapprove.png?rlkey=2akbvpgd22wpvvhdf2h9rfnsa&st=3hfsh83b&dl=0

Now, i dont have a top of the line PC, so i use NovelAI, meaning i dont have LORAs or other tools, Just the Furry Diffusion and the NAI tools. I tend to fiddle with images and odd concepts till they hit the good enough phase, but even though they look 98% good, small things (imo) would likely cause it to get deleted. Like i am going to guess this gen of mine wouldn't pass because the the eyes/mouth, even though i don't think it is that noticeable unless you look for it.
(NSFW, Female Nudity, TF as an fyi)
https://www.dropbox.com/scl/fi/pwt0ybgiya2xr1exjpfa9/nsfw-explicit-sfw-detailed-background-high-detail-best-quality-so-s-2547809367fx.png?rlkey=jfqhpeay6f2ni245un4k0e31d&st=tpjbv60z&dl=0

Im guessing the strictness came about because of some people mass uploading very similar raw gens and such (which is annoying, pick your best attempt and refine it), though other's are getting hit in the crossfire.

fluffscaler said:
I guess I have to explain why auto-approval off of score is a terrible idea after all. If you have a large fanbase or are popular then you are immune from quality rules no matter what you do.

Can you please provide three examples (just to ensure a larger sample size)? I'd like to see an AI furry creator who is popular, has a fanbase, and consistently scores above 50, even with poor-quality images.

I believe we're currently inventing problems that don't exist yet.

ayokeito said:
[...]
I don't think I understand. I'm suggesting a simple rule for all posts, regardless of content: if it reaches (random number) 50 score in 24 hours, it stays—even if it has a watermark all over the image.
[...]

OK, then let me explain. Linking the approval to score simply breaks the principle of neutrality and equality.
Let me give you an IRL example for this: If I park my car, let’s say on a fire lane (or some similar emergency access road etc.) I will get a heavy fine for doing it, because it’s against the rules (and these rules exist for a good reason.)
Now imagine some celebrity doing it and they can just say “hey, I’m popular, these rules don’t apply to me!”. And they get away with it. Do you wanna live in such a society?

And for e6ai this would mean:
“Hey, I’m [POPULAR TWITTER USERNAME], I can upload here whatever I want, my dedicated followers will upvote just anything! Those silly rules only apply to you unknown, insignificant users!”

But you know what? Neither your suggestion (auto-approval) nor mine (hard limit) will ever be implemented. Because e6ai website code is linked to e621. And implementing one of these new rules would require substantial change of the e6ai website code, which would lead to incompatibility with e621, creating a huge amount of extra work for admins upon every update. So the chances are 99.9% that none of this will ever happen.

As for Patreon, I wouldn’t know

Well, I don’t know if it’s actually successful. But you can see there are posts like “Visit Patreon for nude version” or “AMAZING 50 IMAGE PACK ON PATREON!!!!” posts everyday. These people only post here to shill their Patreon, not for the community, nothing else.
And imo the purpose of e6ai should be that furry AI enthusiasts can share their work, and not a place for advertising.

And yes, everyone should be able to participate equally. That’s why the hard limit would be so beneficial.
And yes, the quality standards might keep some people from participating, too. So reviewing that again might also be a good idea, but only in combination with the limit (that will never exist), otherwise you would favor quantity over quality too much.
Some users who have been posting rawgens for a very long time have absurdly high upload limits. These users can fill a whole page with their posts within minutes (or even faster if they use API). Nobody can tell me that this is desirable, fair or beneficial for the e6ai community.

busahou said:
[...]
The one time I tried to use Civitai for feral porn they blocked my post within like 30 minutes. Granted I haven't tried uploading there since then.

Yeah, realistic feral is probably the only bad thing to upload, despite Civitai not stating in the rules that it is forbidden. Rules only say human on feral is forbidden, not feral furry as a whole. You can find some feral on the site however, if it’s not too realistic. Talked with a user about it who had some of their feral content removed. They said “pony” is OK “horse” is not. Pretty weird policy.
But Civitai has zero quality control. And aside from the feral thing they really allow almost everything. You can see many deepfake posts of real persons there for example...

scout said:
[...]
Putting some real effort in a gen and then seeing it basically get ignored simply because a bunch of users decided to batch upload 20 variations of discutable quality of the same seed right after you and your single post just gets buried under the slop is just as disheartening feeling as seeing a post you cared about being taken down. In both cases you feel like you were wasting time taking the time to upload.[...]

I wish I could UPVOTE your post somehow! There are very few people who are able to understand both sides.

fluffscaler said:
[...]
There's definitely a lack of focus. But this thread has also lost focus. I've said my piece and I'll leave it at that.

What is the intended focus of the thread? Quality control and how it affects the site and the community, right? We are still talking about that?!

ayokeito said:
Can you please provide three examples (just to ensure a larger sample size)? I'd like to see an AI furry creator who is popular, has a fanbase, and consistently scores above 50, even with poor-quality images.

I believe we're currently inventing problems that don't exist yet.

No, examples should not be provided. Nobody wants that we start pointing at certain users, that can’t end well.
But they do exist. Pay close attention to what gets uploaded within one day and you will notice absurd misalignment between score and quality of some posts, I assure you.

silvicultor said:
And for e6ai this would mean:
“Hey, I’m [POPULAR TWITTER USERNAME], I can upload here whatever I want, my dedicated followers will upvote just anything! Those silly rules only apply to you unknown, insignificant users!”

Look, I'm not here to brag, but I'm in the upper bracket of Twitter users by follower count (i'm at 9K now).

Just five uploads ago, I made a post that was completely ignored—post #91034.
post #91031 and post #91030 were also largely neglected (7 and 8 uploads ago).

I stand firm in my position that there are NO "popular AI artists" in the furry genre at this time.
If any one of us stops posting, no one will care. There’s no competition rushing to upload our pictures, and there aren’t people lining up to order commissions for $10, let alone $2,000.
And there are certainly no fanatic followers going around the internet upvoting our posts or defending our content against the anti-AI crowd.

