🏞｜general-with-images

[smoky oak]

currently at 1500 words, looking to double it by release

[wispy nest]

do we get a1111 working with it?

[daring kernel]

What happened to your gpu

[smoky oak]

it works, but its not ideal

[smoky oak]

its too much to get into. Still an ongoing process

[wispy nest]

oh boy, i dont want to use comfyui

im too used to a1111

[daring kernel]

Me either, looks complicated

[wispy nest]

yeah i just didnt like the interface lol

[daring kernel]

It's very scientific

[wispy nest]

haha

[smoky oak]

@wispy nestmy workflow for SDXL gets better results than auto by using a new method of diffusion I worked on with some people here. It is also very easy to use, as I will be giving a beginner and advanced workflow for those who wanna toy around with it

[daring kernel]

Oh

[wispy nest]

ok but which UI is that? are you refering to comfyui?

[smoky oak]

Comfy, yes

[wispy nest]

uh oh

[smoky oak]

it will all be documented and setup for use out of the box

if you can use auto, you can use comfy. Its basically identical with how I have it setup

[wispy nest]

but hold on, dont we get support for a1111 down the road?

[smoky oak]

you add a positive, a negative, and image size

[smoky oak]

SDXL already runs in auto, but not with my diffusion method, which is faster and also looks better

Only Python Diffusers and Comfy UI have my workflow implemented

[wispy nest]

i thought that was a 3rd party extension that makes SDXL work on A1111

[smoky oak]

oh, if you mean the one I think you mean, then it legit just loads a compy of comfy UI in auto lmao

I have seen some people use that

[wispy nest]

i heard some people mention A1111 will have newer version natively compatible with SDXL 1.0 shortly after SDXL release

[smoky oak]

yeah, but its not gonna be the ideal workflow

although, it seems liek 1.0 may run different, so we will have to see

[wispy nest]

i can wait for a few weeks till they get a compatible version

really don't want to switch from a1111 at this point

the thing about A1111 is that i have several usability features I'm used to and comfortable with and it'd be shame to ditch it after having used it for so long

[daring kernel]

Now that I think of it

Comfy UI is a very ironic name

[wispy nest]

im sure its not that complicated to use it but i dont like the idea of switching from one interface to another after getting to used to one over a period, just cause a new version of SD is out, don't seem fair and strategic

[daring kernel]

I understand

It is like moving to a different house

[wispy nest]

you know, it has that kind of feel lol

[smoky oak]

The only reason its called Comfy is cause the creator is Comfy Anonymous

Just like A1111 is made by Automatic A1111, and VLADomatic is made by VLAD

[calm canopy]

what if he named himself after the ui

[smoky oak]

Then that would be interesting haha

[last cosmos]

long neck lmao

#🍥｜anime

[smoky oak]

Currently losing my mind over this GPU

[rich sky]

Help

Help. Pls

It dissapeared now

Like crashed

[livid crystal]

Alr. What model

[livid crystal]

Tends to work better linux

[plucky tartan]

anyone trying to make loras? i wrote a guide on how to make them in 10 mins w/o code or scripts lol

https://lorai.notion.site/Making-LoRAs-with-lorai-art-006e398a5ed34ce2a145537696dc9d89?pvs=4

[rich sky]

I got windows

[livid crystal]

I can tell

[shut sinew]

#🤝｜tech-support

[strong sonnet]

you wanna share it with me?

[chrome kite]

It depends on what you want. I’m not an expert, but what I’d recommend for photography is Realistic Vision 4 or I Can’t Believe It’s Not Photography. I use Dreamscaper for less realistic images. I mainly make photography so I don’t know much about the other types of models.

[crisp horizon]

Hi all, anyone have good prompts??

[grizzled sage]

naw. proper configuration is what counts. linux benefits in a few areas but not for a beginner just getting started. it's not friendly for beginners especially in this realm.

[raw quiver]

Almost perfect text

sad second s there

[rain gazelle]

nice!

[dense solstice]

It was your post on reddit?

[languid pagoda]

Why are you making me hungry

[daring kernel]

Cool shit

[daring kernel]

It even has tomatoes and everything lol

Those tomatoes are good

[stoic pivot]

SDXL Bot is a bit out today, I still have time to go out get a pint.

[chrome kite]

weird image but kinda cool

[jolly grove]

I don't know if I'm in the right topic, but does anyone know if there is a model that does a color treatment, with the example below?

[grizzled sage]

the tomatos look like jalapenos

[rain gazelle]

i can add jalapenos if desired

[grizzled sage]

nomnomnom

[ripe cedar]

My face detailer just gets skipped. Anyone else having this issue?

[livid crystal]

I use both and i still get foul images

[livid crystal]

That is debatable. Personally i have never been able to get stable diffusion to work on windows

Which with Linux and a less intense DE it seems to launch fine

[warped basin]

#💬｜general-chat message

This image looks much more realistic than

This image because the background is not blurred out

I see people do it on civitai community page but I have no idea how

[dry crow]

something like these?

you need to use an realistic model, then dont prompt for depth of field or bokeh effect. That will make the blur. Use tags like photo, portrait, realistic, and for the details you can use the add_detail Lora or a latent upscaler.

[odd hearth]

didn't they make a flipless one??

[odd hearth]

if you find one lmk. I'd like to apply that to my moebius style stuff

[odd hearth]

NE'FLIX

[acoustic moth]

been trying to figure out the same thing

did you try CS1o's recommendations?

[wispy nest]

photo of a beautiful girl, 22 year old, hazel eyes, (highly detailed face, eyes:1.3), neon street lights, (blur background:1.1), (upper body shot:1.3), (highly detailed:1.3), HDR

[acoustic moth]

were trying to avoid the blur

[wispy nest]

you want to blur the background and not use blur in prompt?

[acoustic moth]

no we dont want a blurred background at all

[wispy nest]

oh

[acoustic moth]

we want it like this first photo

[wispy nest]

then you don't use blur in the positive

if you do get blur maybe mention that in negative

[acoustic moth]

yea it usually comes out blurred anyway

im thinking of resetting my stable diff add ons, might be some extra thing i added

[wispy nest]

possibly

i never get blur since i use a default style i wrote that negates blur

[acoustic moth]

you wrote it into the negatives?

[wispy nest]

yeah a prompt

could use something like blur, blurry

[acoustic moth]

i think ive tried specifying deep focus in the prompt, can’t remember if i tried the negative route but ill give it a go 🤝

[wispy nest]

should work and gl

ok check Deep focus is a photographic and cinematographic technique using a large depth of field. Depth of field is the front-to-back range of focus in an image, or how much of it appears sharp and clear. In deep focus, the foreground, middle ground, and background are all in focus.

i was checking if that blurred background from foreground but it doesn't

[acoustic moth]

yea it should cause everything to be in equal focus

[tawdry stump]

@small hearth https://media.discordapp.net/attachments/1128636013573574686/1131336054637600940/00149-1558615466.png?width=895&height=671

[grizzled turret]

So this is 'Emma Watson' in Horizon Zero Dawn in 1.0 via ClipDrop

Celebrities still seem to be frankenstein versions of themselves

[split valve]

it cant draw.. it has no idea of anatomy.. bone placement, muscle placment, limbs, fingers.

Its progressing, each model is better and better, but im not going to over hype it.. at the moment its like an artist that has studied a lot but never studied anatomy, and does not understand the concept of hand or fingers.

[wispy nest]

lol this tech is still in its infancy btw

not much older than 6-8 months, but so far it's pretty interesting what it can do

[split valve]

I agree. Its good at understanding concepts, and producing abstract art and maybe faces to an extent.. But beyond that its still learning.

