#🏞|general-with-images
1 messages · Page 104 of 1
i mean ive posted very sexy artistic images
Theres an automatic filter as far as i know.
Very happy with this one
is it SDXL?
No, 1.5
seriusly 1.5?
looks better than current SDXL
Yeah, Photon model
well photon model and 1.5 its not the same xd
no sorry man I understand you.. but this is impossible in the vanilla 1.5
Anyway its a incredible image... seriusly
Definitely these days of waiting are going to be eternally long. They are magnificent.
1.5?
XL
Thats well done, its handling the symmetry well and the lines of the features are accurate.
maybe they are improving it daily behind the scenes?? i use clipdropio
Im out of credits now...
me too... im waiting for my daily renew thing.
Will SDXL be free and locally run?
Yup
How many generations does it allow you to create... in my case it's rare to get more than 6 if I remember correctly, sometimes less and I get the message "enough for today my boy :)"
LOOL, yes this helped a lot...
Those images i posted well they are just a few out of the current 3000 +
from 2 days
I find myself in a serious and paralyzing doubt, on the one hand 10 dollars is nothing... on the other I'm going to have it for free in a few days... as I say I can't decide.
But i think of my electricity bill. its not free to run it on your own card
That is true, another aspect to take into account.
And then what speed would you be running at on your own card? versue instant
they probably run on H100s

That is one of the things that I would love to know, because being used to waiting a few seconds on a 3060, things are going to be hard.
Pay for one month. bookmark the cancelation page.. just see how it goes for a month
how much for a month?
Maybe they keep updating the terms im not sure
It's easy to get out of doubt, I'll take a look
The only major advantage for running it on your own card is NSFW content... But clipdrops filter misses it quite a bit, im getting images i cant post here
If I'm not mistaken, 1000 credits 10 dollars, it indicates approximately 5000 images but it is not correct since XL usually consumes 1.60 approx. You should have a really advantageous old plan, I wouldn't lose it 🙂
Pro
5£per month
Skip Queue Stable Diffusion XL (1500 images per day)
Skip Queue Uncrop (1000 images per day)
Unlimited Background Removal HD
Unlimited Cleanup Pictures HD
Unlimited Relight HD
Unlimited Image Upscaler up to x16
Unlimited Web Editor
Unlimited Text Remover
Unlimited ClipDrop iOS
Unlimited ClipDrop Android
Unlimited Replace Background
Sorry my fault, I was looking at the prices of dreamstudio.
Never used it tbh, i mean Niji
And stopped using MJ since back in december 2022
So even before V5
Only Firefly/Generative Fill atm
Firefly 🤮
I have both on Photoshop, and I can say that Stable.Art does 100 times better than Firefly
this was v5 niji, I can beat it with locally run sd1.5 custom models, lora and controlnet etc. but it was fairly impressive in it's ability to draw with minimal prompts and get characters right.
Not in my experience
Well depends on some stuff
SD allows more control to say the least which ofc is a bonus point
I used SD before but i knew from the beginning it wont end up in my pipeline
anybody know where i can get advice for lora making?
im making a lora of a very complicated character
the new SDXL paper throws shade on MJ 5.2 saying it doesn't do prompts as well as 5.1 lol
@untold zodiac #🔧|finetune
thanks
censorship and art dont mix
this was when I gave up
scannable QR's
In case of AI art generators they do 😄
I mean ofc its also easy to do so compared to do it in digital art programs like Photoshop
Disagree
Also i dont care about certain degree of freedom
Also here where i live some art is forbidden, not because of art but because of propaganda it might propose
We have harsh anti nazi laws
yea i dont think there shoud be any censorship in art. But putting certain art in certain places, thats where the problem happens.
is that 1.5? the details are coherent.
I'm using models from Civati but it's not SDXL
SDXL draws a perfect violin.
something weird happenning to strings
its not entirely perfect
Anyone fix
this is why i use 1.5
what’s this an old family photo of yours?
I’ve found using certain painter references gets the details on figures right in clever ways
missing chinrest, extra pegs, wrong body shape
Finally got around to testing SDXL. It totally blows.
Zero precision.
Limited vocabulary, too.
It's a f* cellolin.
if it were a cello it would have 4 fine tuners and an endpin
also the fingerboard is always one piece of black wood
was it simply less censored?
im sure the model is censored, but i dont have some overlord checking my every keyword
thats creativity 101
anything that says “oh but that won’t be good because —“ has to go
i didnt encounter any of the censored words before but i also didnt use Stable Diffusion for long
but i know that Firefly has a lot of censored words...for now
more than Midjourney
lawsuits are coming and big corpos have everything to lose
what lawsuits? On AI art?
on its sources and its abuses
In that case SD and MJ amongst all are the ones that have more to lose than Adobe, Nvidia and eventually Google
theres also possible PR problems
the bigger a corporation is the more it has to lose
no offense but i double SD market shares are bigger than Google
you mean their AI art product?
if i remember correctly MJ is having the biggest market share as of now
i mean the net worth of the corporation lol
oh, SD is nothing in that case
not an economist dont know the lingo
hmm
hi im just wondering if somone can help me i want to get more detail and less blurryness in the city what do i do ----prompt : "realistic photo, science fiction, futuristic colony, science fiction cityscape, industrial city, highly detailed, intricate, ultra hd, sharp photo, the city is very detailed and high quality, 4k,
Negative prompt: deformed, low detail, blurry, disfigured
Steps: 50, Sampler: DPM++ 2M SDE Karras, CFG scale: 9.5, Seed: 1218656311, Size: 512x512, Model hash: 980cb713af, Model: reliberate_v10"--- i have tried variations of this and still same thing
agreed lol
pfft AI art makes human art even more valuable.. so the lawsuits are stupid!
get it out of your system, pal
As an analogy, machines learn to make chairs en-mass and floods the market with cheap chairs.. But a hand made chair with hand carved wood.. That's what people see value in because of the human artistry involved.
But do 90% of people really care about the fancy handmade chair? How much stuff in your house is handmade vs machine made?
no 90% of people do not.. But that just makes the hand made chair even more valuable. People will always see value in human craftstmanship.
Im just saying, artists get the wrong end of the stick, they think AI art will replace them... Hows that???
Yeah some very rich people might, but the vast majority will take whatever is cheapest and fills their needs
oh yea im going to buy an AI picasso, versus the real thing.
wheres the value in the AI art??