A few days ago, I even made a likes/reposts goal post, and the numbers were laughably low—even for a new account with fewer than 1,000 followers.

There are no true fans of "AI artists". They don’t exist. They’re just a concept from the far future.

RE: "Batch upload 20 variations" hits too close to a similar situation where an indie dev published a game, and a few hours later, EA started a huge sale:
https://www.reddit.com/r/IndieDev/comments/1ba4d01/my_indie_game_i_worked_on_for_10_years_was/

This is random. You can't protect yourself from that. You can't expect anyone to protect you from that. If you make a post right now and I post 20 variation seeds, your post is gone—it's not seen.
Now, let's assume all 20 posts were low-quality garbage and got removed an hour later. Nothing changed. Your post is already too far down. The current rules on quality simply do not protect you from variation seed spam.
For you, it was a picture you worked on for five hours. For the slop generator, it was just Monday.

silvicultor said:
And implementing one of these new rules would require substantial change of the e6ai website code, which would lead to incompatibility with e621, creating a huge amount of extra work for admins upon every update.

Come on. I have no idea how e6ai code base looks. But i can bet you a 50$ i'll make a script that automatically approves posts that are above 50 score in, let's say, 3 hours.

class PostApproval < ApplicationRecord
  belongs_to :user
  belongs_to :post, inverse_of: :approvals

  validate :validate_approval

  def validate_approval
    post.lock!

    if post.is_status_locked?
      errors.add(:post, "is locked and cannot be approved")
    end

    if post.status == "active"
      errors.add(:post, "is already active and cannot be approved")
    end
  end

  def self.approve_high_score_posts
    system_user = User.find_by(username: "System") # Ensure this user exists

    return unless system_user

    Post.where("created_at >= ?", 24.hours.ago)
        .where("score > ?", 50)
        .where.not(status: "active")
        .find_each do |post|
      PostApproval.create!(post: post, user: system_user)
      post.update!(status: "active")
      puts "Approved post ##{post.id} (Score: #{post.score})"
    end
  end
end

namespace :posts do
  desc "Approve posts with a score greater than 50 within 24 hours"
  task approve_high_score: :environment do
    PostApproval.approve_high_score_posts
  end
end

every 1.hour do
  rake "posts:approve_high_score"
end

Anyways, nothing will happen since mod team has already went with making examples for what they consider bad quality. We'll see how it goes and if it will reduce the amount of whining. Pretty sure nothing will change.

ayokeito said:
[...]
Just five uploads ago, I made a post that was completely ignored—post #91034.
post #91031 and post #91030 were also largely neglected (7 and 8 uploads ago).
[...]

Huh? Ignored you say? No, these are safe posts, and for safe posts these scores are quiet high, not the highest yes, but certainly far from being ignored. A good deal of good quality safe posts only have a single digit score, which includes my own avatar btw.

There are no true fans of "AI artists". They don’t exist. They’re just a concept from the far future.

If you consider 5 years far future, probably it will take less time considering how fast things are changing...

This is random. You can't protect yourself from that.

It is like that now, but that doesn’t mean it can’t be changed. As I interpret the uploading guidelines as they are now, the purpose of e6ai is to be some kind of showcase for your best AI art. And the fact that some users have absurdly high upload limits due to prior posts and can fill whole pages with API within seconds, contradicts that purpose blatantly. That is a design flaw, not bad luck.

Come on. I have no idea how e6ai code base looks. But i can bet you a 50$ i'll make a script that automatically approves posts that are above 50 score in, let's say, 3 hours.

I’m not a developer. But I know running big website is more than just executing a random script written by LLM. I have now idea how that would affect compatibility with e621.
If it’s that easy, what are admins for? Yeah, let LLM run the site, would be funny experiment for sure.

silvicultor said:
And implementing one of these new rules would require substantial change of the e6ai website code, which would lead to incompatibility with e621, creating a huge amount of extra work for admins upon every update.

Not trying to say you're wrong, and I have no idea how to fix this issue in a way that everyone would like.
But I don't think it would be THAT hard to implement your proposed hard limit.

Go into the source code and check file app/models/user.rb

def upload_limit
      pieces = upload_limit_pieces
      base_upload_limit + (pieces[:approved] / 10) - (pieces[:deleted] / 4) - pieces[:pending]
end

There's the current code for calculating upload limit. Just remove addition on the third line and upload_limit should equal base_upload_limit.
base_upload_limit seems to be here and I think it's tied to each user rather than a global variable. I don't do much with databases but I think it would be trivial to lower it for new users and easy enough to do so for existing users.

Maybe it does run deeper than just these two files but still.

IMO the rules tend to have detrimental effect. Instead of high-effort, high-detail pictures, where you risk some tiny mistake getting through, and hours of your work going down the drain because a background character is missing one ear, I've noticed an uptick in uploads of pictures where hands are cropped or hidden behind back, pictures so simple there's very little that can go wrong, series of cookie-cutter poses with one prompt word swapped out to change the color, 'cause someone finally arrived at one working pic and they now want to milk the effect to fullest extent. The arbitrary "correctness" standard has chilled efforts to try to make pictures that are complex enough to have a potential to be incorrect.

And a lot of people have great ideas, but work with Purplesmart.ai dreambot, or Civitai's on-site generator because they can't run models good enough locally. And that means no inpainting, no correcting one detail while leaving other unmodified. That means the only thing you can do if you spot a mistake is hit 'generate' again, pay the buzz, and pray this time the mistake will be gone. That unique, one-in-a-million facial expression? Down the drain because six toes. Eventually you may download the result and do some manual editing in gfx program. But if I can get 250 buzz daily for free, I can choose to spend it:

- to generate five pics I really like personally, and I'm sure most of users here would like too. But they won't pass moderation due to minuscule flaws.
- maybe one that will pass the moderator scrutiny here, if I try really hard, spend all the free buzz I got and get lucky in the end.
- or I can generate a series of twenty pictures from which most will pass moderation, get 5 upvotes max each, and clutter the timeline for everyone, 'cause they are too simple to be flawed.

I know which option I prefer. I can create just for myself, stuff I like. I get nothing from posting here, I don't care about gaming the system just to have something on site. I'm just tired of this. And I guess I'm not alone in this feeling.

