someone released a chrome extension that does just that for her styles

@thewinterwren would you be so kind to point me to that extension?

@thewinterwren Where is the Linksauce?

CSS:
.mantine-lw13s0 {

display: block !important;

}

@thewinterwren could you please link it?

You can simply use CTRL+F, since the page loads the entire list. This'll break when or if civitai decides to have dynamic loading. :p

The support ticket guy said they are thinking about it. As a senior web developer and php programmer I could have it implemented in.. 5 minutes at most.

@thewinterwren please link it

@ThatGiantessGuy this doesn't help if you're just trying to browse

LoRA with Frank Frazetta style is in another thread. But it's not very similar.

@freckledvixon that would be awesome, thanks for lookin

@garifetdinov604i@garifetdinov604i have a very good xl lora I use for Frazetta but it doesn't wanna work with pony

I am looking to do the same if you want to collaborate. Cause i didnt find one and I think it is missing

@Pagame lol yeah. I made a spreadsheet that catalogs each of the loras I've downloaded from the set so far. There about 38 on that list, and the complete set is probably at least 8 times as large. So we're looking at a list that's over 200 entries long and growing longer by the day.

@freckledvixon I can't wait to see it!

You are amazing @freckledvixon! Downloading this right now :D

Living for the sass at CivitAI's dev team in the readme! I love the community on this site, but I wish everyone would migrate to a faster, more customizable platform. I daydream of searching by seed distance or something and we can't even get a good display for model versions.

what a true hero.

wow haha awesome stuff!!!

Probably unhappy from styles being mixed

Nice of them to do something useful instead of just messing on my car.

edit: apparently I'm an idiot and birds can draw

@freckledvixon I can't seem to find it. How do i look it up?

@Ziggai In case you haven't figured it out yet, just right under the title of ''Vixon's Pony Style'', you'll see a row of listed names for different models. Keep spam clicking the right arrow to go through the list until you find the model named ''Arcane''. Vixon's a fantastic creator but she tends to be stubborn and it's a pain in the butt to search for her older models..hopefully, CIVITAI would tackle this with a drop down menu at some point.

@Ziggai ctrl+f inside this page

@Cronoe like @obiyomida said ctrl+f (find in page) function actually works great, bro's a genius, i use that fuction a lot but to use it here it didn't cross my mind, thanks a lot

That would be much appreciated.

Finding or exploring for new styles is so hard with all the clicks and scrolls

Yes so hard !

@TanjiroMuzan Yeah. It's not great. We're cooking up an update that will make it so that collections like these can be shown more like items in a YouTube Playlist.

@JustMaier Any chance of a 'download all' sort of button as well?

Like Vixon stated, it's a trigger word for the model ''Detailed Painting v1.0'' Hope that helps.

those are good stuff :D

there are tons of artists in here lol, one of them is as famous as Gorillaz's character designer/artist Jamie Hewlett, at least one of them is probably gonna have a problem with their art being AI trained

It means you only make pictures for yourself because these are trained on the works of real living artists and can't use the results to make money for yourself and that if you do, Vixon and Civitai are free of guilt when the artist comes for your neck.

@skorpious_ Good luck proving you created a new style. Must be awesome, i will claim the genre jazz lol. I think its hilarious to claim a style, genre, writing style etc. Only creative works should be protected, not the general idea, method, or style behind them. Without inspiration no art, no music no nothing. But I respect the rules. Will get rid of these "unique" styles 😊

Alternatively, people could start using collections to display collections of different models.

And have one trained model per model page.

made a little script that helps: https://greasyfork.org/en/scripts/508495-civitai-model-versions-scroll-remover

@kaljinn oh hell yeah

@kaljinn you're script is godsent brodah

well, organizing the loras could need a little more work. It wouldn't be too hard to expand the scheme for the db for some new features when it comes to grouping things up.

Do you have any specific request on how you would like the LORAS to group? Is it just a dropdown? Nothing else?

@freckledvixon So why don't you put it in a new lora of its own, because it can't be found by searching because civtai considers it as a version not a lora

With this work, many beautiful styles are lost. I mean bro u have 11k flowers and ur top on civitai so what are you afraid of?