The point of buying an original Picasso is as an investment so that you can resell it later for more
Most people are just as happy with prints
AI art has no value.
It can fulfill a purpose, like making a book cover.
yes most people are happy with Prints.... but that does not change my point. AI art has no vlaue.. Real artists who paint or draw or sculpt will not need to worry about their art losing value.
if you make a million plastic chairs, does that change the value of a hand carved wood chair? nope.
It depends on what sort of artist they are and who they're targeting. A fine artist making oil paintings won't be affected, but a digital illustrator making book covers, concept art, etc. now has more competition
So the artists throwing a fit and wanting lawsuits is stupid.
morning, how are we today?
yes they do but their illustartions will hold the same value to people that want real hand made art.. And those people wont dissapear because ther are a trillion AI images
The more the market gets saturated with trillions of AI images, the more the real human made art will be saught after and hold high value.
Most people who commission artists like that have a specific idea they want created for them. If the AI gets to the point that it understands their prompt perfectly, there's no reason to commission those artists if the AI can do it for free
well there is because what i just said. Human made art holds intrinsic value because its made by a human.
AI art is not special.
thats not the reason for lawsuits
People can pass off AI images as human made, if they want that money
Art forgery its been around since forever
Yes, but one forger can't work faster than 200+ artists
thats a what if scenario and also most of the people of the AI art community wouldnt be customers of artists anyway
or many
aren't the lawsuits from artists whining about losing money because AI learnt from their art or took their art to machine learn on? isnt it ultimately down to artists thinking they lose money?
I've already seen people who used to commission artists start making their own AI images because it's free
I think AI art tools are special and valuable, because you can create new art by blending styles and themes with infinite possibilities very quickly, even if the output is not perfect, it let's you broaden your imagination
from artists and from Getty for example
yes, copyrights are the primary reason to do so
it means AI art can invent new styles and new images... But it does not give the images themselves value, because theres a trillion others out there on the intenret.
oversaturaion
saturation
especially all the anime/manga stuff
yup the market will become saturated with a quintillion AI images
Which means it will flood the internet and drown out actual artists. The algorithm prefers AI anyway because it's faster and can make more content than manual art
one of the reasons i was against allowing AI art on Artstation
Then the real art will be still wanted and still hold its same or even higher value.
and i still am
Adding more competition doesn't make things more valuable
there wouldn't be trillion out there if the piece was unique in a way, like not just another portrait, oil painting done in the style of artist x
Yes it can create unique combinations of art styles and each image unique... But theres trillions of similar AI images...
also the fact that its AI image alone...
And there's trillions of similar manual art images too. How many anime artists exist on the internet?
I think it's possible to create something unique with AI art tools
yup unique... and still good.. But still just like a unique plastic machine made chair.
plastic chairs are awesome in their own way too, not in their craftsman ship, but the design, use of materials, functionality, affordability
yes.
as an artist i dont agree with pessimism of some other fellow artists
same with that ultrahype and toxic positivism by some AI art people
the person who designed the plastic chair may have spent a lot longer than someone making a chair by hand
As somewhat of an artist (digital artist before AI) i would be pleased if my artstyle came out in SD or MJ if i typed it.
I've discovered new artists by prompting for different artist styles
im not sure in my case tbh
im on the artist styles page all day clicking through artists and photographers to mix in.. it really does build awareness
I don't think it will destroy all art, nor do I think it will magically make manual art more valuable. I think the market will change, just like it did with photography, recorded music, etc etc.
but if my copyrighted material would end up on someones AI image i would eventually sue him
depends
depends on how much of a blatent copy it is or not.. if it just captures the essence,then i think thats good.
the market changes definitelly because AI art has its niche of consumers and buyers
which is already a market change
I've also generated pieces mentioning random artists, found a style I really love, gone to google them and found I didn't really like their original pieces. But I really love the style applied to a different topic they never covered in their work
im going to post some of todays SDXL stuff i made.
i mean about so much that the character for example is easy recognisable, also i would sue out of the box necessarily
it always depends
i personally dont have in mind right now at least to do this creative stuff as full-time job as of now, dunno about you guys 😄
im doing all of this on the side and also have a vision to earn a secondary and passive income, but who knows
for me it's a hobby, not to make money. I am enjoying sharing the process and results with others here
oh ofc
this is a joy for me in the first place, earning on that would just be an ideal case
not a necessity
especially considering how much i spend on digital art as hobby
damn i chose more than 10.
Easily far more coherent than SDXL...
But im sure SDXL models will end up that good.
Yeah
Still struggling with waves I see
lol
no thats what you get when you combine waves with architect styles that have distinct curves or shapes.
i want to cry
Guess my last post was too sexy for the filter
Here's a king chess piece I designed instead
guys what is the meaning of this, I can`t fix it
Do you update your extensions regularly?
Also #🤝|tech-support
SDXL is no joke really good at handling certain prompts
When a fallout waster meets a camera
Literally the prompt almost made it into the image. LOL
"Fallout Wastelander"
Creating wild west animals on a 2.0 based model; something hilarious about this goat getting a hat that says either
GOAT WITS
or
GOAT WTFs
on his head. That is all. It is nothing special, I am making thousands of them this one just stood out
can't wait to try this with SDXL
how do you generate actual real looking photos
most of the models like realistic vision and epic looks sd generated or at least studio quality with expensive setup
is there a checkpoint model that's trained on smart phone photos
the background, neck etc is most of the times blurry, probably to make the full picture of higher quality but it's a dead end give tail that its fake
AI going so fast and GMO so slow, I want cats like this now 😄
These c a t s
AI art is a skill that will develop into a trade. Have you used A11? Have you even scratched the surface of how deep this goes?
Mans doesn't know how chairs are made.
The end is near...
pseudo-real on huggingface hub works in invokeAI and is heavily trained on film photos, mostly kodachrome but also kodak 400
pseudo-flex-base uses that model's text encoder but has a retrained unet that allows it to do multi aspect eg. widescreen and portrait gens with a base resolution of 1024x1024
and for either of these models. prompt like ethnographic photography of a california house party, 1998, nikon, f1.4
Where/how did you learn all this crap? I would like to learn. I'm studying Transformers for LLMs, right now, but that's mostly unrelated. What resources will teach me the technicalities of SD and the skills I need to train models properly?