Sometimes in the afternoon I check this site, find five pics I love, so I add them to favs with intention to download at home in the evening. And in the evening four of them are gone. Some artists who'd been prolific here, hardly ever post anymore. But I'm sure I'll see another five pics of Loona with hands behind her back being fucked missionary style first person view, 'cause the lora's strong, the fingers are hidden, and the pic fulfills the site's requirements.

Is this really what the mods want to do with this site?

ayokeito said:
I stand firm in my position that there are NO "popular AI artists" in the furry genre at this time.
[...]
There are no true fans of "AI artists". They don’t exist. They’re just a concept from the far future.

An opinion I'll disagree with. Maybe you have some more demanding, more advanced definition of "true fan". Maybe for you a true fan is someone willing to order a $2000 commission, and a freeloader who only likes and occasionally comments is not a fan.

I'm sorry, but I'm totally a fan of Terraraptor, Hyperion, Thunderwolf, and the anon behind Invoke High. I love Corsetlover's deers, and Langri's Mirage.

Maybe in your eyes I'm not their true fan. But I would totally miss them if they are gone from the site.

sharpy said:
IMO the rules tend to have detrimental effect. [...]

Despite me being someone who has the luck to be able to gen locally, I can understand your concerns. It’s no fun if you feel excluded from the “party”.
That’s why I suggested a hard upload limit as quality control to address the issue with the many low-effort uploads. Then everyone can contribute something, local genners and Civitai users alike, but only the best from each.

I often spend hours refining one gen, but I don’t do it for the sake of e6ai or because some sort of “elitist attitude”, I do it because otherwise I can’t really enjoy my own work, it's only my own motivation, I can't expect everyone to do the same. Plus I somehow love and hate the process at the same time...

Also I agree that there are users who know exactly how to circumvent the “showcase principle”. Like you said they only gen lots of extremely simple stuff with hidden hands and so they can avoid to get in conflict with the quality standards, while it’s still just quantity over quality.
In the worst case it could lead to these uploads being the only ones on e6ai, because creators who want to make something more complex feel intimidated by the quality standards.

e6ai is still “young”, as AI art itself. So we probably haven’t found the right balance yet, that excludes nobody but also avoids low-effort spam, but I’m optimistic that it will happen.

i think it's unreasonable to expect user upvoting to be useful to match the needs of the site. it just replaces e6ai's de-jure quality standards with a worse de-facto quality standard. both result in the same situation—uploaders are incentivized to make uncreative porn to not have their upload limits reduced. e621's popular page is a poor representation of the creativity that traditional artists can display through their work, so why would that change for e6ai when the toolkits are even more limited?

e621 users are often not thinking creatively about the things they upload. they often don't have any real relationship with the image and may not be artists themselves. this leads to a good amount of selective uploading where only the pieces an uploader finds "hot" are put on e621. an artist I like often only has their solo female anthro -rating:safe art uploaded to e621—that should sound familiar to users here. users on that site are often not thinking about the artistic process at all.

e6ai has a much more tense relationship with moderation, because the most prolific users (not in terms of upvotes, but in terms of community engagement, comments, uploads, etc) on this site generate ai images themselves, so there's an investment in making "interesting" images. the janitors now need to be arbiters of taste as well as quality control. any moderation that stems from how e621 works needs to deal with this site's different relationship between its users and its moderation.

with that in mind, i agree that the upload limit system, which made sense for e621 but not e6ai, is the first thing to address. minimal quality flagging and a flat upload limit that doesn't change can allow uploaders the flexibility to make however many creatively safe or interesting posts they want per-day. no one has to worry about a negative upload amount or getting images removed/"curated" for arbitrary reasons, especially if there is no source to point to. janitors no longer need to be quality control, which in this context can feel more like arbitration of taste. the site would still be slow enough to make moderation possible.

at the same time, uploaders that create their own images should also develop thick enough skin to be inured against not getting too many upvotes on this site. many traditional artists that make good art don't end up on the popular page of e621.

Maybe it does run deeper than just these two files but still.

Your post limit still gets negatively affected for every flagged posts. - "pending or flagged posts"; whether it's flagged by you because it's the only way to replace your submission with a better quality one, or randomly flagged by someone else.

"Hard limit - pending/flagged posts" can cause I lot of potential issues I reckon, since anyone can flag a post. I don't know if it's "approved" flags, but even if it is, directors who flag their own posts to replace their own image still get penalized.

i think it's unreasonable to expect user upvoting to be useful to match the needs of the site. it just replaces e6ai's de-jure quality standards with a worse de-facto quality standard. both result in the same situation—uploaders are incentivized to make uncreative porn to not have their upload limits reduced. e621's popular page is a poor representation of the creativity that traditional artists can display through their work, so why would that change for e6ai when the toolkits are even more limited?

I agree. It's giving too much power to the regular users here to circumvent e6ai's own sets of rules. And over-exageration case is a bot attack. Let's say a regular artist that really hates AI art happens to learn that AI art can get deleted if they got a low-enough score? What's stopping them from creating a couple dozen accounts and mass-downvoting everything? Anyone can do it, you only need to have a valid unique email address and wait a week (or maybe not, it may just be for uploads that you have to wait). Users having a grudge against each other could also mass-downvote submission out of spite. And since votes are all anonymous, none would be the wiser.

Let's be real here. Saying that keeping ai content based on the submission's scores is more objective than a specific sets of reasons already established, documented in the uploading guidelines that some may not even be bothered to read before posting, have a clear reason stated upon deletion, and now even have a new wiki page stating what kind of error means what visually, is being delusional.

If you want to appeal a deleted submission, you know exactly which janitor to reach out for. They freely take the time to respond cordially to you, gives you educated hints on how to fix it (because they have experience with AI tech/art software) and will consult each other behind closed doors when they unsure themselves to make sure their decision isn't biased. If that's not being treated fairly, I don't know what is then.

Now, how would you even appeal on a deleted submission when it's deleted by score? Reach out to the whole community? The barely active forum on here? How long would that take? How can you make sure it's not a subjective decision again, when you don't even need to provide a reason for voting?