[wispy nest]

yeah and your comments are helping towards focusing on how to make it better

[split valve]

i doubt, it, theres a lot of geniuses working on the tech.. But actually my comment was wrong, it does not understand concepts.. Not properly. Thats the thing with AI, can we get it to truly understand the concept of something.

[lilac pilot]

eventually

[wispy nest]

obviously not at the current stage of development and status, but its getting there bit by bit

one disappointing factor about Ai development is how politicians force to censor it and and adapt to the human cognitive limitations

[grizzled turret]

Which is why open source is important

Because that's where the community steps in

[split valve]

I've never whitnessed a poltician do anything right. They should not be in control of science or engineering development what so ever.

we are at the crossroads with this AI tech and our civilization.

[wispy nest]

i'm certain we are not yet seeing exponential pace of Ai growth

because of all the social noise

[split valve]

The scary thing is to watch AI get really good at the jobs people want to do, but get almost nowhere on the jobs people dont want to do.

so much for "learn to code" more like "learn to plumb"

lmao.

[wispy nest]

well thats a vast area of topic, lot of the jobs are not meaningful and are just to make a buck, while a lot of the jobs are dangerous and inefficient, Ai can help in that area at a massive scale, but that will also remove human labor

[split valve]

imagine AI doing things of actuall value.. Like "hey AI, go down that mine and get me some gold and diamonds", "hey AI, go and plant me a supply of crops and maintain it"

"hey AI go down that sewer and unclog it"

[lilac pilot]

They cant censor I imagine its too late. Open source millions of people already creating stuff

[wispy nest]

well that's the idea of it, Ai and smart machines will eventually have to manage the production and distribution chain, rendering human labor obsolete

[lilac pilot]

The box will never be closed.

[wispy nest]

there are key factors behind why Ai has to automate things, 1) safety, 2) efficiency 3) freeing society from boring mundane jobs

[split valve]

AI rendering human labor obsolete, i shall believe it when i see it... "hey AI go build me a house", grow my food, and generate electricity for my heating, and maintain it"

All the jobs we thought computers or AI would conquer first, it has not conquered.. The things we thought it would not conquer, like Chess, Art, Music, programming, its started to conquer.

[wispy nest]

Ai is still heavily censored in social landscape

[lilac pilot]

Playing off a cynical approach do you thing the world will make it to the point of robots reaching that level anyway.

[split valve]

Well if they did reach that level of robotics, i worry about the military implications of robot armies.

its doubtful we'll make it without global conflict.

[wispy nest]

there is a fear among those in the control structure that Ai could collapse current social paradigm

[lilac pilot]

I mean the earth in general. Do you thing Ai will save the earth?

[split valve]

I think the Earth its just fine, and will outlast us

I dont think AI is particually censored... or if you think it is, say in what area

[wispy nest]

but i should just add here, in the next 5 years we should be seeing massive changes in the economy, jobs, research and medical fields that could revolutionize the way we operate in society

[split valve]

It will be taxed and politicians will require a licesnse fee, no dobut. because Governments fear losing power and control over normal people.

[split valve]

i agree but i dont think we'd see much in the medical fields... Because the current buisness model, status quo would have to change... That is making patented drug treatments to make big pharma richer, whilst ignoring ethics, morals and the goal of finding real cures as much as possible.

[wispy nest]

right, that will change by how Ai will change the economic model

[split valve]

Theres no money in people being healthy.

[split valve]

maybe, i agree the global system will have to change a lot. but big pharma have a lot of power.

[wispy nest]

those critical factors are rooted in our predatory economic system

[split valve]

Im going to be a cynical realist.. My prediction is a future of human suffering, where the rich rule over the poor, and the poor do all the manual mundane jobs,... think the movie Elysium.

[wispy nest]

thats one side of the possibility and it would matter which path the collective chooses

and to realize what those two sides are 1) thrive 2) collapse

[split valve]

open source will help the collective to chose the right path, and naturally they would chose the path because people are in general good natured as in they want a utopia...

Its the minority that want to rule and control the world.

[wispy nest]

that's another critical aspect of the conversation, education is important otherwise fear of Ai will lead to missing the opportunity to divert society from collapse

[lilac pilot]

The earth will fry us and robots will be the next life form

[split valve]

in a world where AI puts everyone on a level playingfield.. we should be able to eliminate the evil, selfish, greedy individuals that want to control everybody

[wispy nest]

ideally yes, but there are other sides too how we can ethically leverage on Ai

[split valve]

The current society has to collapse, to make way for a new one.

[lilac pilot]

No way this will increase those things. The rich will become 10x more wealthy, greedy and selfish.

[split valve]

All men are equal, and with AI they are even more equal.. lol.

[split valve]

in a post scarcity society, monitary wealth does not really have much purpose.

[wispy nest]

Ai can definitely lead to mass scale automation, removing all obligatory human labors, along with this technology we can pursue a new economic model like a resource based economy, those are the few things that will have to work in conjunction to change the course of society from decaying

[thick palm]

Can you please throw at me some examples of text generated by SDXL?

[split valve]

should be possible, if we can master getting off planet cheaply... then we have unlimited resources

[wispy nest]

you mean colonizing space? that will eventually happen but not by neglecting our current habitat

[split valve]

colonizing it eventually, but my primary goal would be mining for resources.

[wispy nest]

maybe

but i wouldn't think of monetizing resources a constructive plan

[lilac pilot]

Are you prepared?

[split valve]

well i mean, input huge amounts of resources into the economy like gigatons of precious metals and such from asteroids or the moon, and then it creates abundance, those metals plummet in cost

[wispy nest]

that could be a good idea lol but that will also take a lot of resource to travel interplanetary hunting for resources

[split valve]

i mean NASA say they're going to mine an asteroid worth like 10 septillion dollars or somehting crazy

[wispy nest]

earth has good reserve on resources and have the capacity to support twice the current population

[split valve]

China think they're going to strip mine the moon,

[wispy nest]

but the current socioeconomic structure is depleting those resources in a wasteful and inefficient ways just for profit

[split valve]

Resources are potentially nearly unlimited though right? with enough technology?

[wispy nest]

not unlimited but we are not managing what we have intelligently

[split valve]

such as what ?

[wispy nest]

all of the current trends of the corporate and consumerism

[split valve]

Yes we are wasteful, but i think technology solves all the material resource problems.

[wispy nest]

technolgy / Ai is our last big chance to prevent mass ecological collapse and not just social collapse, if we can utilize the technological potential effectively society can literally usher in a golden age

[split valve]

Well i cant predict the future... but i dont think human nature will change.. you can give humans more technology but we'll still have the same nature.

[wispy nest]

true. society at present is in a bipolar state

[split valve]

its like wakanda or something from those mavel movies... you can have a society with crazy tech level but we'll still end up going to war or some dumb shit

[wispy nest]

there are other factors to social conflicts besides the economy

there are hard challenges for society to overcome

well for the lack of a better term or not, I'd say you can never put human qualia into machines, that's not possible

[split valve]

Technically they could be, but if we wanted robot servants we'd limit the compelxity of their minds so they are not sentient.

im hoping for one of those NEXUS androids from blanderunner.

[lilac pilot]

Yes!

[split valve]

its what all men want. And maybe women.. a perfect supermodel android companion. That is the holy grail commodity of the AI future. lets not beat around the bush.

lets not pretend that people want a boston dymamics dog robot that can load the dishwasher

[wispy nest]

there are many fascinations revolving around Ai and how much of our inputs can shape that, but if you are considering Ai to become sentient, that's similar to hoping for a pet dog driving your car

[split valve]

define sentient then.

[wispy nest]

matter of having consciousness, machines are void of that

[split valve]

But define consciouness

[lilac pilot]

We never know. Right now its really considered augmented , but it will be full fledged AI

[split valve]

its a hard problem right?