Wdym develop into a trade?
It'll be an instrumental part of Graphic Design moving forward, and a vital tool in not only visual art, but in music and poetry/lyrics. It's going to be a job.
it took me about 2 months of failing at training stable-diffusion-2-1
ask a lot of questions and abuse the hell out of GPT4 when no one is around to answer
I've been using GPT-4 a ton, but I don't know what I don't know.
I suppose SDXL isn't terrible, there's just no reasonable way to prompt it, and it ignores most of my input.
Yes, I've already incorporated it into workflows at work. It's not replacing artists here but allowing them to do more
If only it didn't take 15 minutes to boot up.
it ignores stuff it's never seen or stuff it's not seen enough of, or phrases that have never appeared concurrently in the training of the text encoder(s)
if you have gotten used to prompting 2.1 models, you have to use that style on SDXL
it's not like keyword spam in 1.5, it's more natural and precise
I have not. Any phunky phresh guides out there? My 2.1 gens are dogshit.
Sytan was having issues making animals look realistic, they would bias to artistic / cartoony because he had too many words in his prompt
i reduced it to simply stunning national geographic photo of a chinese panda in a bamboo forest and bam
Ew. I don't like that.
you can not ask SDXL for realistic, hyperrealistic etc
How do you do artistic styles, then?
they're literally just prompt keywords..
"Bad AI, very AI, go stand in the corer then read more books". Prompt was "Victor Frankenstein sipping tea".
add in the style of Bob Eggleton to the end of your SDXL prompts.
That's what I'm saying. Without keyword spam it's imprecise and vanilla.
i have the local weights, and i'm telling you it prompts like SD 2.1 does
I believe you, that just makes me sad.
SDXL?
Yes, but same in most AI Text2Image, they swoop the open Internet and also pick up on all flaws, lies, fakes and misunderstanding
yes they do.
Ew. Can't boot A11. Import error?
ImportError: cannot import name '_compare_version' from
'torchmetrics.utilities.imports'
(/usr/local/lib/python3.10/dist-packages/torchmetrics/utilities/imports.py)
What the hell.
Part of creative workflows? Optionally yes, but it still needs some big improvements. Vital part hardly
Let alone a job for itself
At best professionals might use generative AI partially in their workflow...but thats not a new job
You wont need a prompter
agreed.
If you think you're going to wake up one day in the future in an AI world where you type a sentence into a prompt and it creates things of value that earn you a living as a job... eer yea good luck with that. Because anyone can type a sentence... what value would things have if anyone can do it?
i think im right in that line of thought.. i just dont see how the AI could generate jobs that are just that easy and why they would need a human in the loop.
@split valve well IF AI gets better enough...why would they hire an prompter who doesnt even know anything in the sector, zero practice, zero experience if they could just do it themselves. People that are in the industry and have the knowhow and skills
Plus generative AI isnt good enough to be the active part of the workflow (canvas)
thats sort of my thinking... but what im getting at is, in order to create value you need to be doing something that is difficult that not everyone can do.. That means you can create a profitable business out of it or product. But to just roll out of bed and type a basic sentence into a prompt, thats not an AI job of the future, it holds no value. Because everyone could do it.
if i can type a basic sentence into say SDXL 3, in 3 years from now, and it creates an magnificent work of art at 4K.... That art would have no intrinsic value.
its never going to be a machine that can make money for you.
everyone can work at mcdonalds but its undesirable because of low pay and high stress. prompt engineer is the minimum wage job of the future
its not... Because you wouldn't need a human to write basic sentences.. LLMs can already write better than 99% of people.
sono, are you still bored or something
you'd just rig an LLM into your AI to write the prompts and leave it going.
as always.
You can make some business with gen AI but its more than just prompting and earning ofc and it doesnt necessarily earn you much as well
Prompt engineers exist already and are specialists withing the AI area, definitelly not wage minimum and nothing to do with the Stable Diffusion random Joe
i agree. It would take a professional.. like an author, or psychologist or philosopher.. currently
The application is not to get anything out of machine art, it's to get something specific out, something that comes from the vision of the creator. That is not something anyone can do and thus will continue to have value
its never ever going to be a min wage job.
sounds like copium to me
you literally just said anyone can do it
im just saying... AI will not create new jobs at the pace it replaces jobs... and being lazy and just typing some stuff into a prompt is not a valueable asset. it wont create value, because any random joe can do it.
And I'm just saying the value is in people who understand how to get exactly what they want out of the tool, not just anything
@split valve look, if i dont respond anymore its merely because i have blocked your messages, i hope you become less bored and argumentative
ok analogy time... if you had a machine that cost $100 and you buy it from amazon... And it creates gold. like 10 KG of it a day... at that point gold would have no value, because anyone could generate a mountain of gold.
your choice.
Thing is to make money you don't use the machine to make gold but maybe platinum or titanium
i can sort of see your point.
you're blocked bad anatomy.
Paladin*
most expensive-ass elements
heres the problem with prompt engineering that i see.. How do you keep the AI from making it too easy, and outputting that great thing without the prompter needing to be smart? its like a constant game of trying to be a useful prompter but the tool keeps getting better.
Can you all calm down? 
We're here to support each other, not to start wars
So i have cynical view on AI creating jobs... i just dont think it will
Almost like being a carpenter or a machinist. This is the problem with people who don't know how chairs are made talking about how jobs work.
i left milk out on the counter overnight, so, i'm not sure

what do you call an intense argument between two yoghourt producers? ||culture wars||
Reminds me of that one episode of love death and robots where the yogurt took over the world
that's grey goo you're thinking about
I loved when AI was wacky and could give crazy result that was vert creative in a psychedelic way, some time it still happen if to give "right" prompt.
the self-replicating nanobots
No there was one with actual yogurt 
Yeah. They were infused with nanobots.
boi you can still run SD 1.1 on there, 1.2, 1.3, 1.4, 1.5 and 2.0 and 2.1 all of which without fine-tuning are "whacky" beyond belief
Smart-yoghurt.
also that Hello kitty looks like it's rollerskating or skateboarding but where's the board/skates
Human want is infinite. The easier things get the more we do, not less. You don't need to worry about these things
I need to try... I forgot to test the oldest junk, hope I get the random things I thought was so funny.
hmmmm.
if none of those work, try Craiyon.com it's DALL-E Mini and it's total trash lmao it's gold for making really weird stuff
like 3-4 days ago i gave craiyon another go, and it has negative prompt support now, which allowed me to get "photorealism" out of it
Hadn't heard of it until you yanked my chain the other day.
i didn't spend very long trying but this is one of the best i got
That image was made with the free Dall-e or Crayion and I think the prompt was "Storm Troopers in Apple advert" if someone want to try improve it in SD 🙂
Can somebody explain the cow to a newbie?