This is random. You can't protect yourself from that.

First of all, that is not up to you, but up to the site's policy in place. It's not because it's like that on other sites that you frequently use that it has to be like that on e6ai.

Second of all, I didn't ask to be "protected" from such behavior. My point was to speak in favor to stricter limits by providing a different point of view to break up the vocal majority of users wanting to critique the new set of rules. And it was related to the "being upset about the deletion of their own content to be point of mentioning suicide in a farewell post". If you either use e6ai as a dumping ground and expect everyone to respect you for it and not have some critique coming your way, or if you use e6ai to make big numbers and gain popularity on your high-effort gens and get upset that you get ignored to the point of getting heavily depressed or worse, you should not be uploading your own content on e6ai at all. It's true for any kind of content posting site; critique and moderation decision (or lack of moderation) will always be upsetting for some users. The only thing you can do to protect yourself against that is to not use their services.

That's why I will never use Twitter or Deviantart personally. I don't agree with their TOS, and I know that debating about it is not gonna change the way things are. So I move on. Some users are only very vocal about it because they can pinpoint the exact janitor that "passed judgement" on them. And then your first reaction is feeling personally attacked. I know, I've been there. But in the end, there's nothing personal about it... I feel bad for the janitor staff to have to handle all that negativity coming their way, but regardless if you agree with their decision or not, moderation is needed on any hosting website. E6ai is no different.

I kinda lost interest in the topic already, but I'll once again say that I wasn't saying that posts with low scores should be automatically deleted. I suggested that posts with high scores should be automatically approved and that's it.

ayokeito said:
I kinda lost interest in the topic already, but I'll once again say that I wasn't saying that posts with low scores should be automatically deleted. I suggested that posts with high scores should be automatically approved and that's it.

That's a flawed opinion. If a post with high scoring was approved and didn't have anything wrong with said post then that's understandable.

But if a post with high scoring had errors within in (Whether it's Anatomical Anomalies or Attempted Signature or anything similar to regards to the uploading guidelines.) it's likely to be deleted.

Scoring is based entirely on user's feedback and has no actual impact whether that post would be approved or not.

angry_puppy said:
But if a post with high scoring had errors within in (Whether it's Anatomical Anomalies or Attempted Signature or anything similar to regards to the uploading guidelines.) it's likely to be deleted.

IMO, if the post *despite* these flaws got massively upvoted, that objectively means these flaws weren't important enough to adversely affect enjoyment of the post for lots of people.

Isn't that very enjoyment the main purpose of the existence of this site?

For whose benefits do you enforce the "quality standard" rules? Some elite art critics? Opposition of the anti-AI crowd? Because deleting what a lot of people like is certainly not for the benefit of the people who liked it (unless you have some secret "long game" plan, "you'll be grateful later"?)

Normally the purpose of existence of rules is to best serve the user interest. Is removal of widely beloved content in best interest for the users? Who's getting actually hurt by a scratch vaguely resembling "attempted signature" in a corner an otherwise excellent image? Because a lot of people get disappointed by the image's removal, so it's better be some important reason why you'd rather disappoint them all.

(and no, it's not always about "my image got removed". I'm not feeling too strongly about my images being removed. But seeing "Deleted: 147" on Hyperion's userpage makes me fume. How many images I'd absolutely love to see got wiped out of the site before I could see them?)

angry_puppy said:
But if a post with high scoring had errors within in (Whether it's Anatomical Anomalies or Attempted Signature or anything similar to regards to the uploading guidelines.) it's likely to be deleted.

Yeah. This is exactly what people complaining on forums are talking about.
The opinion of a janitor or moderator is valued more than the opinion of any number of viewers.

angry_puppy said:
Scoring is based entirely on user's feedback and has no actual impact whether that post would be approved or not.

Which is kinda strange because some might assume this website is made for viewers, not just the moderation team.
It doesn’t matter if a post had 1 or 100 favorites—if you didn’t like it, it’s gone.
Once again, I don’t even care that much; I have three deleted posts, one of which I deleted myself. I’m just pointing out how strange this whole thing sounds.

sharpy said:
[...]
(unless you have some secret "long game" plan, "you'll be grateful later"?)
[...]

I assume the long term plan is that e6ai becomes an archive for high quality synthetic data for future furry model training. And this synthetic data should not only contain maintream content, otherwise future models will also be able to produce such content only...

And attempted signatures and broken hands would be like poison for the future training. Of course you could exclude images with attempted_signature tag and you could also establish a broken_anatomy tag, that would also be excluded. But that only works if tagging is done flawlessly, and we all know most uploaders don’t care much about tagging...

Also finding the right balance to serve this purpose AND allow as many users as possible to participate AND avoid low-effort spam is not easy.

silvicultor said:
I assume the long term plan is that e6ai becomes an archive for high quality synthetic data for future furry model training. And this synthetic data should not only contain maintream content, otherwise future models will also be able to produce such content only...

Soooo it's a kind of honeypot, give us training data and we'll let you see stuff others post? Not sure if I like it but okay.

Of course you could exclude images with attempted_signature tag and you could also establish a broken_anatomy tag, that would also be excluded. But that only works if tagging is done flawlessly, and we all know most uploaders don’t care much about tagging...

It would not take any more effort from the mod team than deleting takes currently. Not even a little bit. Maybe even add these to default blacklist so logged-in users need to enable them manually if the mods don't want to show that stuff to outsiders?

Also finding the right balance to serve this purpose AND allow as many users as possible to participate AND avoid low-effort spam is not easy.

Finding the perfect balance isn't. But currently the mods overcorrected so ridiculously far away from the "sweet spot" that making the situation much better than it is would take no effort at all. Not perfect. But much better.

silvicultor said:
And attempted signatures and broken hands would be like poison for the future training. Of course you could exclude images with attempted_signature tag and you could also establish a broken_anatomy tag, that would also be excluded. But that only works if tagging is done flawlessly, and we all know most uploaders don’t care much about tagging...