Probably the best lora out there

@Yeatluffy Open your eyes a little and you will find what is better

@freckledvixon calm down 🗿💔

I don't think the market exists for it yet, once it's fully cracked I imagine most creators will

@freckledvixon Yes but is still a massive list, this is one of my most used loras if not the most and after over a month on civitai and i fell that still haven't discovered what i offers fully, and that's counting my attempt to test effect of most of those styles on one of my characters, the more styles are inside that one lora the harder it is to explore them

@freckledvixon Thank you, I kept scrolling and it took FOREVER to get even close to the start of the list. I appreciate all your efforts!

I made a script that makes it a little less annoying: https://greasyfork.org/en/scripts/508495-civitai-model-versions-scroll-remover Still would be better if Civitai implemented a fix natively.

Lately ive had a hard time deciding something i always knew the answer to-which was my favorite artist: there's the greats like Gustav Klimnt and of corse Monet- both have their huge contributions to art history but "Trippy Mush and Stars" reallly has burst onto the scene lately in the past decade with some truely mindbending pieces,not to mention the works of "Dark Purple & Creepy" have really been restoring my faith in modern art especially after DP&C's showing last year at the Phila. Art Museum.

@freckledvixon take a glass of cider, put on the cam, and enjoy. You can see it as the natural next step in your journey as the first true AI Artist on youtube :D . do it!! :D

If your LoRA weights exceed a sum of 1 the risk of artifacts will go up progressively. That doesn't mean 1 is the cap. There is no hard rule. Often you can hit a sum of 1.3-1.5 with no notable issues. It depends very much on the LoRAs. For example, a LoRA that only affects skin tones will have little impact on the overall composition even at higher values (when combined with other LoRAs) unless it was poorly made. I have seen poorly made style LoRAs that have a huge impact in the composition, even at lower values of 0.4 even. Then I have seen well made style LoRAs that have very little impact on anything but style.

Lower CFG means less burn in. I've found a CFG of 6 works best with Pony with 7 working fine. The longer the prompt the worse a higher CFG will impact the image because you're telling the model to do too many things at once and also emphasize them.

People also often forget about prompt cohesion. A bad example but let's say you have a longer prompt and in one section you wanted the character to stand and in another you said they should be sitting. That breaks cohesion. In some images the model will create them sitting, some standing, and some might cause artifacts where you can't tell if they're sitting or standing. The more breaks in cohesion like this, the worse the images will get regardless of what LoRA combination you have. Combine bad LoRA weights with bad prompt cohesion and RIP your images.

I only use denoise for ADetailer.

My final note is that if you can achieve something reasonably well with prompting alone, don't rely on a LoRA. Every LoRA you add will have an impact on the comp that you might not want, sometimes you won't even realize the impact without testing. With LoRAs, less is more usually.

@atypicalnavigator Well, you are correct about the 'well made Lora' thing. Just for example, I had a Lora that's only supposed to affect one small part of a character, but if I turn it up too high, it affects the entire image. So, I get, first-hand, where you're coming from with that.

I don't have enough experience with Adetailer to really get where your comin from with that(think I had it for a minute back when I was still using A1111 a week or two ago. but since then I went to SDN, and now I'm in comfyui). I'm still trying to figure out which components work 100% offline. Cuz I'm 100% not down with corporate data collection/harvesting. They want my data, they can pay me for it.

Prompt cohesion can be tougher for some than others. Until I get my new PC up and running, I'm doin this on an 8gb vram laptop. So Not as much computational resources or ability to use better models/methods that comprehend prompts better.

That is to say, if you're using an FP32 thing or flux or w.e, then you can probably type a sentence or two, and get the results that you want. But if you're not, then you're gonna be frustratingly adding more and more to your prompts and negatives and weights just to get anywhere close to what you wanted. And if you don't want female characters in your images (got nothin against women, but prefer male-centric images), then that's at least a full line(by default) worth of words and phrases in the negatives just to give a better chance of your male characters not being turned into women. and probably end up with a full line in the positive prompts as well to that end.

but yea, if you break cohesion, then it'll probably forget half of what you put in, and make your characters into women anyways. So, for those on lower end systems, that just comes down to striking a balance between short and concise, while still keeping enough in the prompt that the model won't force things that you don't want into the image. If all I wanted was to see women with oversized tits, I could prob use 3 word prompts with nothing in the negatives, and get all I want, all day long.

Oh, my final note, sometimes ya do need the Lora to get what ya want. The models have a lot, but there's specific characters, styles, or even things that a lot of models have restricted, like blood/gore, real world landmarks(though the Loras for those are extremely limited), NSFW, ect.