[wispy nest]

you can study the papers on consciousness and how it influence all things living,

[split valve]

Maybe some super LLM that has read every text and seen every image and video, can define exactly what conciouness is, then it can set about creating an AI that does embody it

its probabbly just an information processing system

[wispy nest]

it used to be called hard problem cause science hit a brick wall due to the limitations of physics nature of things, but now its more researched and taken seriously by larger science community as the fundamental essence to our reality

[split valve]

Ive not studied that but i am guessing that its theorized as some kind of quantum information processing system that is sort of everywhere, and that emerges within our brains.

[wispy nest]

wel on the contrary its not an emergent product of anyone's brain, consciousness is fundamental to reality, and our brain are able to perceive it

[split valve]

thats sort of what i meant. i guess the theory is consciousness i everywhere, and our brains have the ability to access it? rather than create it.

[wispy nest]

yeah to our current scientfic understadings

[split valve]

Thus if an AI is on a microchip or software, then it can never access consciouness..

unless its linked to a brain, with neuralink

[wispy nest]

brain chip can augment your brain ability in ways that are not native to your brain now, but the chip itself cannot be you

[split valve]

it could potentially steal or read information though, and access consciouness. so if true that sounds a dangerous technology... Like a parasite the nerualink could seek control

[wispy nest]

yeah, and mind you when you talk about stealing your cognitive information it involves 3rd party wanting to access what you have that he doesn't but through technology its quite possible to tap into the source of information for anyone wanting to steal anything from you

[split valve]

so then we'd have factions that dont believe in keeping any information a secret and will want a hive mind.

its all out of sci-fi novels.. all been thought of before.

[wispy nest]

singularity is not a new concept

[split valve]

its so hard to predict which way civilization will go though.

[wispy nest]

idk

im not bothered with it

[rain gazelle]

Today's Tifa.

[wispy nest]

is that SDXL?

[rain gazelle]

naw

[wispy nest]

aww shoot, im wanting to try out SDXL 1.0 but can't have access to their models

[rain gazelle]

haven't really messed with sdxl yet

[wispy nest]

huggingface link is 404 dead

on wait its up now

https://github.com/Stability-AI/generative-models

[fossil gyro]

can you use the refiner on ComfyUI

[stoic pivot]

Some time in my nightmares AI give me

[honest tree]

Hi

How would you prompt this?

I'm not getting a hammer no matter what I try

[wispy nest]

easy way to find out, load that image under Img2Img and Interrogate clip

if you are using a1111 that is

lush green valley, water streams, cozy cottages, night sky, twinkling stars, fairy tale theme

gotta love the quality of the sdxl render

[honest tree]

Okay I'll try tomorrow and tell the results

Thanks!

[lunar mantle]

Is SDXL 1.0 allowed to be used commercially?

[tawdry stump]

should be

[toxic pulsar]

Generate comic characters

[last cosmos]

what the hell man 😭

dude what is this 💀

[split valve]

dunno what is it? is that crimes commited inside a video game?

[last cosmos]

roblox sentenced me to death man 💀

[chrome kite]

done with ICBINP

[split valve]

MattVidPro says SDXL cant come close to DALLE-3 preview... But i quick test shows it easily can

[ripe bluff]

my fucked LORA

[boreal falcon]

those will be interesting times in the coming months

[muted mulch]

If you see this in the forest, what is your reaction ?

[stable abyss]

This seems harmless

[dull cosmos]

I posted my images on Adobe's Discord and was informed that sharing work generated by other engines is not permitted. Isn't it interesting?

Could it be a hint of apprehension about acknowledging the true potential of FREE STABLE DIFFUSION?

[quick current]

wicked

[boreal falcon]

Nah

You cant post DALL E either etc

And its not free

Also it was Firefly server, they only have to deal with FF there ^^

And they both have different target audience anyway 😄

[dull cosmos]

yes I am agree with you

[boreal falcon]

Its always interesting for me to know how they use this tech and how they workflow lools like 😄

[boreal falcon]

100 people, 100 different workflows

[west tree]

Fecunda Ratis

[willow field]

just tried sdxl 1.0

thoughts?

the prompts are in png info if yall wnat

want*

[split valve]

incoherence is holding it back

[muted mulch]

Cute monstruosity

[hollow vault]

"Colors of Bangalore (India), Embracing the Beauty of Famous Spots Through Artistic Vision!"

[tawny forge]

GTX 1650 --lowvram SDXL 1.0

[mystic ledge]

why does my donkey kong look like this

[surreal thistle]

Some fake Chinese movie posters I made.

[cyan snow]

hey guys, I'm not updated- is 11.8 still faster than 12.1?(CUDAtoolkit)

[jovial umbra]

i die and go hell

[left sparrow]

Today i generated some ai lineart and then painted it with my markers

[daring kernel]

And then he sings the halo 2 theme

[steel apex]

"Master Chief, mind telling me what you're doing on the toilet"?

"Sir, big bean burrito"
reverb fart

[rugged dagger]

3d geometric MPAP

[grave flax]

yall gotta see my grilled cheese

[unborn aurora]

Very work in progress, but, at least it's semi legible.

[grave flax]

almost looks like nvidia canvas?

[unborn aurora]

I'll have to check that out, control net with normal+depth+canny from a 3d model.

[grave flax]

oooo cool

i wanted to try canvas but... i have a laptop

[wispy nest]

SDXL 1.0 base model, honestly... disappointing result

not a very complex prompt either

[proud dagger]

this is one of the cases where refiner is still needed - small faces are the real killer

[wispy nest]

could be algorithmic issue how the engine parses because prompt wise its not prioritizing the face any less, its that the image resolution is large, and we have other factors like a dog and sea beach. its the same issue with previous SD models too

[wild sorrel]

https://stable-diffusion-art.com/adetailer/
This might help

never tried it tho, usually just going for closer views

[wispy nest]

thanks, will take a look

[wispy nest]

all the faces across different models are coming out nicely, wonder if that's also compatible with SDXL

going to put that off for later

actually i'll try one render to see if its compatible

no glitches, seems to work ok!

[unkempt maple]

what's this

[west tree]

Testing a Lora without a prompt and it gave me the Ayatollah

[west tree]

Small faces are the little-death....

[calm canopy]

@snow pewter you selected negative prompt here right? This is in "Settings" -> "Live previews"

[snow pewter]

yes, I selected negative prompt

[serene dragon]

Upscalers go in this folder to improve image resolution

[unkempt maple]

well shit

I'm using invoke rn

[calm canopy]

hmm strange, selecting negative prompt and saving the settings should work 🤷‍♂️

[unkempt maple]

and it doesn't even want 2 work

[rugged dagger]

up/down aging works great in 1.0. This was an older man^

[split valve]

here are my negative prompts.. when generating humans. disfigured,mangled, burned, scar, broken, deformity, alien, animal, reptile, amphibian, insect, limb, blur, smudge, incoherent, ugly,duplication, clone stamp, duplicated, crayon, draft, pixelated, rough, drawing, food, deformed, squashed, smear.

The humans still turn out mangled... but ah well.

[west tree]

From my upcoming music video

[rain gazelle]

5760x2400

[nimble geyser]

those feet are on point

[surreal thistle]

So apparently this is my single most popular AI image. It received 37 likes in a single day on Playground AI.

[surreal thistle]

Are you a Mario Pasta fan or Luigi Rice admirer?

[rugged dagger]

playgroundai.com just launched filters for SDXL 1.0 !!!

They are so cool !