Wouldn't examples of various issues (properly tagged) be useful for training? They could be put in the negative prompt to hopefully enhance the quality of the output.

sharpy said:
Soooo it's a kind of honeypot, give us training data and we'll let you see stuff others post? Not sure if I like it but okay.

I don’t say they that e6ai team is doing it. But from what I’ve heard the team had the training data issue in mind when setting up the new rules.
You know how furry AI models were made? Someone took a “normal”, censored image gen model like base-SD1.5 or base-SDXL, scraped tons of images from e621 and finetuned it with these images. That’s how furry AI was born.
And these attempted signatures come from the real signatures from all the artists on e621, they are sort of an echo of these.
Future AI models will be much bigger than current ones, so “real data” won’t be enough to train them, because there isn’t enough (high quality) “real data”. So you need good synthetic data. And the better this synthetic data is the better the future furry AI models will be. We will all benefit from this, no matter if you use Civitai or gen locally.

sharpy said:
It would not take any more effort from the mod team than deleting takes currently. Not even a little bit. Maybe even add these to default blacklist so logged-in users need to enable them manually if the mods don't want to show that stuff to outsiders?

Adding it to blacklist isn’t really helping. Because the problem would be the posts that are not correctly tagged. We would need many extra tags for all the quality issues to solve the issue. And many users are not aware themselves what is wrong with their image. But 6 finger hands, attempted signatures and GAN artifacts are all poison to the training.
So we would need many tags like 6_fingers, GAN_artifacts, melting_fingers, double_knee etc. And all these tags would have to be applied correctly by the uploader, which wouldn’t happen. So instead the staff team would have to add these tags. But doing so would consume much more time than deleting the post. And remember, the janitors are doing all this in their free time, nobody is paying them for it!

tyto4tme4l said:
Wouldn't examples of various issues (properly tagged) be useful for training? They could be put in the negative prompt to hopefully enhance the quality of the output.

Not really. It is less problematic if the are tagged, but it’s better if they aren’t there in the first place. I know that from my own LORA training.

sharpy said:

Finding the perfect balance isn't. But currently the mods overcorrected so ridiculously far away from the "sweet spot" that making the situation much better than it is would take no effort at all. Not perfect. But much better.

I don’t dare to judge what is the right balance. But I can really understand someone feels frustrated because they can’t really participate. It would be really great if everyone could join the fun while avoiding the above-mentioned pitfalls. But I sadly don’t really know how to achieve this.

silvicultor said:
So we would need many tags like 6_fingers, GAN_artifacts, melting_fingers, double_knee etc. And all these tags would have to be applied correctly by the uploader, which wouldn’t happen. So instead the staff team would have to add these tags. But doing so would consume much more time than deleting the post. And remember, the janitors are doing all this in their free time, nobody is paying them for it!

staff already has to flag posts for quality issues, document what they think is wrong with the image, and then wait for possible reuploads of the same image. it would end up being the same amount of work—the difference is that users can then account for the quality issues themselves by looking into the tags. if we assume that staff perfectly flag posts, then we can also assume that staff can perfectly tag quality issues in posts. if you want to train on e6ai, you can filter out all the negative tags anyway and not even have them enter your training set. e6ai as a training set is already kind of moot given that the site has a huge grandfathered backlog of low-quality warped images.

this also assumes that only uploaders or staff set tags. on e621 non-uploading users often start tagging projects as they please. i dont think the idea of having "bad-content" tags is ridiculous at all, and i think i would prefer that system over what is in place now, especially with upload limits. most genai issues are easy to categorize—extra digits, face warping, eye warping.

e621 has a new page of posts every hour, according to re621. e6ai has a new page twice a day (albeit with a harsh removal policy). its not like the posting is too much for the users here to handle. ill probably start tagging warped posts myself

silvicultor said:
synthetic data

That's gonna take a loooong while and a significant breakthrough on the model side. Because currently all synthetic data is the poison.
Check out the Computerphile video on the subject. https://www.youtube.com/watch?v=dDUC-LqVrPU

silvicultor said:
Adding it to blacklist isn’t really helping. Because the problem would be the posts that are not correctly tagged.

I still fail to see how, for these stated purposes, that differs from the current situation with posts being deleted instead.

Is it really that much more work to add "too_many_fingers" tag, than to delete with reason "Bad anatomy (too many fingers)" for the mods? Especially that users would tag lots (most?) of the pics, reducing moderator's workload. Currently every deletion is done by the mods.

silvicultor said:
But doing so would consume much more time than deleting the post.

How? The tags for flaws don't need to be any more detailed than current delete reasons. Hell, even a single tag to cover every kind of flaw would improve the current situation.

silvicultor said:
Not really. It is less problematic if the are tagged, but it’s better if they aren’t there in the first place. I know that from my own LORA training.

If you download the entire site wholesale instead of leaving off whatever you don't want, then this is indeed a problem. If you don't bypass the blacklist on download, there's zero difference between what you downloaded vs what you'd download if the blacklisted were deleted instead.

xerxes_i said:
staff already has to flag posts for quality issues, document what they think is wrong with the image, and then wait for possible reuploads of the same image. it would end up being the same amount of work—the difference is that users can then account for the quality issues themselves by looking into the tags. if we assume that staff perfectly flag posts, then we can also assume that staff can perfectly tag quality issues in posts. if you want to train on e6ai, you can filter out all the negative tags anyway and not even have them enter your training set. e6ai as a training set is already kind of moot given that the site has a huge grandfathered backlog of low-quality warped images.
[...]

It could be a possible alternative to deletion, yes. But it’s not realistic to expect that it won’t results in more work for the staff. Setting all this up alone will require substantial work.
And considering how bad the tagging is generally on the site I have serious doubts that it will work in the end.
But I don’t say it’s impossible. I know for example TantabusAI (the AI-friendly derpibooru clone). They have a tag “generation errors”, that is blacklisted by default.

But in the end it’s for the staff to decide if they want to implement such a system or not.
Personally I don’t really know if it would be a real improvement, also from the user perspective. Because if your post has a tag that is blacklisted by default, you are basically “shadow-banned”, 99% of the users won’t see your post anyway, so it’s not so different from being deleted.
Who would remove “generation errors”-type tags from the default backlist? Would you? I certainly wouldn’t.