I do get what ya mean though, and you're not wrong, I'm just adding that caveat.

@Lazman Seems like you have the right idea. I use LoRAs, just as minimally as I can while getting the desired effect. Like I have a skin tone LoRA because most models default to pasty-white and sometimes no amount of prompting for darker skin tones makes a difference because none of the training data had anyone other than light skin tones. But if I crank it too high it can start to do more than just change skin tones, though mostly it is well made and has helped a lot.

I started out with 8gb vram as well and having 24gb vram now is indeed a luxury. It's nice to come across someone else who also cares about data collection, a thing that seems to only get progressively worse.

I wonder how you like ComfyUI? I have thought about giving it a try for a while now but just haven't because my setup works fine for now.

Good luck with the new PC.

@atypicalnavigator ComfyUI is good. I highly recommend. It's got fewer hidden components than the others, so lower chance of undesired data collection. But, more than that, I've learned after using comfy, that the others, I don't remember enough about easydiffusion to say, but deff A1111 and SDN have built in blockers that steer your content and make it harder to get what ya want. Comfy looks intimidating, and admittedly, there's still some basic things I need to re-learn how to do in there that were simple in the other ones, but imho, it's worth it in the end.

There's a tab manager thing for it that makes it way easier to manage, but I'm on the fence about getting that, cuz I'm a bit afraid it'll turn it into another A1111/SDN, by bulking it up with censorship and surveillance. The more bare bones something is, the harder it is to hide that kinda stuff. And corporations do tend to use attractive words/phrases like 'ease of use, increased security, ect.' to take away user control and make you more dependent on what the corporation provides, on top of siphoning your data to sell for a profit.

Speaking of model training, I stumbled on another annoying thing that corporations have done in recent years. Conveniently just as AI is becoming more mainstream, and more people are getting into training their own models, they've removed support for SLI/Crossfire, and cranked up the price of their top ranking cards.

To clarify on the implications of that. In a typical corporate fashion, they've once again held production resources high over the heads of those on a lower budget. The things that corporations are getting away with these days, makes me sick. Things like free market, and competition these days, are nothing but an unfunny joke.

Anywho, I could go on for days about that shit.. Got my new PC, it's great. https://www.youtube.com/shorts/2tNUlMZ2xz8

It beats the hell out my laptop. I went with Garuda Linux on this one. Linux is still a pain in the ass, even on this. Currently got some issue where I can't drag widgets onto my desktop cuz something is off with the display server. Honestly, the entire display server thing in Linux has been one of my biggest headache across distros.. With windows, it just installs a display driver, and it works, no prob. Linux has some arbitrarily complex display server thing that has a habit of randomly screwing with anything and everything else on your system for.. reasons..

That aside, I much prefer the look and feel of Linux over windows. It's much more customizable and feels far more personal. Windows has... wallpapers.. And forced updates and forced telemetry, forced adware, bloatware, spyware. Like yea, you paid for your system, you 'own' it, but as long as windows is on your system, M$ owns it, you're just being granted use of it. Also, all that major, windows-exclusive software comes with a big price, whereas most things on Linux you can get for free via open-source options/alternatives. Some Linux distros are bad for the data stuff too, but they're mostly obvious about it I think, like, if you use the google chrome distro expecting data privacy, then ya kinda deserve to get screwed, lol..

@Lazman Super late reply...haven't been on much.

Agree with what you said and yeah market competition is a joke now because it barely exists. When I saw the projected 5k series Nvidia GPU prices I was really glad I got my still over priced 4k series card because the new prices would cost me selling vital organs on the black market or turning into a drug dealer lool. Gotta love how our legislators just let this stuff happen but hey between stock options and getting your own slice of the pie they hardly need to do anything but pretend they're doing their job for a few years.

I'm still on Windows 10 and it's the last version of Windows I can see myself using because having built some computers for others and installing W11 on them, I can see what a nightmare it already is and how much worse it could become. I'm undecided on an OS but thinking about Linux Mint. Yes it will be a headache any time you switch away from an OS you used and became technically competent with for years but sometimes the growing pains are necessary and worthwhile in the long run.

I will have to look into your claims of Automatic1111 content blocking/manipulation as I'm unaware of this. The only time I have felt I've had issues with content using A1111 is based on the training of the model. Like if the model was poorly trained in an area then it won't perform too well with certain prompts. Likewise some keywords are overemphasized in training data and others are under emphasized. Once you figure out what those triggers are it becomes a bit easier to control.