[wispy nest]

wouldnt it be cool if we could select an area on screen over a video playback and let Ai trace and learn about characters in few mins of video playback?

that technology probably already exist, and general public don't have access

[rigid fractal]

so far I am extremely impressed by how easy it is to prompt with SDXL

[sturdy mica]

DreamshaperXL10_alpha2Xl10
Prompt starting with: Portrait of / Wide-angle shot
1024x768
2x Upscale with Ultimate SD upscale + R-ESRGAN 4x+ Anime6B
Cherrypicked from 30 or so images and sharpened with Photoshop.

[nimble geyser]

I like the analog portrait option

I'm curious to see where it's at in a few months with all the loras and new tools people will come up with

[wispy nest]

HI, am unable to makle DS create a UFO coming out of a big Digital Signage Screen on Times Square, the AI does not get the "emerging from the screen" part... any ideas? Here is my last attempt prompt: "a UFO is emerging out of a huge Digital Signage screen, 50% of the UFO are not visible yet, because still behind the glass, on Times Square"

[grave flax]

not safe for the dinner table

it happened again 😭

[acoustic moth]

is sdxl the one that has a ui like this? im wondering what this is and if anyone thinks this ui is better than the default prompting workflow, seems like it gives you much more control

[neon crystal]

Just an appreciation post to Stability AI team. THANK YOU for blessing us with this release WHOA 🤯

[charred ridge]

That looks like a ComfUI workflow you can also use A1111. But the experts say to go with ComfyUI cause its more versitile and powerful. SDXL is just a model lyou can use with several different interface programs

[split valve]

i give negative prompt of face, people, character, eyes. And it still gives me images of faces.

[stoic pivot]

They fixing the sad bot now?

[chrome kite]

How did you do that? Nice job

[muted mulch]

@chrome kite https://civitai.com/models/118536/voxel-xl

[west tree]

Experimenting with higher resolutions

[grave flax]

not ai i just want to share my silly kitty

[west tree]

https://tenor.com/view/hot-dogs-gif-20437077

[grave flax]

he is one big hot dog

just tried to put my face on his tummy and he bit me. cant say i didnt deserve it

[stoic pivot]

Techical plan drawing of an automatic cat thrower by Leonardo da Vinci, it may solve the lazy cat problem.

[grave flax]

😭 omg

we adopted him fat but we may need this

[stoic pivot]

Oh no no no no, you never throw your own cat....

[drifting flower]

Merged characters generated with SD XL

[stoic pivot]

Cool, I was also a bity into dark horror, mine became... "the screaming horror ginnungagap, dark smoke, death gue".

[stoic pivot]

I need to show this. It is a old photo I took over ten years ago in Stockholm Sweden, I described it for SDXL 0.9 (the bot here) with:
"a narrow street with a slight bend in a Swedish old town with a few buildings on both sides of it and snow on the ground", I got it on the 3rd generation

I need to point out that many of my images been on Flickr as CC0 and they later been spread to other sites as "public domain", so maybe maybe I helped with inspiration for SDXL Stockholm images 😄

[stoic lotus]

__Crystal-Clear XL - Photorealistic __

[stoic pivot]

Another with original image and the AI after the description that failed in a good way.

[split valve]

The incoherency is beyond a joke... this is with lots of good negative prompting.. also does the same without any negative prompts.

[rugged dagger]

I've had major struggles with characters on the ground. Whether it be yoga, pushups, situps or janitors laying under a desk scraping bubblegum. There's something about characters not standing up that the AI hates.

If I try to get this janitor to lay under the desk the AI murders him and rips him into 50 pieces.

[split valve]

and if you ask for a person walking on a beach they will be waist deep in sinking sand

[rugged dagger]

Yup. I find that controlnet (control traits) do help. But it sucks we cant use controlnet with image to image.

[split valve]

i wonder if it was lack of training data that caused this...

[nimble geyser]

yes. well, lack of ways to quantify it visually

[split valve]

yea, i think it lacks comprehension of specific bodyparts and the comprehension of anatomy. how do you train an AI that

it does not know its ass from its elbow

[nimble geyser]

yeah. it has no concept of what anatomy is. so it's tricky with things like that

[rugged dagger]

Well I presume (complete speculation) part of it is PhysX in games and other software. It's needed to create collision and gravity mechanics. But I find words like "add gravity to scene" help. Animals on the ground work great. But something about humans.... just dont know.

[nimble geyser]

might have to do with us wearing clothes

so it's visually inconsistent

what we look like sitting on the floor or whateve

[split valve]

i dunno. seems they would need to combine a large language models understanding with the diffusion model. would be a lot more processing.

[rugged dagger]

They have access to pose bone structures from sites using controlnet like playgroundai.com.

[split valve]

GPT-5 they say they will train it with videos. so it will understand the concept of everything, such as walking... from that knowledge it should be able to generate images of somebody walking

but i dont know all the theory behind it im a layman when it comes to this

[rugged dagger]

Or maybe just do a massive deep learning voodoo thingy that maps bone structure in all images of humans they currently have.

[nimble geyser]

I think AI tech will evolve into modular specialized models that can compliment one another. so down the line it wouldn't necessarily need to be a large monolithic text to image large language model

it's already beginning to go that way

[split valve]

i dont think they do anything special... They just feed it a gigantic dataset of images that have basic captions along with them, and then it diffuses random noise on the data based on your prompt.

Its not an actuall AI. it has no comprehension.. its just an algorithm

[rugged dagger]

We're already working with AI generated microvoxels. But we had to import additional software for bone structure. It's mind blowing.

[nimble geyser]

well AI might be a misnomer. but then if you want to get philosophical, aren't we just diffusing random noise in a sense?

but yeah, the general pubic seems to vastly misunderstand how it works. I know I did at first. still can't say I understand how it all works exactly, but it's definitely been demystified for me

[split valve]

is it truly random noise. i dont think so. i think its special noise. because it can be reversed by following the chain back.

its more like compression noise.

[nimble geyser]

I like how this prompt encoder allows me to adjust aesthetic score on the input. I'm going to see how that actually works. curious how it'd be able to have an impact on aesthetic score without loading a model to calculate it

ahh, it does nothing, lol

what sort of things are you guys doing with sdxl? I have no idea what's actually optimal. I just try lots of things and sometimes they work. but not the most efficient approach

[stoic pivot]

Midnight server bunny.

[tawny timber]

/private

[grave flax]

heyyyy well that's something!

i've never trained a lora before and this was just via a google colab lora trainer but i'm kinda proud it actually worked to any capacity

if anyone wants to help me out that would be greatly appreciated (heads up before anyone asks for the file, this artist does a lot of gore/guro and body horror)

just figured out how to use civitai nevermind that first part

oooo it's progressss

for some reason it shows up darker in the photo gallery program

for some reason i'm stuck on the 'start upload' part of civitai
edit: nvm we good

[odd hearth]

that's huge for a LoRA

I mean I am gonna run out of storage if the new lora standard is nearly a gig

[muted mulch]

Yeah, 1Gigs for a lora is so heavy, i hope they will found how to reduce the size

[grave flax]

i didn't even link it yet bruh

[boreal falcon]

DALL-E 3 seems to be big this time as well

[odd hearth]

dalle is infinitely more expensive than SD. If I had to pay dalle prices for what I've done with SD, I'd be out like ten thousand bucks

not to mention you can't customise it and you have oai breathing over your shoulder

[odd hearth]

that's a photo

[boreal falcon]

@odd hearth thats true ofc but as with other alternatives to SD they serve their purpose and have their target audience and niche of users, although it will be hard to compete with those as well

[odd hearth]

I rememebr being on the free beta and I'd burn through my allotment of daily images in like an hour. super frustrating.

