🏞｜general-with-images

[subtle condor]

Anyone know where the FAQ is on nsfw guidelines or are we left up to our own to decide? I couldn't find anything. This server has lots of channels.

[muted mulch]

Will you eat it ?

[mint current]

i'm trying to make a Saiyan Yordle

Here's the result

[wispy nest]

Thanks for the nightmares

[fringe lake]

anyone want "Oreo Chezz"?

[rain gazelle]

awesome

[candid vigil]

Looks great, I'll have some

[wispy nest]

Looks like English biscuits

[subtle condor]

Finally got the third arm added.

[royal tangle]

the first time sdxl produced something akin to art

[dry crow]

#✍🏼｜rules-and-tos Rule 4.

[stoic pivot]

"Mandelbrot set" worked, the prompt was, "crisp colorful intricate Mandelbrot set shaped as a wolf neon silhouette"

[stoic pivot]

"Cat girl" as a test can be horrible if you forget to switch back from the Refiner model.

[subtle condor]

No NSFW Content
Excessive gore/blood/nudity or anything that exists to suggest sex including sexualized poses. No illicit imagery. This includes NSFW content involving or depicting underage characters such as schoolgirls, lolicon, etc.

[late crane]

dark suishi?

[dry crow]

I haven't seen them so idk how nsfw they were

You better make a ticket at #1010934719455707218 if your really unsure. There multiple eyes from the team can check, not only me

You also can flag messages by using ⚠️ as reaction then it will be checked by the team

[subtle condor]

Showing them is probably not acceptable, its the "suggest sex" or "sexualized poses" is where the soup gets stirred.

Oh the days when ankles and wrists were sexy lol...

[wispy nest]

Where is Shrek at

[royal tangle]

this was generated by just one image - video 100% synthetic

[subtle condor]

Yep, very terrifying.

[royal tangle]

nice hehe. it's called diffused heads on github

[hidden violet]

has anyone gotten transparent backgrounds with SDXL?

[candid vigil]

nah its meinamix

[languid bloom]

Sorry for the NOOB question but how can I avoid that, when I skip from 512x512 to 1024x1024, the subjects in the picture become 2?

Dude, I just copied the prompt from the first image on CIVITAI deliberate model's page, don't see any Freud stuff in that 😄
I'm using AUTOMATIC1111

Hires is disabled

So, is it normal to get this kind of issue in 1024x1024? I was just wondering if my installation of AUTOMATIC1111 was buggy

That's ok. Anyway, negative prompts failed. Still 2 girls. Insted, in portait mode 768x1024 works perfectly

[cedar stirrup]

That's expected. As @west tree said, the model isn't trained on this resolution so it will struggle with it. As Deliberate is 1.5, you can avoid having two characters by using controlnet, to control the pose, but you will anyway get all sorts of malfunctions to the generation. Nprompt won't always work. Hi-res fix really is the best solution, if you want to make your life easier.

[muted mulch]

someone know how to setup auto1111 to auto select vae when i switch to SDXL model ?

[candid vigil]

wait i forgot to mention, i generated the picture with meina and then used inkmecha lora in img2img

[foggy holly]

#🏞｜general-with-images lotus,dream catcher,bird catcher

[subtle condor]

1girl, solo,

[foggy holly]

Tafti

create a black and white sketch

[candid vigil]

[wild sorrel]

Without hires fix - yes, it's normal.
Most models trained on 512x512.

Some models trained on higher resolution and you can straight up generate higher res, but even they doesn't guarantie you good output, base 1.5 is 512x512 anyway

[manic mist]

anyone know how to get clothes not to stick to the body like that

[wild sorrel]

probably model specific thing if happens too often

you can try adding "tight clothes" or something like that into negative, but I'd just swap model, there are plenty of anime-like models

[late crane]

Thanks

[opal warren]

am i doing this comfy UI thing right?

[heady anchor]

needs more nodes

[opal warren]

is there some sort of loop or something?

I am using the KSampler with different denoise values for each node

this feels wrong and its a lot of connecting

what even is denoise!?

I have no idea what i am doing but images are coming out so surely im doing it right

why are these so pixelated!?

seemed to work fine before, i got this one before

is it the low denoise value? idk. the que is coming down to the ones with higher denoise values

oh ok, its getting less and less pixelated

[small garnet]

majicMIX

[lilac wagon]

if anyone likes it let me know'

[hybrid coral]

ArcaneXL Lora is amazing. Almost look like it's from the show itself

[subtle condor]

I so miss that show. I think I have rewatched it 4 times.