After doing some research on ComfyUI I decided to stick with A1111 (for now). My understanding was that ComfyUI is easier to manage with its UI but has less overall features, settings, and depth than A1111. So ComfyUI being fine if you don't need as much granular control or advanced features. That is just what I have heard/read though.

What I can't figure out is why I can't get Flux to work despite that I'm running the latest stable versions of xformers and PyTorch. Never had a problem with SDXL model generations but Flux just craps out noise and then a blank image having tried two separate models hoping to rule out just one model causing the problem.

That's a slick PC build btw.

@atypicalnavigator "When I saw the projected 5k series Nvidia GPU prices"

I hadn't even looked into pricing yet, but I saw projected/rumored specs, and I was fukin livid.. apparently the 5060ti, next year's version of my current card, is gonna be HALF the Vram.. What in the AF.. and to add insult to injury, the new PCIe 5.0 slot is throttled to half lanes for any card that can be got for under 1000$ (CAD). Well, assuming they stick to this year's pricing scheme, if they boost it like rumors say, the 5060ti could end up costing 2-300$ more, basically 1000$ with tax, and borked specs to boot. WTF is wrong with that greedy fuckin corporation.. The only card they upgraded the Vram on is the retardedly expensive 'consumer' card that's gonna cost at least 2-3 grand.. and they still only boosted that to 32gb.

They're working their asses off to ensure that AI training remains delegated to the wealthy.. At least until other corporations manage to monopolize it..

Also, just found out today that apparently OpenAI is in bed with M$. Even they couldn't manage to stay afloat without partnering with M$ and receiving a billion dollars from them. So, I think it's safe to say that M$ pretty much owns OpenAI, even though it hasn't yet been made official. But, as soon as they start struggling again, big daddy M will surely come in and take over.

Really though, if a company like OpenAI can't make it into the big leagues without being taken over, due to the greed-based strategies of other corporations, what the hell hope does any business have of succeeding, unless they hide away in a small town somewhere..

I've been preached to my whole life about how we live under a free market economy, and competition and blah blah, but at this point, I can't even see there being any new amazons or googles or whatever, cuz the current ruling corporate structures will consume them before they can become competition.

PS: apologies if I'm coming off too ranty or 'conspiracy theorist' at any point. I've just come across too much agitating info today, and in general, as someone that's autistic, I'm a natural expert in pattern recognition, and the patterns I've seen in our society, of human behavior and toxic borderline evil corporate practices.. OK, I don't use the word 'Evil' lightly, cuz it sounds downright cartoonish, lol. But just look up Purdue Pharma, and they aren't even the worst one, believe it or not.. Nor the most recent.

"they hardly need to do anything but pretend they're doing their job for a few years"

Face it, ''Lobbying" ruined any chance of living under a non-corrupt government in a capitalist country. It's just a synonym for legal political bribery. And with no wealth cap on the rich, things can and will only progressively get worse. Cuz, little known secret, corporations don't suffer under recessions. I mean, maybe not 100% true, some do, I think Target is going under and a few others. But I'd say it's logical to assume that the ones that do go under, either go under because they're run by people that actually care about their customers, and/or they're competition for the bigger names (like Walmart, in the case of Target).

But, back to the point, when a recession rolls around, modern corporations treat it like a fire sale. They have so much expendable income, that they use it to buy up other failing businesses and buildings on the cheap. So, they don't have a 'boom and bust' like the rest of us. They have 'invest and reap profits' and that's it, until they swallow everything and eventually end up just being 5-100 corporations that own 99% of the production and distribution.

"any time you switch away from an OS you used and became technically competent with for years but sometimes the growing pains are necessary and worthwhile in the long run."

Preachin to the choir here. Ya don't even know until ya do it, lol. I'm not trying to urge you not to do it though. I think everyone should bail on windblows. M$ has monopolized the OS scene long enough. Them and CrApple with their overpriced CrApple PCs (Well known for FUN things like.. Image editing!).. You can even buy a thousand dollar monitor STAND to go with it XD..

But yea, Linux is a different beast. I'm NGL, I kinda evily smiled inside when I read the part about 'technical competence' Windows is autopilot compared to Linux, everything just works. Now, if you can manage to get the hardware to agree with Linux, then that''s the biggest headache out of the way, but there is a lot to watch out for as a new user of the OS. There's also a lot of fun and exploration though, it's not all bad. Since it's not locked down like windows, you can do almost anything with it, and if you know any coding/scripting, you can really have some fun with it.