And for the grandfathered content: Yes, if you collect training data here you would want to exclude everything from the SD1.5 era. But that can be done (more or less) by ignoring all content that was posted before a certain date.

sharpy said:
That's gonna take a loooong while and a significant breakthrough on the model side. Because currently all synthetic data is the poison.

That is only true if you take raw outputs from the model and use them again for training. Yes, then every following model-generation will be worse. But hand-picked and refined high quality synthetic data is as good as “real data”, or even better.
I trained LORA on 100% synthetic data with zero quality degradation. (if you train SDXL on flux outputs it can even improve the quality!)

sharpy said:
If you download the entire site wholesale instead of leaving off whatever you don't want, then this is indeed a problem. If you don't bypass the blacklist on download, there's zero difference between what you downloaded vs what you'd download if the blacklisted were deleted instead.

If you only train LORA and just need 30 images it’s OK, you can just check every image yourself. But if you wanna finetune with 100k images it’s obvious you can’t do that, you just scrape everything from the site. Than you have to trust the tagging of the site more or less. And let’s face it the tagging is pretty bad. And adding the “generation error”-type tags would most likely make it worse. But I don’t say it can’t be done (see above).

I just spun up two sets—set:warpedhands and set:warpedface—where grandfathered posts with common generation errors can be organized before tags are applied. if you want to be a maintainer let me know. you can also make your own sets

i think usage of synthetic data from this site is necessarily imprecise. you're just depending on the moderators to catch everything as opposed to the moderators + users. based on the posts already in my sets, both recent and older posts have these generation errors that slipped through the cracks. it'll be a problem regardless of the start date you specify.

i don't think its a foregone conclusion that generation errors would be part of the default blacklist. ive marked posts in my favorites as having generation errors. its just part of life, and as users are exposed to more images their eye will be better trained for generation errors—or at least I think that will be the case.

Is it really that much more work to add "too_many_fingers" tag, than to delete with reason "Bad anatomy (too many fingers)" for the mods? Especially that users would tag lots (most?) of the pics, reducing moderator's workload. Currently every deletion is done by the mods.

The answer to that is yes. A handful of janitors have to look at and approve roughly 950 posts per week (as of this week. This number will keep growing). There are less than 10 janitors while there are currently 2325 director tags (that will also keep growing). And it is not factoring the annonymous_director or posts without any director tags.

It was never the janitors' responsibility to curate user tags. Just like it isn't on e621 either. Because it's a monumental task. That's why any user without a privileged account can edit any post's tags.

There is so many anatomy errors that would need to be described using tags for it to be useful and we would need thousands for each one... No one here wants to see a few hundreds two_anus tag or merging_limbs or poorly_detailed_eyes tags. No one wants to tag them, no one wants to debate them, and anyone can play tag war with each other to remove those "negative tags" from their own personal posts right after.

Considering most of the less well tagged submissions are also posted by the ones that don't feel like fixing anatomy is worth the effort, or don't even see that there's any anatomical errors in the first place, no they wouldn't be tagged properly by the uploaders.

Most of the useful negative tags to fix broken anatomy comes from embeddings anyway, which is seperate from the model's training data. Manually curated by humans and tested, because anatomical errors are too complex to be described by tags alone, and even then it's a hit or miss if those embeddings do something useful in the end. How do you explain the concept of "impossible_geometry" to an AI. The answer is, you don't, apart from omitting to feed it training data with impossible geometry in it.

sharpy said:
IMO, if the post *despite* these flaws got massively upvoted, that objectively means these flaws weren't important enough to adversely affect enjoyment of the post for lots of people.

Isn't that very enjoyment the main purpose of the existence of this site?

For whose benefits do you enforce the "quality standard" rules? Some elite art critics? Opposition of the anti-AI crowd? Because deleting what a lot of people like is certainly not for the benefit of the people who liked it (unless you have some secret "long game" plan, "you'll be grateful later"?)

Normally the purpose of existence of rules is to best serve the user interest. Is removal of widely beloved content in best interest for the users? Who's getting actually hurt by a scratch vaguely resembling "attempted signature" in a corner an otherwise excellent image? Because a lot of people get disappointed by the image's removal, so it's better be some important reason why you'd rather disappoint them all.

(and no, it's not always about "my image got removed". I'm not feeling too strongly about my images being removed. But seeing "Deleted: 147" on Hyperion's userpage makes me fume. How many images I'd absolutely love to see got wiped out of the site before I could see them?)

I can assure you that our team is doing only what is right for the site. All actions taken are purely to help benefit the quality of our post and nothing more. The quality of our uploads is our #1 priority. Sure, it's a shame that some of those post might not get the approval they would expect but an example has to be made regardless of what should pass for quality or not.

Example, your pfp. Despite having a high scoring. If it wasn't Grandfathered, I would've deleted it without hesitation.

scout said:
There is so many anatomy errors that would need to be described using tags for it to be useful and we would need thousands for each one... No one here wants to see a few hundreds two_anus tag or merging_limbs or poorly_detailed_eyes tags. No one wants to tag them, no one wants to debate them, and anyone can play tag war with each other to remove those "negative tags" from their own personal posts right after.

this is a nonsensical claim. "thousands of tags" are not necessary to capture the most significant and prolific errors. mods do not give thousands of reasons for a post deletion. e621 does not support thousands of colors for fur tags.

also, posts can be tag locked. this site already has dramatic users that fight removals. its not the end of the world. you already say that users that dont care to fix their posts dont care to tag. do they care enough to check tags and get into a tag war?

angry_puppy said:
I can assure you that our team is doing only what is right for the site.

i think the question of "what is right for the site" is what this thread is asking about

this is a nonsensical claim. "thousands of tags" are not necessary to capture the most significant and prolific errors.