Check my join date
Dude. Just upgraded my A11. This is fly like my daddy.
What the fuck does that even mean
Not a Wayne fan?
Nep is legit 16 and doesn't care to learn, don't bother explaining anything
Seriously, somebody explayin the f* cow please.
Polish cow meme
From old Dall.e 256x256 to SD (A1111) 1024x1024
Still blurry
@west tree hi, just a reminder we can see the msgs you delete in here etc-- please keep it civil with the conversation and or arguments in the future 
CRT tv
?
Better discord
Wdym copium, there are already prompt engineers and those are not what prompters sometimes like to label themselves as
This, and not "i prompt good in SD, therefore im prompt engineer"
I prompt with ai
that's just a poorly labeled Data Scientist role
they used to work with tech like Hadoop and Cassandra and now it's LLMs and text encoders.
I'm sorry. d* are funny.
@oak osprey i just took one example, the point is that prompt engineering is already a case and its not what prompters here on Discord do
Neither is that a small salary job
first one on the chopping block
what an insane salary range by the way
95k to 210k
Wdym by that? Replacing?
i made about that much at my prev job and i worked there for 6 years and when the pandemic hit, the engineering team was nuked from orbit to show that the company financials were strong
Jerma's (Currently Non-Existant) Mystery Meat Cooking Special Stream
Lol
He had a spicy one
i have a better job now where i make less but i like that i'm not as appealing to get rid of for a quick money fix
and better job doesn't always mean paid more, i just genuinely like my colleagues and the work we do
I wanted to mention this but I couldn't find the image.
I have since found the image.
Your Profile Picture reminds me of this sort of generation I did
what happened to make the frostbite on his cheek bones?
He's a messy eater
(I guess)
"shrek" yeah lmao
If I'm wrong correct me
if shrek had uncanny valley as fuck features
True
it's supposed to be shrek, the prompt was a delicious jar of shrek, next to a bowl of shrekuamole
lol
it made a shrek mogwai instead
Is this SDXL?
that's some random SD 1.5 model
I see
i think maybe double-exposure-diffusion
Mmm
i know for sure this one is
look at all the little shreks running around the shrek-hulk
"shrek-hulk combo" at 1920x1080 made that with double-exposure-diffusion
@oak osprey oh, definitelly. Good job isnt just about money
i really loved all the deformities of the high res in 1.5
so unique in terms of produced art
i honestly think i'll just always prefer SD 2.1 over anything else
it's like the kiss by which i will judge all others
golden kangaroo retriever sheppard
treebeard
my #best-of channel on my discord server is like 90% SD 2.1
these cross-animal hybrids it did in high res without duplicates made me go
and i deleted all my 1.x models
it does people without merging them, too
Lex Luthor and Snoop Dogg put together a cannabis startup
and SD 2.1 has a very charming view of Danny Devito's roles in 1985 cinema 
Has anybody tried making art of Himiko Agari on SD? my laptop doesn't have the GPU or RAM for SD
And the free sites don't get good images, I think you need a LoRA
overtly sexual stuff not allowed here
@hoary pelican please don't post nsfw in channels
Are we getting diaspora from Unstable?
Anyone know if you can create ai scenes with lots of detailed characters in it?
Is there a way to do that?
Regional prompting in A11.
You define a region and then create a prompt for it.
have you guys been able to get any good web designs ?
At work right now, but I don't recall seeing something called or related to regional prompting. Is it an extension I need?
Or is that just a method of inpainting or something?
It's an extension.
Ah ok.
Adicts refer to it as A11. Regional prompting is that hard white. That and dynamic prompting.
This I assume?
Hello @ClashSAN, not sure what's the format for proposing extensions be added to the wiki, sorry if I missed it. https://github.com/hako-mikan/sd-webui-regional-prompter It is similar in purpos...
Aye, sir. Smoke that rock.
I am out of ideas.
Boring, derivative shite.
I've got an adetailer version of that first one. holdup.
Any other good extensions to get btw? To maximize the ai's capabilities and convenience. I already got control net.
There we go, okay gens.
Get dynamic prompting. That's my favorite.
What's it do?
It allows you to {pick one option|or the other}, at random.
Little vague. Where exactly do the options come into play? In the results? Or somewhere in the setup process?
It's any keyword, in brackets, divided by a pipe. 2$$ picks two keywords, 3$$ picks three, etc. EG:
Photo of {a roast beef sandwich|your mom's twat} would generate the same image. If you added 2$$ after the first bracket, it would generate two pictures of your mom and/or a delicious reuben. Same difference.
Just kidding. I love you.
@west tree If I use the regional prompts extension and come up with something rather complicated, should I be worried about overloading the ai? Like it has so many specific things to worry about that it starts to suffer in quality overall.
Do you have the gpu to handle it? Should be fine
Yeah, an evga rtx 3090 ftw3 ultra
I guess I'm wondering if ai need to increase steps or something
Nope. Keep your step count low unless it's crushing the image.
What noise sampler are you using?
I forget. Something dps or dpm something 2m
DPM variants yield the best results, imo, but they tax your system. You may want to try unipc.
I mean it can tax my system all it wants. I want results
Unipc's good though. Similar results, just a ton more efficient. Unless someone steps in and corrects me.
I don't remember seeing unipc. Is that in the sampler list by default?
Man, those are both rough honestly. x)
just getting in on the QR code bandwagon
FYI, this was 30 steps, which is high.
good thing this in greyscale lol
SDXL so bad at anatomy currently everyone looks like they were born in Chernobyl and inbred at the same time.
very coherent.
I have a type lol
thats miles beyond SDXL
2.1
Has to be because SDXL is in 0.9 beta
If I had a faster card/newer card I could do better but it takes forever
Will be interesting to see when SDXL has a ControlNet equivalent
I barely use controlnet so wouldn't matter to me since the main ones I want never came to 2.1
The ability to draw something and then have it come alive is too good for me to give up
all I want is depth, style, tile, and inpainting. 1 out of four is all I received.
i swear they took out all nudity from the training data so it has no clue what to draw... And for artistic nude images.. thats not good. or maybe im just wrong.