To be fair about the A1111/SDN thing, I did swap those and easydiffusion out pretty fast, I just felt that the latter two felt a little more passively restrictive than ED. It also depends what you're into. If you want 'big boobad waifus', you will NEVER EVER feel restricted in any model EVER.. Even most of the male centered models are inclusive of female outputs, if not outright biased towards them. But, I digress on that point, I've done enough ranting for one post, lol..

"keywords are overemphasized in training data and others are under emphasized."

Yes, trying to train a lora on my local machine. But I'm a total perfectionist, so I will likely end up going over every image caption and doing them manually. The primary methods, WD14 and Blip(1 or 2), are wickedly flawed.. with captioning methods like these, it's no damn wonder ya gotta contort the prompts the way ya do in order to make the models do what ya want them to. the stupid thing probably tagged every non hairy/hulkish male as '1girl'. Well, even some of the hulkish ones come to think of it.. Cuz I've seen way more hulkish looking dudes with futas than I ever wanted to.. And that is a bizarre thing to see on a realistic looking human man..

"That's a slick PC build btw."

Thx. I do take pride in that. This is easily the most beautiful PC I've ever owned, and I've been building my own rigs for over 25 years now. Never thought I'd really be down with a white case (not that I'm the casist type..). But, this thing is mesmerizing. ARGB lighting is so soft that I can sleep peacefully with it right on my face (the light, not the case).

"ComfyUI is easier to manage with its UI but has less overall features, settings, and depth than A1111"

Dang dude, I think whoever told you that, was dyslexic, lol.. You got it full on backwards. Comfy is the one with max customization. It's like Linux, you can do anything with it, but it just takes a little extra work. As for getting flux to work, I had a hell of a headache with it myself. I could guide you with comfy, but not with A1111/SDN, cuz I haven't the first clue about flux for those. as for granular control, there's lot's of customization in comfy, though, thinking about it, it may lack some of those hidden settings options that the A1111/SDN had. in the settings menu (it just occurred to me) there is very few options in contrast, the thing does work well non-the less, and I've even managed to get flux Dev working in it. and flux shnell in comfy with UI manager is autopilot (dev is the better one, but took a bit more work).

What's more, about A1111, the comfyui manager has a menu for adding stuff from A1111, so you can get a lot of it back. Though some may require a little extra tweaking to get it working. Typical linux type headaches. dependencies, conflicts, etc. That's the thing with Linux, ya learn patience. Oh, final note about Linux, distro matters, but not as much as the main type. IIRC, there's Arch(hardest but most custom), Debian(Ubuntu types, relatively easy mode), and red hat (the middle-ground type, Fedora being the flagship distro for that). Of the ones I've tried so far.

I think I like Garuda most for the visual appeal, Majaro(Arch) for it's professional look and it just worked better for me than fedora. I didn't hate Fedora, in-fact I rather liked it other than the fact that I couldn't make it function well on my laptop, but that's a hardware issue. But also when I went to uninstall Firefox from it, the stupid thing wanted to uninstall the kernel with it, which is insanely stupid, lol. As for Mint, Haven't tried it, IIRC, I think I liked how it looked, and it was on my list to maybe try sometime. You'll have to let me know how it goes.

Oh, also two different desktop managers, Gnome and KDE, I prefer KDE. Gnome is pretty, fisher price look, might be good on a tablet. Hated it for the most part though. Garuda blends use of the two, so ya get the 'pretty' without the 'pretty-stupid'.

I actually just use a modified version of the Vixon-Style-Explorer chrome extension made by markuryy on github. I got it running as a userscript. Its not much code and it seems to do the job for my needs.

Just figured out that you can middle mouse click the styles to scroll through them very quickly

I can't help but wonder how many people are unaware of the functionalities available to them because the creators are using the UX in ways that differ from how the webmasters programmed it.

Probably wouldn’t recommend it. Looks like a bunch of artists may be flagging these models and since that might be easier to flag as use of their art, selling it could turn out poorly for you. I’m not a lawyer so you do you, but just fyi

@freckledvixon thank you for your reply. I felt exactly that way. I had to create my own art.

Sorry M'aam, good job there

@freckledvixon Not denying that at all. I'm simply saying I appreciate your current one. It is a certain style that I really like.

@freckledvixon I had to give up and try to find the end another day :D