I don't think you quite understand how AI training works. For a small, very specific LORA, sure, some would claim that above 100 tags would be sufficient to get "good enough" results, to nudge a particular model in the right direction; using recommended weights, for a model in particular... "Oh but when you switch to a different model, it doesn't work that well? Why is that?": because there's not enough training data. And yes, loras (that are working well) are very specific. We're not talking about a small lora here, we're talking about a much larger scale model, and the target is "anatomical errors"; which is a very large scope that includes tons of errors. It's not by coincidence that models are getting better and better. Sure, better tagging and better quality (higher resolution, uncompressed image data) helps a lot, but also the training data keeps growing. But every time you add data to a training model, it introduces a new bias based on the content you feed it. And the more training data you have in it, the more you have to be careful to balance it out to avoid those bias. Feeding it bad examples of what you don't want to see in an AI model is probably the worst approach at refining an existing model.

e621 does not support thousands of colors for fur tags.

Furry models weren't trained from scratch solely on e621. They are a merge from the standard Stable Diffusion base model, which had a lot of training data on colors already.

mods do not give thousands of reasons for a post deletion. e621 does not support thousands of colors for fur tags.

Because they don't have the time to give out thousands of reasons. They would need to quadruple the number of staff to put this amount of care into every rejected post. And why would they do all the work, when some users don't even seem to put any effort on their end? There's a reason that regular artists are so high-praised. It's because they have to deal with people's criticism constantly, from regular users, commissioners, etc. which gives them the motivation to improve their skills. Why is it so wrong to treat AI artists the same way (on e6ai specifically, because they clearly state that their #1 aim in their upload policy is quality; nobody is stopping you from using any other posting services)?

also, posts can be tag locked. this site already has dramatic users that fight removals. its not the end of the world. you already say that users that dont care to fix their posts dont care to tag. do they care enough to check tags and get into a tag war?

Yes, but only privileged accounts can manage locked tags on submission, which brings me back to; why is it 100% the staff's responsibility to uphold quality standard, when it's the user's responsibility to respect the upload policy and they do not?

And they probably will if they feel like they were "wronged". Not caring about omitting or two many positive tag is easy. Having tags that focus on all the negative aspects of your post, I think that lots will take it much more personal.

scout said:
I don't think you quite understand how AI training works.

im talking about making this site functional as a gallery, not as an training dataset. which is what i believe the point of this thread is. if anyone wants e6ai to be a training dataset with readymade prompt tags to generate slop, then that's a different conversation (and probably a bad one). but to be a functional gallery we don't need thousands of tags.

the point of this thread is not to determine the efficacy of training on variably-tagged synthetic data. its about people not having what they make deleted.

scout said:
Because they don't have the time to give out thousands of reasons.
...
Yes, but only privileged accounts can manage locked tags on submission, which brings me back to; why is it 100% the staff's responsibility to uphold quality standard, when it's the user's responsibility to respect the upload policy and they do not?

the point you seem to be making is that everything is the staffs job, but the staff cant be overworked, so we have to settle for this unhappy medium. i dont think the staff have to do all the work. thats why i made the sets that i did. you didnt acknowledge the fact that most people that dont care about tags probably wont care about people tagging their work as having generation errors.

i think the rest of your post follows that misapprehension that im talking about something that im not. e6ai is a gallery site, lets make it work like a gallery site.

scout said:
because they clearly state that their #1 aim in their upload policy is quality; nobody is stopping you from using any other posting services

i dont think incentivizing solo female loona big_breasts with her hands behind her back encourages quality. its true that people can use another service, but its also worth discussing what makes this service worth using

ultimately i dont know what youre saying. artists improve with critique, but some will take negative tags personally, so thats bad? or good? staff work too much but need to handle deleting all posts every day and cant offload some responsibilities to the users?

silvicultor said:
I assume the long term plan is that e6ai becomes an archive for high quality synthetic data for future furry model training. And this synthetic data should not only contain maintream content, otherwise future models will also be able to produce such content only...

it seems like this speculative post is why the conversation pivoted. if this is really the goal then this site is going to have a hard time being useful considering all the grandfathered content and gen errors that slip through the cracks

xerxes_i said:
it seems like this speculative post is why the conversation pivoted. if this is really the goal then this site is going to have a hard time being useful considering all the grandfathered content and gen errors that slip through the cracks

Nah, it more than just pure speculation. I’m an active user on Furry Diffusion Discord. And in the e6ai-channel you can get many insights about e6ai, because the staff also is active there. And the training date issue was mentioned there several times in the past.
It’s certainly part of the motivation for the new quality standards, but not the ONLY reason.
If you join the server and read there a bit, you will understand much more what is happening and why, so I can only recommend it to everyone who is interested in e6ai.

i think any training ideas need to be secondary to making this a functioning gallery site. its unfortunate that this all has to be hashed out on discord, which is difficult to search and temporary, instead of on the forums that the owners pay for

e6ai is a gallery site, lets make it work like a gallery site.

If this if how you view it, then I understand our difference in opinion. A "gallery" implies that every user have their own personal account, can manage content however they like (add, delete, manage their own comment section for all of their posts, etc.), have a valid source to later claim any ownership claim on the stuff they posted, and more.

That's not what e6ai is, and that's not what e621 is. Those are archive danbooru. E621 is not used by traditional artists as their primary "account", because there are no "personal account" on an archive sites. Most content management has to be done by management staff. The source code itself was never meant to be for traditional artists. Just because you do your own uploads on an archive doesn't mean that you can get special treatment.

But yes, essentially it is not a training dataset. This is a compromise between some users wanting to treat it as a personal gallery vs. regular users coming here and having an enjoyable experience with at least some level of quality control. If you want a service similar to e6ai that acts as a personal gallery instead of an archive site, it'd probably be easier to code from scratch rather than to rewrite all the current code of e6ai to make those changes.

scout said:
If this if how you view it, then I understand our difference in opinion. A "gallery" implies that every user have their own personal account, can manage content however they like (add, delete, manage their own comment section for all of their posts, etc.), have a valid source to later claim any ownership claim on the stuff they posted, and more.

That's not what e6ai is, and that's not what e621 is. Those are archive danbooru. E621 is not used by traditional artists as their primary "account", because there are no "personal account" on an archive sites. Most content management has to be done by management staff. The source code itself was never meant to be for traditional artists. Just because you do your own uploads on an archive doesn't mean that you can get special treatment.