Give it time and we will train that shit back in
does that mean i have to run it on my own card?
yep
Then i need money... And a new card.
hmm my 1070 is not going to fare well either.
My entire system is brand new except the motherboard just for a new video card. :/
i dont upgrade usually.. i just keep the PC as is until i have to replace it.
14 year old case was 15-25mm too little to hold any of the new cards
dunno if my zen 1 system would be compatible with new cards... even a 4060
It could have the FE but they don't make those now and I refuse to pay scalpers
not that i would even consider buying that absolute block of crap.
still waiting on nvidia to release a viable upgrade with at least 16GB
My Zen 1 (1600) I just upgraded to a 5600 and 48gb of ram (b450) and it will handle a 4090 with ease. Now for gaming, which I don't do, I need a 5800X3D, or 7800X3D which is AM5.
for this stuff Zen 1 is fine
i have a good 850 watt psu and an Asus Prime X370-pro. 32GB Ram
370, ahem
quite new bios or i could flash it to newest
not sure what it could take but it could certainly take something above a 1700X
But i dont even need a better cpu. i just want a great GPU
4gb ram but it works
Right now the iron is hot for pricing for ram, and AM4 motherboards. If you wish to upgrade do it now, as I did.
CPUs too
AM5 is out so AM4 is being heavily discounted
i was thinking of trying to get 64GB... as i already have 32GB with only 2 dimms.... BUT never in my life have i ever successfully upgraded the ram on a PC... whenever i fill all dimms with an upgrade theres a problem
five years ago i didnt know a gpu had ram at all xD
even if all the dimms were identical there would still be an issue i just know it... Like windows dont support that much ram or something stupid
Well, I had 16 then purchased 32GB and stuck them in. A lot of issues but I finally managed it. 48GB. Now if I upgrade to another 32GB (going to) I can move up in ram speed. I should be set until AM6
hmm but how much ram will windows support...
win 10?
i used to be big into Blender, and i easily needed 64GB... but now im not so bothered
you can purchase AM5 128GB systems with W10 on them right now
Never liked Blender I was more 3dsmax and especially C4D
3DS max was good, i used it in the early days, but then needed something free.
I started to love Maya then Autodesk took it over and I fled
Where can I use stability diffusion?
Autodesk known for just being greedy..
?
assholes too
yea all those professional 3D softwares are great,, i would have used them if money was no object.... i used them at university when i was doing a degree in 3D stuff
do it our way or don't do it. Okies, I don't do it
but i ended up dropping out of uni...
SD and NSFW good luck as base is all SFW
you have to reserach how to install the models to your own PC.
How?
I use Mac
i dont know, because i have a GTX 1070 so its not ideal for running SD
Let them... we need competition... Intel could come through as well.
I agree
Ah ok
The more the meerier I think
yes
How?
Go on Deviantart AI section to see what can be done.
Ok
My hope is Battlemage starts to worry AMD. Druid then begins to worry Nvidia. Celestial comes and both are put back into their place and spanked for this gen. Punish them.
Does anyone know of an OpenSource that does the same thing as Wonder Dynamics?
Nvidia will get spanked its inevitable... In time... Because Nvidia are holding back massively and over charging at the same time.
So we just have to wait.
Well, as I save for a 4090 a 5090 will be so close I might as well continue to save. As it is I can't even get the card I want or the alt card AND I really only ever purchased EVGA so this sucks
yup i think just keep on waiting... Those high end cards are power hungry monstrosities, you want something that is capable, a reasonbale size lots of VRAM and not priced to the moon...
My 850 was getting old so I just replaced it with a platinum 1300W
Yea my 850 was purchased in 2017.. dunno if thats old for a PSU or not.. But i dont want a GPU that has to be fedd by a PSU over 1000W its stupid
I would have purchased a 3090ti but not from ebay unless 700 or less because you never know and I don't play with money
man even if i had the cash i would not buy one of those highest end cards... i think its just a gimmick to see how far they can push it to the limit... i dont want something on the verge of meltdown thats the size of a breeze block
Mine is from Dec 24, 2016
Apple or intel will blow Nvidia out of the water in the next few years.. and it will be a 300 -400 watt card i hope
If you remove the heatsink the actual card is small. For me I don't care I am supposed to have upgrade this gen but not at these prices.
A 4090 is almost the price ofa used car here. A 5090 will surely be.
Wouldnt it be nice to run it with no heatsink what so ever tho.. And that would require a massive underclock and under volting.. but on a 4090ti that would still spank a GTX 1070
From what I am gathering the 5090 will be 2-2.7x the speed of a 4090 at 1.5 times the price. At that I am out 100%
I finally went water cooling and just wow
IF they keep doing that they increase the density of transistors and the clock speeds they just increase the heat dissipation needed...

what gets me is the AIO on video cards as they will break down then forced to buy a new gpu is not my idea of a wise investment
it will be another card on the verge of meltdown to satisfy rich fanatical gamers
Big think on the next two years.
Nvidia are just throwing dumb rich gamers carrots, and laughing all the way to the bank as they focus on AI data centers
Well, I don't care about games any longer as I want AI YET here come neural compression which does AI not one bit of good
neural compression is just a new way to keep you from having the ram you really need as a consumer. want more ram buy a pro card for thousands.
lag
yea.. Nvidia like to do that... try to force professional artists to pay a premium.
the latency is three times longeron a 4090. Around the 6090 it should be faster than non compressed textures. By the time the 7090 hits all cards from them will be great. Of course we need ram not shit like that
if they can compress games somehow and keep 8GB forever they would.
that is their goal
never in my lifetime
There should be an interest in making it happen, so AI consumer devices can run without the cloud.
i think they intentionally increase the disk space required for modern online FTP games in order to bias their audience to players with larger disposable incomes
and to reduce the amount of other games on their disks
maybe if the ceo of AMD and Nvidia die from old age and some new blood hit, but Jensen says he wishes to be turned into a cyborg and CEO Nvidia to 150yo
naa i dont thinkso... disk space is quite cheap or you can just delete a game and redownload when you want to play it.