SD is so amaze

[boreal falcon]

Well what was it, 15$/115 credits

But it doesnt serve the same purpose as SD, nor do the same people use those

DALL-E 3 will be basically replacement for Midjourney and complement for Firefly

You pay as you go, not like in Midjourney

[odd hearth]

yeah the credit thing will mean it can't compete with mj

to say nothing of SD

I don't think OAI really cares about Dalle as a standalone product

so they aren't making it competitive

[split valve]

It wont replace anything, it will be more akin to an outdated economic model desperate for relevance against the looming steamroller of Open Source technology.

[boreal falcon]

For a niche of users its a better alternative than MJ tho

[boreal falcon]

Open Source is underrated and overrated at the same time

Its hard for them tho with MJ, SD and Firefly being here

Because the way i see it DALL E is best suited for people who are into MJ like generator but with money saving if they dont use it as much

For MJ you pay 10-60$ no matter if you use it only few times

[split valve]

DALLE would work if the technology was not known to the public. If they had the secrets stored in a vault and patented. But its not. The cat is out of the bag, people can even run it on their own computers.

[boreal falcon]

lets see how it will go, but they better bring even better images etc.

im unlikely to switch from Firefly to DALL-E anyway

or use it as a "company" to it

[split valve]

well it will just carry on the way it happened with the first DALLE.... Because StabilityAI have a roadmap and plan for continuous improvement. SDXL is not the end.

[boreal falcon]

they all have that 😄

[split valve]

im not so sure, DALLE is a for profit model, and OpenAI need to justify its cost. They might not have even made DALLE-3.

They are not a generative art company

[boreal falcon]

Its not their focus probably, GPT is

[split valve]

true

and this is a snipped from their mission statement. Our mission is to ensure that artificial general intelligence—AI systems that are generally smarter than humans—benefits all of humanity. Read our plan for AGI Read our Charter We are building safe and beneficial AGI, but will also consider our mission fulfilled if our work aids others to achieve this outcome. Our work Research

DALLE-3 is just a side project technology demonstration...

A company whos buisness it is to make digital art tools like Adobe they are sure to steamroll DALLE-3 by OpenAI

[boreal falcon]

Its hard for DALL E at this point to compete with SD, MJ and Firefly (and the unreleased Picasso as well likely) at this point. All 3 of those are collecting their target audience and niche of users

If that is even what they want

Midjourney may end up being even cheaper for consumers, especially if they use it a lot. The only thing is...they dont have inpaint etc. yet.

Stable Diffusion is free if you are not into Dreamstudio and Clipstudio and you have a lot of customization options to play with with all the add ons, custom models etc. It has its drawbacks definitelly, like its setup as well as need for a at least decent PC and its de facto not really commercially safe for example.

Adobe Firefly is free...but as it gets out of beta not anymore and will remain a tool for those who are Adobe customers anyway and is now already highly customizable with even more features to come and more importantly...natively integrate into Adobe apps which is a huge plus for those who work in those programs anyway. And its commercially safe but since its in beta you cant use those for that purpose now and you have watermark unless you are in generative fill in PS directly. A big drawback for some users or consumers is that its dataset is limited to what is legal for them to use

Picasso we have yet to see (by Nvidia

[wispy nest]

fingers fingers fingers

[grave flax]

its general with images lmfao

[grave flax]

i retrained the lora and it works so much better now!!! it's exactly how i wanted it to turn out

[wispy spindle]

i love this style so much!!

[grave flax]

yess it's saccstry's art style! ive been following her since like 2013

[wispy spindle]

omg, i knew it looked familiar-- i rembember her from the big tumblr days! impressive how you managed how to get out her gore/body horror style here! i like how it leans more ball jointed doll here! how big was your dataset?

[grave flax]

i think 384 images

it took me longer to scour for them all since her dA got taken down than it took to actually train the model 💀

it can probably do body horror, i just haven't tried yet since pix.ai only lets you generate approximately 0.0003 pixels for 47889567 tokens

OMG these turned out so good! we even got a cyclops

[left sparrow]

SDXL is freaking amazing

from my doodle i finally managed to make something that actually looks sort of what i want this character to look like

[grave flax]

oooo

cyclops characters with dreads are so based

[left sparrow]

😛 a cyclops wouldn't be an issue really, it's the fact that she's a gazer and needs to have five eyes

no model so far has really wanted to give her many eyes

which i understand, the models have been trained to avoid making people with multiple organs

[crisp bloom]

Diablo 1 classes

[stoic pivot]

I still can't waifu, but I getting closer.

Prompt: "funny and playful furry waifu in a bunny fursuit dressed as Alice in Wonderland repairing a car engine in a garage".

[stoic pivot]

I try to make my own Anime style and I found out that if I add "over feminized" in the negative prompt I get that simpler look of the characters that I like. Image name 1 is without negative and image name 3 is with "over feminized" as negative.

[grave flax]

ooo it gives that older anime art style

[indigo aurora]

Any idea, How can I achieve such results?

[upbeat remnant]

what's the resolution size, sir?

[analog jasper]

So after one heck of a battle I have the final design of the currently unnamed "main" character of one of my many, many, custom worlds(I have a literal server full of worldbuilding, stories, and art). Note: the background is irrelevant, the character design is the priority as I can make a good background later.
Thoughts? Suggestions? Ideas? I take all constructive criticism and advice, still kinda learning here and I currently believe this is the best AI art I've made to date so it's a good point to jump off from.

[heady anchor]

try to play with patterns in clothes prompts

[analog jasper]

?

[lunar mantle]

Error occured while processing model:Input type (c10::Half) and bias type (float) should be the same
Getting this error while trying to run SDXL

[quiet knot]

Ohno what have I done?

🤣🤣

[dull cosmos]

do you think is it interesting?
https://youtube.com/shorts/CJrS8J4h6Lw?feature=share

[pseudo solstice]

How does this work?

I can't get post processing to work. Do i have to enable restore faces? That kinda defeats the purpose

[quiet knot]

Thomas the Shank Engine

😄

Featuring a 120mm smoothbore main gun

[west tree]

https://tenor.com/view/sweet-dreams-good-night-sloth-yawn-sleepy-gif-15429452

[violet radish]

Any thoughts on why all my daytime images are washed out? Prompt: ecovillage. Setup: Auto1111 SDXL base with no refiner. Thanks

[late glen]

not idea, maybe it's related to the model, try negative prompts lilke, no blurr, no low quality, no distorted pictures. Hope it works

hi everyone, I´m new here, I'm trying to figure out what prompt to use to get a path of footprints in the beach, i haven't had luck so far. This are the details and what i got: raw picture, view point, shoe footprints in the beach, sunlight, 8k, hd, realistic
Steps: 20, Sampler: DPM++ 2S a Karras, CFG scale: 7, Seed: 1645164675, Size: 1080x608, Model hash: a0d9fe1383, Model: thisisreal_v20, Version: v1.4.1

[heady anchor]

try by removing the unwanted footprints on photoshop then img2img with low denoising strenght

[late glen]

thanks, sounds like an interesting idea to try

[steel apex]

What happened to stable

[dull cosmos]

My New Tutorial 😉
Pix2Pix + Animating any parameter and or prompt in ComfyUI

https://youtu.be/tyPt7ge2VO0

[stoic pivot]

I made a prompt and when I use "girl" I can see the face, when I change it to "boy" I can not get a visible face:
"computer generated art, a strong space boy commander dressed in slim silver spacesuits is sitting in a futuristic chair, seen from 45 degree side angle, open helmet, open visor, clear helmet visor, visible face . graphical render, digital smooth clean colors, mildly detailed, clean shading"

The original prompt was:
"computer generated art, a tall space girl commander dressed in slim silver spacesuits is sitting in a futuristic chair, seen from 45 degree side angle . graphical render, digital smooth clean colors, mildly detailed, clean shading".

Any suggestion how to get a visible face without trying to change the girl image bias on the internet?