But yes, essentially it is not a training dataset. This is a compromise between some users wanting to treat it as a personal gallery vs. regular users coming here and having an enjoyable experience with at least some level of quality control. If you want a service similar to e6ai that acts as a personal gallery instead of an archive site, it'd probably be easier to code from scratch rather than to rewrite all the current code of e6ai to make those changes.

sure, i meant an archival gallery as opposed to furaffinity. the question is whether making generation errors needs to be a "content management" issue i.e. the responsibility of staff, or something users can handle on their own. i think its the latter. this is more accountable than having the moderators apparently arbitrarily determine what posts should be allowed. i also think that uploaders (who are often generators) should have upload limits that threaten to go into the negatives and incentivizes posting a lot of safe slop

xerxes_i said:
i think any training ideas need to be secondary to making this a functioning gallery site. its unfortunate that this all has to be hashed out on discord, which is difficult to search and temporary, instead of on the forums that the owners pay for

Maybe. Ideally it should fulfill both. But I decide nothing here, I’m just a normal user, like you.
But Discord server is still a good place.
Just minutes ago a user there asked about a deletion and got a detailed answer about the flaws in the image. So if anybody doesn’t fully understand why an image was deleted, the e6ai-channel is the way. You will also get help there to improve.
You say Discord is “temporary”, which is somewhat true, since posts will get buried pretty fast, but you can still search. And also the discussions about quality happen almost everyday in e6ai-channel, so you don’t really miss anything.
It’s a nice community of it’s own, and it also has image channels were you can post your stuff. And the FD server has way more relaxed quality standards than e6ai itself.

Personally, I'd rather trust the judgement of the staff in place which, some were there from the very first creation of e6ai, and others joining the staff after having been active in the AI community, active on the Furry Diffusion discord server, have proven experience with AI tech, and had to make that hard decision of raising the quality standard after looking through all 92k posts for approval to have a better picture of if it was needed or not rather than a few vocal regular users and other AI directors which (I don't want to discriminate) but most have joined later, so they might not see the bigger picture if they only see a small window of e6ai's current situation.

But that's just me.

the site getting handed off to power users is not on the table. what is on the table is making staff not have to be the sole arbiters of taste/quality. a post can be interesting to a viewer and still be "wrong" if a barely-in-frame hand is missing a finger. let it be marked as wrong but kept up, with other tools in place to prevent spam

Yes I agree that it can be interesting to look at for some, even if it has wrong anatomy. But why do you absolutely have to share it on e6ai specifically?

Someone has to pay the bill for all that storage space, for hosting. If we keep everything, regardless of how strict should be the quality standard argument consensus is, it's gonna be a problem. For everyone.

Cost increase for the owners, difficulties to search content relevant to the user due to the spam and poor tagging, encourages users to post or repost very poor quality, badly compressed media just like on any trash social media sites or excessively upscaled images at an absurdly high filesize. It's their right to want to make things differently with their own set of rules.

You want unfiltered questionable quality AI art to pass without problem? Upload on https://rule34.xxx/ instead. It's basically the same system as e6, except for the fact that anarchy reins, everything gets accepted, or automatically archived by bots from other sites, user comments are the creepiest, and it also allows traditional art and human-only/anime/non-furry digital content. It's a great way to experience what a website's condition can become with next to no useful moderation services in place.

scout said:
You want unfiltered questionable quality AI art to pass without problem?

i dont think im being interpreted in a reasonable way here, so i think im done with this particular conversation

of course no one has to use this site. e6ai doesn't owe anyone serverspace, but no one owes dragonfruit ad revenue. thats just part of the deal. oh well!

Sorry if I was offending using "you" to provide an example. It wasn't meant to target you specifically, but a more general "users that don't want to have strict quality rules to be enforced on their own submissions".

Ad revenues are the most common profitable income that any free hosting service can get. E6ai is not special; all of the current alternatives also have selfish profiting interests. At least e6ai doesn't sell my online activity/interests/shameful degen furry fetish to some random third-party companies for targeted ads, unlike some others. That's the way the Internet works nowadays.

So let's agree to disagree.

As a new "director" myself, it is definitely disheartening to see your work deleted en masse when you feel you'd made a good and honest attempt. I myself recently uploaded 10 images (that were well received even!) and had all but 1 deleted. Obviously I want to do better and get better, but I thought they were good attempts from a beginner and honestly hadn't noticed the flaws until pointed out.
I feel there should be a little more middle ground between good and bad; there are glaring and obvious errors that readily distract a viewer from the overall content of an image and then there are less extreme errors that may not be noticed unless specifically looked at.

I don't think we should allow any and all, but certainly as E621 allows less than perfect works from newer artists, we might have room here for our new artists to upload their early works and seek feedback from the people who will be enjoying it.

At the very least, the formula for determining upload limits should be altered so that it does not punish newer/less experienced artists/directors for small faults they may not have even noticed or considered egregious enough to be deleted.

On an unrelated note, does anyone know a good community for budding directors? A discord or such?

lilac_clouds said:
As a new "director" myself, it is definitely disheartening to see your work deleted en masse when you feel you'd made a good and honest attempt. I myself recently uploaded 10 images (that were well received even!) and had all but 1 deleted. Obviously I want to do better and get better, but I thought they were good attempts from a beginner and honestly hadn't noticed the flaws until pointed out.
I feel there should be a little more middle ground between good and bad; there are glaring and obvious errors that readily distract a viewer from the overall content of an image and then there are less extreme errors that may not be noticed unless specifically looked at.

I don't think we should allow any and all, but certainly as E621 allows less than perfect works from newer artists, we might have room here for our new artists to upload their early works and seek feedback from the people who will be enjoying it.

At the very least, the formula for determining upload limits should be altered so that it does not punish newer/less experienced artists/directors for small faults they may not have even noticed or considered egregious enough to be deleted.

On an unrelated note, does anyone know a good community for budding directors? A discord or such?

Howdy. You find a lot of how-tos, information, and people to ask questions from here: https://e6ai.net/static/discord