Thinks should happen, new technologies should happen... Think Tesla and their optimus robot.. Elon wont want to to have to rely on the cloud for its compute, he will want it to run locally like his self driving cars
Well, I am not so optimistic as you appear to be as I know their types, and their greed.
Consumer devices running big LLMs and generative AI locally have advantages and can be always on which is critical if we are going to rely on AI technology for life.
What we need, and right now it is prime, is for a 3rd party to enter and devestate them.
imagine being in a tesla that did its compute in the cloud and you went into a tunnel and it loses internet connection
The cloud is for greedy corproations to exploit.
If Intel can sell their card for 199 then that means the hardware shouldn't be costing anything like AMD, or Nvidia is charging.
100%
yea its interesting that..
I hate the cloud since inception and despise, loathe, wish to destroy, anything as a service.
Well i think if Tesla make household robots it will force the technology to come to furition.. they will need stacks of cheap low latency ram.. it will need to run locally not cloud
That tech should cross over to PC GPUs as well
UNLESS it is all, somehow, tied in with Starlink.
naa satellite communications is not reliable,, not enough for critical stuff and to avooid customer lawsuits...
like your robot is doing something like saving your life and loses connection to the cloud.... lawsuit time
My keyboard is dying on me and some keys I have to hit multiple times. Gotta say this is my longest keyboard since the 1980s.
i have an AUKEY mechanical keyboard but it feels jankey i cant type well on it for some reason
neural link soon
That would be a great lawsuit since it was helping you it lost its signal so you died but you were already dead if it had not been there to try and help.
may look like im typing fast but my touch typing technique is haphazzard
lol
lets just hope for optical or crystal memory technology
petabytes of fast memory
I have been waiting on that crystal memory for 25 years now. wtf?!?
Things are moving fast now.
lol
or...
crazy thought... what happens if neuralink gets good.. can it run software on your brain? how much memory does the brain have
I have been saying that the age of silicon is over we need new tech as we have pushed this as far as we can, or extremely close to it. At least TSMC has now said we are almost at the end.
have your brain execute an LLM
I read a long while back the estimated amount of memory a human brain has.
its true we are almost at the end of silicon... i guess they will take it down to 1nm
TSMC said 1nm wafer will be almost 1 million dollars
@sono 2.5 million gigabytes
As mentioned in an article in Scientific American, the memory capacity of a human brain was testified to have equal to 2.5 petabytes of memory capacity. A “petabyte” means 1024 terabytes or a million gigabytes so that the average adult human brain can accumulate the equivalent of 2.5 million gigabytes of memory.
I remember that article
im trying to get an answer out of ChatGPT it is falt out refusing to give me even a ballpark figure.
give me a min to argue with it
range of approximately 10 to 100 petabytes
But thats total... so not sure how much spare working memory it has
gpt is not knowledgeable on this as this article passes its knowledge range
2.5 peta byte is probably right as our spare memory.
do remember this the human brain does garbage collection too so discards shit to make room as needed
interesting.. or the neuralink could have the memory and the brain gets to do the processing.. eithe way it would be interesting future tech
the goal is to merge with AI
I would fight to my death against that. I am me and I will use AI but I will not merge with AI.
is this a fire hazard
ah yes some people are expect to feel that way... Ive read it in the sci fi novels.. neural purists they called them
prob not, but the 5090 will be
Neither card is going to love that
you do realize gaming is actually less taxing than genning is, right?
sad
I more wait for DDR5 RAM cards to fix the damn issues with 128gb RAM
My saler doesnt even want to build that setting as of now
Because of stability issues with DDR5 above 64 GB RAM
@iron field Can you tell me the model you used to create this redhead portrait?
why is it that it cant draw fingers, technically>
is that XL?
Hands, feet are the rarest items in photos so it doesn't actually understand all that well
then you just don't understand how it works
no i dont understand it.. i want to know technically why
surely it has a massive training data
what I I don't understand is why someone hasn't trained hands and feet model
It's Photon model. I use it with ControlNet to take some of my favorite digital pieces and see them with realistic hotties
photos have never really shown hands and feet all that much. Fact is people are repulsed by both to the point some content creators wear gloves so as not to "trigger" people.
the model needs repeating patterns, and the hands are difficult in that sense, the same goes for all the facial niceties.
Is this a stupid idea... could they just put a notice out to the whole stability community to send in photos of their hands and feet, so that it can be used as training data???
it needs the connected parts too
Seriously, the image is fantastic, how did you get that magnificent lighting if it can be known for sure...
when we train we need various backgrounds, and body parts or it locks on and suddenly we gen and it will always be locked in
its terrible at anaomy in general.
that is due to removal of naughty bits
yeah, draconian measures by SAI led to this
its like that movie total recall??
Thats just not good... thats woke and stupid.. Anatomy is a massive part of art... the male and female form potrayed ARTISTICALLY is a huge part of art.
They try to make the case its just a dumb porn generator churning out specifically deep fakes to men can jerk off like cavemen... wtf is that as an argument... We have internet porn for that.
Plus all the images that SDXL are trying to generate are very artistic.. certainly not porn.
I have seen some nipples and it is really disastrous in anatomy without a doubt, but it seems that it is the stability policy, the community is supposed to fix that, the problem is that fine tuning always leads to overtraining, the models become too "hot" " 🙂
Dissapoining to hear they have to go those measures.. For an art generator..
"dramatic lighting, skin pores, subsurface scattering, (masterpiece, highest quality)"
They are not the only ones
Midjourney and especially Firefly as well
imagine if photoshop banned nudity...
They cant
@iron field thanks so much Cpt Nemo
Photoshop is different
But if they could they would.. Because they are also draconian and woke most likely.
Bet Adobe AI art generator bans nudity
It does
Yes
so going against what they have had for years which is allowing artists to express human nudity as art
Generative Fill too but its basically Firefly
moronic
Does Adobe Stock even have nude photos?
I doubt
If they dont then you couldnt generate those anyway
NOT PORN
Because their models are trained mostly on Adobe Stock
its trained on art so its outputs are artistic nearly always. its not a porn generator.
They are not trained just on art
Right,... But its outputs that im getting are always artistic..
Have you prompted for outright porn?
pretty much, but only as an experiment with keywords to try to forcec it do do anything remotely nude.. And it still comes up artistic
it wont do porn, its not traine don it.. SD removed it.