I found a solution:
"computer generated art, a strong space furry fox boy commander dressed in slim silver spacesuits is sitting in a futuristic chair, seen from 45 degree side angle . graphical render, digital smooth clean colors, mildly detailed, clean shading"

[violet radish]

Thanks. When I try the same prompts in DreamStudio with SDXL they are not washed out. I upped the steps to 45 and cfg to 9 and added fog as a negative prompt. That did the trick.

[ripe cedar]

When I use img2img (with facedetailer in comfyUI) I get weird stuff in the forehead as though its trying to make a face on the face. Someone said to increase the drop_size and I did. It makes that stuff go away but the rest of the face loses detail as though I never used the face detailer.

Another example

[west tree]

That's not a solution, it's the problem.

[lilac pilot]

^When they waste life I am generating away.

[royal tangle]

i'll buy myself a huge phat billboard and print this

[royal tangle]

vermeer; i have ones too. at home

the antichrist uses windows

[wild sorrel]

Doesn't work on my phone

if that supposed to be qr code

[heady anchor]

yes its from the qr code controlnet model on civitai

[royal tangle]

depending on your monitor/screen you have to take a shot from a decent distance

it's not perfect but this is funny to do for commercial things

[wild sorrel]

nah, it's sucks as actual qr code...everyone should be able to scan it without any problems, with any phone

when you have to find good position \ do dumb things for it to scan it's just not worth it anymore

[lilac pilot]

awesome

[royal tangle]

for a technical reason yes but that was made by one person tho. give it another month. but it does the trick for adverstisement. it's not about the linked url but the interest about it

[left sparrow]

i love doing quick doodles and then run them through different filters

[stoic lotus]

hello everyone
here are some of the latest images from the development of Crystal Clear XL - Photorealistic

[quick current]

beat this.

[quick current]

great work though man

[stoic lotus]

thank you

@quick current i'll ask my testers to come up with something #🏞｜general-with-images message

[wispy nest]

what's the workflow for it, do we have to use refiner too?

i feel discouraged to use SDXL for the 1 min wait to load up and switch between base and refiner models on my 16gb ram

[stoic lotus]

these images did not use the refiner

[wispy nest]

ahh very nice ty

[quick current]

@stoic lotusThe issue is that this was base.

If your plans are to make something better and or just the same but actually good (there are a lot of oddities within 1.0) then you need to match this (in my opinion).

you know whatI mean.

base as in standard not base model literally.

[stoic lotus]

i passed it to my team

[quick current]

you know what I mean.

[stoic lotus]

we'll see what they come up with

[quick current]

https://tenor.com/view/jerma-jerma985-chat-bubble-p_0_s_t_i_r_0_n_i_a-gif-25694764

noice

[stoic lotus]

if you have anything else, let me know, i could use a few base images as reference

[quick current]

totes

[stoic lotus]

@quick current a few exampples, feedback is appreciated

[quick current]

I can see this looking even better with an upscale.

I actually had this as my prompt.

bladerunner photoshoot (analog photo:1.2) [cosmic:1]

[stoic lotus]

i can ask my guys to use that

[quick current]

rad

[royal tangle]

lol

sd dc is the uno reverse card of this

[slim sorrel]

how i make him standing on a hill

[pastel mango]

"Son, let me tell you about the time we ruled the world. It was 65 million years ago. We were overgrown chickens with barely any feathers, but we were 6 to 10 meters tall son..."

[quick current]

Was there an update to what we were talking about?

[charred ridge]

@stoic lotus when do you expect to release your new model?

[stoic lotus]

@charred ridge sometime this week, but i am being bombarded with requests which might put a delay on things

[charred ridge]

I hear yah thanks for the reply

[muted mulch]

Look what i found yesterday

[fringe shore]

I'm being amazed by the amazing results from SDXL1. Turned my girl's car into a overlanding beast

[west tree]

Ripped Aleksandr Dugin Riding a Bear

[west tree]

@wild sorrel No sense of humor, eh?

[wild sorrel]

It's just...not a good person, let's put it this way and portraying him in this way or promoting in a way is just discussing

[west tree]

Not going to argue geopolitics here.

It's funny.

[west tree]

Although I do support Czar Nicholas XX's expansion into the outer territories to fight the space-bolsheviks

[west tree]

NecroLenin

[raw quiver]

have some money

[severe vapor]

with sdxl out not sure how many people are looking for a new 2.1 model but https://discord.com/channels/1002292111942635562/1136121154168426626

[royal tangle]

100% gpt-4 created from template ffmpeg; cutting; caption; cloned voice and tts

[severe vapor]

but is the video rendered with stable diffusion?

[heady anchor]

yes u can tell because he was on new york and wasnt mugged or punched in the neck

[royal tangle]

it's as much stable diffusion as the other filter on 3d models ig

btw do you guys know if there's already a tool to generate phonk?

[modest marsh]

hi is there anyway i can use reference only on colab

[raw quiver]

hi there

[waxen crag]

hi everyone
don't know if i am at the right section but would like to know what prompt would lead to this kind of result ?
thank you all

[raw quiver]

childrens book lineart maybe?

[waxen crag]

hi !
thank you, gonna try this

[waxen crag]

not bad but a still have to try

[raw quiver]

yeah, idk how to get the exact kind of style

[vast minnow]

I’m trying to reverse engineer this prompt. Any ideas on what art style this is called? Speed painting comes close. Any others? Thanks!

[wispy nest]

Hello, very nice pictures, I like them very much

[wispy nest]

Hello, very nice picture, it's very mega, I like it very much.

[waxen crag]

yep,
it's not common to see the result in here. pretty close by the way ! thanks for your sharing

[ruby quest]

got it to work

so amazing

[dry crow]

Does it work ?

Looks awesome

[green mountain]

Updated for Modern NY Audiences.

[blissful cairn]

I have tons of Trump 😏

Mostly eating potatoes though

[unkempt isle]

somebody guide me through this

[quick current]

What resolution are you trying to render?

[unkempt isle]

512x512

[quick current]

I'm not sure if I can help. To me it seems that you actually shouldn't even have a real way of running stable due to the fact that you have such little Vram.

I recommend asking the same question but in the tech-support channel as well as the SDXL channel.

[lime igloo]

Trumps lol

[lime igloo]

Lol

[lime igloo]

Po-ta-toes

[ruby quest]

yes try it!

[dry crow]

Nice!

[ruby quest]

using QR code monster controlnet model

https://civitai.com/models/111006/qr-code-monster
and any SD1.5 based model

[ruby quest]

idk if my settings are the best but you can try to change it untill you have something you like

[dry crow]

Thx i need to try that ^^

tried an other controlnet qr model a month ago but it wasnt that good

[ruby quest]

yeah, this one is more artistic than the other ones

[dry crow]

Looks really promising

[dry crow]

Hey, you need to add --xformers --medvram in the webui-user.bat at the line Commandline_ARGS=

[ruby quest]

this is also a valid qr code

lol

[limber horizon]

it worked 😄

[ruby quest]

i love this pixel art lora

what is currently the most easy/ effecient way to generate myself into pixel art like this?

[celest bone]

Hey! Dear community I am confused...I have to do a test job I should create 5 images same as I have shown below to get hired but i am working the second day without any rest and still didn't receive the same result...using a lot of technics, models, configs...etc...HOW this art was made?

[grave flax]

mental illness

[nocturne frost]

hello

[celest bone]

where i can to dowload these xformers?

[sacred ore]

😮 Wow, for this prompt, it returned two images that match almost perfectly, one as a continuation of the other, amazing!

[gaunt heron]

I love how often I get glowing nips on my cyborgs

[gaunt heron]

pip install xformers

[celest bone]

still can't to start webui

[ruby quest]

here is the one i sent earlier

is this something you like too?

then i can help you

[celest bone]

not bad!