So all that must remain in their training data is artistic nude.
Thus it cant do porn.. thus even an NSFW filter is pretty dumb and unncessary.
MJ does have those filters as well and yet some dude published a image of like 20 naked and fat imams in a room touching each other
Yes, BUT I get why they did it as they had to as the lawsuits were incoming.
@split valve more nipple problems
what lawsuits?
One in London, and a few others working themselves out. This is why 2.0 was rushed and why, as emad said, you may have nude or children but you cannot have both in the model.
The lawsuit tries to make a case, but at the end of the day its not being trained on CP and its outputs are art. But i can see how some might try to make deep fakes. And that would then be a problem...
But they wont be likely to win a lawsuit against art. Artistic nude.. its not illegal.
@iron field precisely today I was testing photon after seeing the results you showed yesterday, and although it is a merge, we must admit that it produces some fantastic results.
The image without the .net reference, the light its too hard.
SDXL bodies look like they have been ran over by trucks..
LOL
That is the problem that both 2.0 and 2.1 had without using the negative tens. Is it still playing at the same scale in SDXL?
It is as simple as gun manufacturers being sued for some idiot committing murder with a firearm made by that manufacturer. In Europe the case is strong in the USA depends if in California or NY or DC or in a "red state" as to what happens. If the court is highly left leaning get ready. In London they would very likely rule in their favour not the favour of SAI. You gave the tools, and the ability, to do CP. They did CP so you are an enabler of it. Now that they removed nudes they can stand back and say we didn't enable shit. What they actually did was give us a base to train on that now has nudes. Hell, when I tried 1.5 again a couple of months ago I couldn't stop getting nudes on prompts I used in 2.x. I tried even neg prompt but no dice it was hard titty time. Volkwagen beetle with tits.
Wallace told me to use corset and that helped some but wasn't right either.
Yea its all so stupid though. Because its not CP.. its art.. Although there could be edge cases where people misuse it. So that needs to be resolved technologically to allow for the legitimate art.
they dont want to train on nude photos in the first place
You, and I, are in complete agreement. I can buy a pencil and murder with it does that mean the pencil manufacturer gets in trouble?
i personally couldnt care less about nudity generated images
yea but thats as dumb as not training a sculptor or painter student with any nudity.
I am a bit of a prude and don't like nudes but it is needed for anatomy as 2.0 was hit very hard at first with their removal.
3D sculptors have to be very well versed in anatomy.
yes
the human form
so if you want to make artistic humans you need to train the AI with the nude human form
traditional art school you gonna get a nude at some point
3D sculptors are a different thing
might be a 1500lb person rolled in by a crane but you gonna get one at some point
@boreal falcon The problem is that knowledge of anatomy and the nude body is necessary to produce correct images, even when nudity is not involved.
Autodesk wont even try to censor what i do in their software (in my case 3ds Max) neither does Adobe with Photoshop for example
similar though as the AI has to depict 3D humans but on a flat image.. its still virtual three dimensions
its also different as said
Maybe Firefly will stop censoring when it isn't on their servers, but I doubt it. It's easier to provide your GPUs and charge a subscription
anyway, this is all rehashing the same stuff from last Nov for me so I am off to try and train. I will be damned if I understand why I can train a dreambooth but can't a lora
no way to stop censoring
forget it
we are living in draconian orwellian times... you can feel the censorship on your minds on your thoughts.
Midjourney and Stability AI have also their reasons not to allow everything
people think we are free and liberated.
thats where Unstable Diffusion comes in
War is peace.. ignorance is strength. Freedom is Slavery.
whats the best way to get a bunch of 3 point perspective line pictures like this for use with controlnet
or is there a perspective controlnet preprocessor already that i didnt know about
Art
Read up on our #✍🏼|rules-and-tos
you guys ever make embeds in auto? do you tag the photos with prompts or just like chaz01, chaz02, chaz03, etc if im making one of myself or trying to
censorship by non-government entities is often a very good thing
aha, can a mammary gland be considered a subversive image worth suppressing? I see that we continue as before treating everything with the same deck sim to attend to the slightest subtlety. I already removed all my images from the stable diffusion forum for this reason, it doesn't catch me by surprise.
This image is titled "hypocrisy"
Yeah it can vary, but I generally run a batch and get some favorites
Yes, the problem is that of all the retrained models, they have their faces branded with fire, they are overtrained, it is difficult to get different people.
merges often produce higher quality results. did you not know that SDXL itself is a merge? they back-merge training checkpoints. it isn't a crutch, it's just annoying that the community only mixes and merges other mixes and merges.
#🔧|finetune better than this shit-show for asking for help with that 😄
Of course, as I have pointed out, merges produce great results, although you have to differentiate them from those that only mix, those that also retrain the model with new images.
why? good results are good results, gatekeeping is weird
Well, it's simple, when you look at a family photo of the models that have only been mixed, you begin to see very similar poses and images... a clear poverty in the phenotype to express it in another way... new images must be introduced.
merging two models together doesn't make a new model that's good at something neither of its ancestors were good at, okay
why do you have to try to be provocative
looks like a pretty young girl you're genning topless photos of?
weird, man
it is illegal in most of the world, i'm pretty sure
We have server rules here where we dont allow nsfw on the server. The rules are here #✍🏼|rules-and-tos
Circumventing filter to still post your image of a naked girl breaks this rule as well
@wispy spindle no weapons fruit?
Nsfw on the server is described in the channel I linked you
i don't know for sure but if you're posting weapons that are aimed at the viewer repeatedly that's a bit different than just posting a character of someone holding a weapon
take it up with your government, not discord moderators, they have no choice here
I never posted a naked girl, I think you are dreaming, a posted a armed girl
don't gaslight the mods lol
Silly me.
this gir is naked, seriusly?
and like, 14
Sure is.
Read up on the rules while you're gone.
Thanks
i respectfully disagree regarding the character's likely age
i was never good at telling someone's age which is why when they look 14 to me i just assume they are 🙃
Looks like a teen to me tbh
even with clothes on i honestly don't understand the fixation. or i do, and i don't want to 😬
a mistake is cool it's the pressing on and on that confuses me
Regardless, I don't know exactly why everyone is fighting the big NSFW war today-- but we have rules on the server where we don't allow full on nudity. Please dont fight mods and make it a fight of morals, ethics, or freedom and be inflammatory about it. This is the server rules we abide by and are not personal attacks on anyone.
in any case we post portraits with the clothes ON in this discord, like it or nay... at least with the SFW policy we don't have to police age estimations as much
i'm just glad i haven't accidentally posted anything anywhere it's not supposed to be. i have read some tragic tales of this
even with clothes on if they're in an overt position it's b&
thank you fruit
ananimasu at risk of upsetting you or others i just wanted to point out that your pfp is like the ultimate irony vs how you conduct yourself
you're a swell person
that is not usually the type for that pfp 😄
thanks 🙂
either that or everything is ironic and i'm a boomer who doesn't get the joke
Regardless, I don't know exactly why everyone is fighting the big NSFW war today
@wispy spindle i just double checked. mercury is in retrogatorade
She must play for the Intergalactic Basketball League.
This feels really familiar
so the img2img functionality of SDXL is so fucking cool @wispy spindle
have you played with it?
snoop dogg and lex luthor
yugioh version
using .6 denoising
My post was fully clothed but was removed. I'm not complaining, but also I don't know what will be ok in the future 🤷
Reminds me of The Boondocks, was that your inspiration?
Nvm
You posted a picture of a red haired girl in a red dress, right?
did you learn nothing from the matrix
Lolllllllllllllllllll
oh u got researchers blessing i see
grats 
yes yes
how many VRAM you need for research
i just didn't use the right email to sign up 😛
Begging era is OVER
I remember it like it were days ago
"you had your chance, fruit", our hero said, as he strode down into the volcano
that'd be a funny line to hear at the end of lord of the rings ngl
big pivotal moment
this model looks great (im testing)
"fruit, it is only a matter of time, before you, and your powerful team are no longer having the grasp of ownership on this powerful machine. we will free this beast. and it will be free!" the crowd cheers with anticipation
Blonde but yeah
Rage Against the Machine
bass guitar string snaps
A subject can be fully clothed and still be considered NSFW. In your images case, her dress was so tight to her skin it left not much to the imagination, considering her nipples were hard and really obviously stated..so I deleted it
poggies
Shoo
i think we've all established i have zero/little self-control

im sorry fruit, i mean no harm lol
it's ok i am used to admin ab00z
in my day we called them SysOps
over on the booru sites they call such quasi-booba "questionable" as opposed to "safe" or "explicit"
ananimasu is a compendium of knowledge on the spectrum of booba
Didn't even ask for them,SD pervy 
do badhands negatives even do anything
you mean the embed?
or just adding it to the neg
brooo not again 🥹 the nsfw train needs to stop choo-choo'ing
and her? after the photos leaked and everything? man, such disrespect
the embeddings on civitai
i want to believe they do something
whether it works with a particular fine-tuned text encoder on any model is up for experiment
these vectors in those embeds are pretty specific to what it was trained on, so if you can find that model, it'll probably work better there
all i can tell for sure is i see less hands in my images at all
It's just a test (photon), a LoRA, a prompt. That's not what I call nsfw.
put it in your positive 
good way to tell what it does is to ask for it
like on SD 2.1, nfixer going into the prompt makes the image all fucked n noisy, which is perfect, because it cleans up the images when it's in the negative
ohhh you know what would be sad
hey fellers
what if badhands removes all the bad hands from your model. but then there's no hands. because it has no good hands 
im sure one of you can sort out an issue im having with a1111
dont think its major enough to warrant tech support
but it's aggravating
there never were any good hands
dood that's what it is then.
this is a damn 1024x1024 image and its 90% rendered and the preview looks like PISH
the end result looks fine but the previews mid-render look awful in a recent update
you need a goodhands embed
put goodhands and badhands and let them fuckin fight each other
i need to see this
I NEED IT

let them fight a thumb war
oh man that screwed my damn eyes up to look at that lmfao
why is it like that Jordan
😭
nah its 100% zoomed
looks fine if you like stylized stuff
my damn eyes
hmmm i haven't updated and mine is old and when i ran that, it had blurry-ass live previews
theres an option for better preview quality
2 secs

glad to solve it for you
"AI doesn't make you more creative"
yeah right who else could come up with fievel going bankrupt
me i couldve done it
you needn't do it
you can really feel the bankruptcy in this one
oh i thought the building was boarded up

It is however what we'd call it
Please refrain from posting suggestive poses
You make me afraid to post now. Is there something you would show me that you are saying is a suggestive pose?
think that's more of a textual inversion thing tbh
the text encoders are responsible for a lot of hand deformities
stop using sd to make anime waifus then
post cool stuff instead
You are new around here so I will let that pass because everyone around here knows how much I hate that shit.
I don't use SD to make imaginary lovers so I can rub one out over either.
oh, really? I think not as my definition of NSFW and that broad shit I just heard without anything to say this is a nono is vastly different. I see shit on here I let slide but I would consider NSFW and they apparently don't. SO, IOW, cope it,
yourself.
Anyway I really have nothing to say to you since I was asking the mod you just sort of added your half a penny insight onto.
when this message gets through then my internet will have returned
testing
testing
123
ok internet is back
mech.
MECH
The image was a girl in a dress with both of her legs spread out and open- crotch center in shot. That is why I called it a suggestive pose
Oh, ugh. :/ Yeah, waaaaaaaaay over the line.
LOL
I dunno, man. Fruit mommy's discipline earlier may have awakened something in me. I might have developed a taste for boots.

looks like watermarking might be fixed..
I like to upscale 😌
bruh
Let's call it an easter egg!
NSFW capability is critical to proper anatomy, but I am mad critical of anyone who makes extensive use of it, particularly for deepfakes or super creepy shit, like weird teen or futa crap. I'm a bit of a hypocrite... I had an initial curiosity about NSFW possibilities, as I have a d*, but that was quickly satisfied. I don't understand how Unstable or, to a lesser extent, the Mage discord exists as a forum for people who recreationally make this their ply and trade.
Heya Bad, do you got any tips for fine tuning small details of generations?
Really think about what you want to see, and realize you're not going to get all of it. Prioritize weight accordingly.
Like if it looks really good and I want to keep it but it has a small annoying anomaly like say a malformed foot with fused toes or something. Or a distant character that's rather garbled.
Inpainting. Mage is excellent for inpainting, in some ways better than A11, or at least better for getting a mask.
Mage?