[ruby quest]

you can get the type you want with another controlnet model

but u can test this first

[celest bone]

but i need to create 6 different styles

[ruby quest]

do u have comfyui

[ruby quest]

easy

[celest bone]

mooo

[ruby quest]

download comfyui

wait

[celest bone]

i like)

[ruby quest]

https://github.com/comfyanonymous/ComfyUI
https://civitai.com/models/111006/qr-code-monster
https://huggingface.co/runwayml/stable-diffusion-v1-5/blob/main/v1-5-pruned-emaonly.safetensors

download these 3

[celest bone]

@ruby quest mate i think my 3rd reinstallation of aoto11111 would be ok and after i will add whole mfkerrs models difusorioss and more)

[ruby quest]

lol aight

[celest bone]

is that correct?

[ruby quest]

yes

[celest bone]

OH Alliluyaaa

so now will to install from first link you sent above//

[ruby quest]

you dont need if u use a1111

its instead of a1111

it's a better UI

[celest bone]

but i have not these extensions i see

[celest bone]

Where? In comfy?

[ruby quest]

yes

[celest bone]

@ruby quest where i need to insert

[ruby quest]

in the controlnet folder

[celest bone]

main dir ?

[celest bone]

done

[ruby quest]

good so for checkpoint u will load up any 1.5 based model

and then as controlnet load QR monster

[celest bone]

okay but againg mad errors .....

[modest vine]

Really enjoying dreamshaper 8 so far, what model do you guys use for general prompts? This model just blows anything I've used prior to it out of the water. Can only imagine using other models if I need something VERY specific that this one fails at.

[celest bone]

why? was okay...

[royal tangle]

GGGGGGGGGGGGGGGGGGGGGGGGGUUUUUUUUUYS

animatediff with sdxl1.0

without controlnet

without guidance fully synthetic

[celest bone]

WOOOOOW

fantastic!

[heady anchor]

you cant use controlnet with animatediff since its not an img2img or vid2vid model

[royal tangle]

? i said ||without ||controlnet xD#🏞｜general-with-images message

[royal tangle]

haaaaaaaahahahaha

jap thats right

[north ginkgo]

/Angry: A woman with an intense expression of anger on her face.
Gorgeous: A strikingly beautiful woman with golden hair.
Hands on Hips: She stands with her hands on her hips, showing her displeasure.
Pink Attire: Dressed in a vibrant pink outfit, adding to her allure.
Tall and Elegant: Her tall and graceful posture commands attention.
Cityscape: The backdrop features the cityscape of Chengdu, adding a modern touch.
Fiery Eyes: Her eyes burn with frustration and irritation.
Radiant Hair: Her golden hair shines brightly, even in her fury.
Pouting Lips: Her lips form a slight pout, reflecting her annoyance.
Emotion in Motion: Capture the dynamic energy of her emotions through brush strokes or lines.

[wispy nest]

going to play with that prompt a bit

mmm Gorgeous woman, Hands on Hips, Pink Attire, Tall and Elegant, Fiery Eyes, Radiant Hair, Pouting Lips, Cityscape

[candid vigil]

Hey! Sorry to ping but I'm new to using stable diffusion, Im just curious how you animate without controlnet 😅 if there's a video link or something that you followed id appreciate it if you could share me the link!

The animation looks stunning btw

[scenic umbra]

there are quite a few ways, apps and models for animation, that one is called "animate diff", you can also hop on the #🎥｜animation channel for anything related to animation

[candid vigil]

Ooh I see! Also I didn't realise there was a separate channel for animation, ty for pointing that out!

[ruby quest]

im so in love with this QR model

[ruby quest]

u have to install pip

go to folder and do curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py

then python get-pip.py

[paper aurora]

guys can anyone tell me whats wrong i installed stabble diffusion

when i typed on webui user bat for the first time , the left cmd appears and it got stuck there for so long time

and when i doubble cliked on webui user bat again this(right) script appeared

i dont know what to do

[frigid torrent]

From the left window I can see it's working normally, to access the UI you need to put that IP address 127.0.0.1:7860 into a web browser on the same computer

[paper aurora]

lol hehe

but if i clear the cmd from the ram how do i get back this ip adress

since the cmd script might not be the same like the one on the left

i guess the scripts will appear like the one on the right

and theres no ip adress there, so if i lose it ?????

[scenic umbra]

please try on the #🤝｜tech-support channel, usually more tech savvy people there

[frigid torrent]

It'll always be the same address when accessing from the same computer, because that address is the local host address for every computer

So whenever you run the webui, that address will point to it

The script on the right is probably different coz there was already an instance running before

[bright harbor]

You leave that window open, if you close it, stable Diffusion won't be accessible in the browser

It's not "stuck"

[frigid torrent]

Yeah it's supposed to look like that. To access the UI you need to use a web browser while that command window is running it

[river glade]

Hello, can any of you help me find the style of this image / a prompt / artist name to generate something similar? I only care about the drawing style. Thank you!

[iron field]

Dang, this one got so close

[paper aurora]

ok got it , ive also genrated few images ,

but what do i do if i have to turn my laptop off and generate again next time later

[bright harbor]

Just run that same .bat file

[scenic umbra]

for help with prompts please try the #📝｜prompting-help channel

[iron field]

parameters: john william waterhouse, perfect composition, high fantasy, romantic lighting, (masterpiece, highest quality) Negative prompt: deformed, ugly, mutilated, disfigured, text, extra limbs, face cut, head cut, extra fingers, extra arms, poorly drawn face, mutation, bad proportions, cropped head, malformed limbs, mutated hands, fused fingers, long neck, lowres, error, cropped, worst quality, low quality, jpeg artifacts, out of frame, watermark, signature Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 6, Seed: 3535580634, Size: 512x512, Model hash: ec41bd2a82, Model: photon_v1, Denoising strength: 0.7, Hires upscale: 2, Hires upscaler: Latent, Version: v1.5.1

If you're curious!

[hybrid coral]

Holy moly dreamsharper/anime diffusion xl + Arcane Lora = This

Does anyone know the prompt for Anime and Digital art also Fantasy art style?

In clipdrop or this discord

[sour summit]

in taiwanese night market, eating stinky tofu

[celest bone]

sure bro

Look Juicyyy we just should to meet with demonstrating screens etc...and will be done) I am absolutelly sure one small piece wich i forgot to put un any place doing this problems

[ruby quest]

i think u have a lot of problems left

[grave flax]

nipples 🤨

[viral urchin]

My dream has becoming reality

[celest bone]

guys and girls be aware before downloading and insalling from here or github....

i dont know where but it is...here

2 days left after i first time loged in here and went from links here to git hub...

[smoky oak]

I decided to give the furry model I use for a lot of hot guy gens a break, and instead try to see if I could hone in on the more violent content from its source material

Safe to say it delivered massively

TW: (Quite bloody)

I now have a dope ref for a wendigo for my new DND campaign lol

[wild sorrel]

What are you trying to install?

[celest bone]

start was from there ...each link step by step ....

befor that was repo automatic1111

[heady anchor]

all those are clean

[frosty kiln]

made with SDXL

[stoic lotus]

Hello everyone, here is something for all of you:
CCXL (Crsytal Clear XL)
https://civitai.com/models/122822?modelVersionId=133832

realism at its finest

[smoky oak]

my tiger :>

looks pretty good

[stoic lotus]

@smoky oak thank you, we got loads more on our discord, we do all our development on open channels
we believe that is the best way of keeping people up to date on news and content

[smoky oak]

glad to hear

[royal tangle]

@celest bone 9/10 times it's the "virus scanner" itself that introduces malware to the system lol

if your ok with corporate spyware just keep using windows and you'll be fine

[celest bone]

@royal tanglea STOP pls! no one speak or tell or say about communiy here or there...if you read my post..it INFORM about maybe risks in the way which i already done.

as you reading repos...there inside the text you informed about everytime...Ohh sure authors love our club too "windows FUN" cK named.

[wispy nest]

I had to reset my computer and forgot which package i used. For AMD Nod AI something

REE

[heady anchor]

repeat that please,in english cause i couldnt understand anything

[celest bone]

means better to let ppl know about risks...if you find or see.

no?

stay safe guys and girls!

[pastel moth]

Finally dipping my toes into SD. Starting with the XL model. Just installed Automatic1111 and ComfyUI.

[royal tangle]

@celest bone bro you should not give advice on anything when you are using a virus scanner and windows /bold

[heady anchor]

plus the repos he linked are clean idk where he got the virus or if theres a virus at all and just a false positive

[subtle condor]

Today's favorite/"could be five fingers day"

[limpid lichen]

a lot of the time, those free virus scanners are often scareware. the gigantic red exclamation banner. at a time malwarebytes was great. in 2023 it relies on potential customers being scared and feeling like they need it.

[wispy nest]

I like the elf a lot :D

[late glen]

El rocket league de hoy

[frosty kiln]

SDXL realism at its finest with my custom checkpoint 😛

[sterile kiln]

Guys, do you have Kohya_SS not working anymore?

Have this error

driving me crazy

[slim sorrel]

when i try to remove lower sails, they become translucent

[sharp merlin]

does anyone have an idea how can i achieve a similar result using SD? im talking about the lines in the image

close up for the lazy people

[muted mulch]

i think you cant achieve this precision

use controlnet to make a depth mask, then use this depth mask as a deformation mask of straight lines

[sharp merlin]

uhhh can't be done without controlnet?

because it's not supported on SDXL

yet

[grizzled sage]

remember, you've got all the design options you've got from before sd existed too. you can generate an image, and then apply that effect in post with photoshop

a simple tiled pattern overlayed on the image should do it

photoshop or whatever you like

[boreal falcon]

wasnt exactly what i wanted but ok 😄

[sudden lava]

What is your GPU

[dense current]

any1 knows how do i input the following values in comfyUI

[pastel moth]

3080, nothing too extreme.

[wet blade]

I wasn't sure there was any thread for this in the art threads so I guess you guys can have it

[wet blade]

oh and this too

[mild tusk]

mewtwo but he have symbiote

[final wren]

There is this huge sculpture in downtown NYC that people take pictures with. I tried transforming some of the animals today:

[sterile kiln]

Haha guys, i have RVC working !
Really nice AI

[runic sun]

Couldn't send humans so we sent this instead

[stoic lotus]

Realism on CCXL

[royal tangle]

guys wtf xray works

[stoic lotus]

a few CCXL shots
https://civitai.com/models/122822/crystal-clear-xl

[subtle condor]

Trying to add a Kuka arm to her shoulder but I cannot for the life of me figure out how. Tried so many types of prompts. Working on a lora to see if I can get it.

[weary sky]

i refuse to switch to 1.5 in 1111 for a comparison because my pc has trouble reloading sdxl but i feel like this is a good example of sdxl understanding language better. Prompt: a woman looking at the intricate magical glowing egg she is holding

also, good hands

[charred ridge]

Nice prompt thanks

[heady anchor]

i tried on 1.5 and it doesnt give the egg an intricate design just a simple egg

[charred ridge]

I am using SDXL in ComfyUI

[lost cedar]

Mow-Ped

[marble wasp]

Did a series of winter portraits with a Victorian theme.

Imported them into gimp and did some touch up and photo editing.

[blissful cairn]

Added @weary sky 's intricate magical glowing egg instead of crystal bubble/orb/ball for last one

[midnight sedge]

Guys, can somebody please help me with something? My brain has basically turned to mush, I forgot everything. All I know is that I made this picture with the Size: 2048x1536, and I dont think it looks nice?

My question is, which size should I use instead for this picture to make it look nicer, and how would I go about resizing it?

Plz, this has been killing me for ages

Clicking on different resolutions in img2img doesnt seem to help

[heady anchor]

you could try to use the "ultimate SD upscale" extension for img2img you need to use a low denoising strenght though otherwise it would add too much details and make the image deformed

[midnight sedge]

Okay, I have that, but what resolution should I use?

[heady anchor]

u dont need an specific resolution just click the dropdown menu where it says target size and click scale from image size and use slider to make it 2x or 3x or 4x

[midnight sedge]

My issue is that I dont think this picture looks aesthetically presentable? I think the resolution I chose is a poor fit for the composition

So what resolution should I resize this picture (2048x1536) into?

Will 2048x2048 be okay?

[heady anchor]

yea you could try to use this method

https://www.reddit.com/r/StableDiffusion/comments/zuh6fl/can_stable_diffusion_be_used_to_expand_a_picture/j1mqrju/

[midnight sedge]

This seems a bit... complicated. There's a method for filling the picture just by using img2img, but thanks anyway

https://youtu.be/vpmy_6cyI7c

basically outpaint like this

My issue is that Im still not sure what resolution would work better

[heady anchor]

u could try with ultrawide resolution

like 2500x1200

[midnight sedge]

Thanks, I'll give it a try, honestly Im completely lost when it comes to resolutions, hard to wrap my head around those

[heady anchor]

yea for scenery images like those ultrawide resolutions work best,for portraits its better to use squares or vertical resolutions

[midnight sedge]

oh wait it actually doesnt go above 2048 in img2img

[heady anchor]

make it 2048 x 1080 then ultimate sd upscale it to 2x

[midnight sedge]

smart, thanks!

Any idea why I'm getting these gray borders though? I'm using low denoising contronet inpait, without inpainting anything

[heady anchor]

try to use higher denoise

[midnight sedge]

oh man, but I dont want to lose the picture composition 😦

[heady anchor]

take the image with the gray borders,put it in img2img inpaint,then inpaint the grey borders with a low denoise strenght

[midnight sedge]

damn you're thinking outside the box, I love you, thanks again!

[heady anchor]

or u could try to use just img2img no inpaint on the grey borders image just use controlnet refernce mode and low to medium denoising

[midnight sedge]

btw, I put the image in 0.55 denoising using controlnet inpaint and still got the grey borders, weird

[heady anchor]

yea maybe it needs higher like 0.75

[midnight sedge]

Actually at 0.55 it didnt seem to touch the composition at all, beyond the border, maybe it is safe to use high denoise

Honestly theres just so many functions and features its hard to keep track of what does what

[heady anchor]

yea u could also try to use img2img on grey borders image with 2 or 3 controlnets(canny or softedge,depth and reference) and high denoising strenght

[midnight sedge]

Okay thats honestly too much work and often results in a mess. So far Im having good outputs with 0.65 denoising on img2img controlnet inpainting with no inpaint. I'll share my final results

[heady anchor]

🙏

[dense wind]

the fingers in the last one tho :/

[stoic pivot]

I do not know if the bots here use AI, but how do that bot find this "bad"?

[naive hazel]

make it as uniqlo shoes

[mild crest]

what model could i use to achieve this type of character?

Like antropomorphic anything (elephant, dog) but in cartoony style.
Stable diffusion seems too generic and anime curated models ive used are too human focused.

[wispy nest]

have a look at this https://civitai.com/models/118086/3d-animation-diffusion

[subtle condor]

Use some hand negative or hand loras. https://civitai.com/models/47085/envybetterhands-locon detailed instructions here https://civitai.com/models/47085/envybetterhands-locon?commentId=166619&modal=commentThread

There seems to be a lot of that going around in the AI art world.

[wispy nest]

Nice

Looks like they're on the moon