[hybrid coral]

Still waiting on season 2 :/

[subtle condor]

Great composition

[subtle condor]

Just sayn.

[wispy nest]

https://giphy.com/gifs/moodman-cJYdjdsQWOKOfPth4g

https://giphy.com/gifs/wow-beach-usa-8AdlIamKVYo084YL4H

https://tenor.com/L6Vu.gif

[candid vigil]

is comfy UI an extension in stable diffusion or is it a separate UI?

looks pretty!

like a lotus in a pond? 😭

i see

oh :o

thats pretty cool

yeah i like it, very unique and interesting for sure!

[opal warren]

yay, i made a wallpaper for my school computer!

andthing i can do to make it look more sharper?

[viral night]

Send image2image with low Denoise + Hires.fix 4x-UltraSharp... maybe

and maybe upscale it first beffor sending it to image2image

but i have no clue about comfy using a1111

so ignore what i have said

[dense wind]

there’s lots here

just random polaroid focused on birds and depressing road signs

[lilac wagon]

if anyone likes this let me know

[lilac wagon]

okay

[mental siren]

hi is there any way to create unique christmas baubles like this? its so urgent if anyone can help thanks

[subtle condor]

https://civitai.com/models/1960

[mental siren]

oh great thanks do i put this in models directiory?

[rancid aurora]

你好

？

[subtle condor]

Put it in the embeddings folder. \stable-diffusion-webui\embeddings

and then use the prompt christmas ornament, xmsorn

[opal warren]

just made a new wallpaper!

[wispy nest]

Oh man the amount of spacey backgrounds I have..

Addicting

need to upscale, my card is very slow

[opal warren]

wow thats a really cool one

[wispy nest]

Haha I just kept throwing words at the AI to see crazy stuff :b

[wispy nest]

Was that one a pure prompt image?

Kinda wanna get some planets like that now

[opal warren]

not sure what you mean with pure prompt image

this is it before i did inpainting

i first created a 640x340 image and then upscaled to 1920x1080 and did img2img

[wispy nest]

meant like if you did use something like controlnet depth to generate where things go and stuff :d

What did you prompt ? I'm not sure how to get this view of a planet

where its a big view of the surface

[opal warren]

i barely know how to use comfyui lol

ill check the prompt

[wispy nest]

oh my god

I just exploded a planet trying that

[opal warren]

cool!!!

the jupiter is melting

this is the original before i did upscalin

haha, lemme do img2img with 0 denoising

[heady anchor]

[opal warren]

and here is how i got that image, prompt was
"A photo of an alien solar system, god rays, panorama, highly detailed, asteroid belt, photorealism, octane render, UE5, 8k"

[surreal thistle]

I spent almost two hours perfecting this.

[barren summit]

?

[summer crescent]

Hey, this server is amazing- is there a way to find a "tutor" or guide on here? I taught myself as much as can reasonably be expected its just the technical details and custom tools now

I usually just mess around making characters for games rather than photorealistic stuff, but I find the idea of controlnet fascinating

I also love the idea of training my own model, based on images generated over the history of prompting. A bit of cherry picked self-referencing I guess

[nimble geyser]

I pretty much know nothing

[summer crescent]

It's pretty intuitive to a point

you successfully made a spaghetti haired woman, ezpz

[summer crescent]

https://civitai.com/models/7094/arcane-style-lora ?

or did you have a hugging face link

@hybrid coral

[hybrid coral]

You can try https://tensor.art/. You can use all models in civitai including lora and sdxl models without using your own PC to run the model

[summer crescent]

Thanks, so far my PC has been handling it really well!

if it is something that requires GPU its a bit slower, but I have a lot of RAM

[hybrid coral]

Also yes this is the lora i used for arcane style (note u should also run the lora using anime art diffusion XL alpha 3)

[summer crescent]

So many synergistic models

I usually use the orange mix for anime

https://civitai.com/models/117259/anime-art-diffusion-xl ?

[hybrid coral]

Yup

[summer crescent]

Hmm

[hybrid coral]

Thats the one

[summer crescent]

Do I use this as the model?

or is it a type of extension

[hybrid coral]

Just use anime art diffusion xl + Arcane lora with weight value around 2.0 to get arcane style

[hybrid coral]

Its a fine-tuned model like dreamshaper (btw same creator)

[summer crescent]

Oh sweet

I love dreamshaper

I will set this all up and see if I can generate some interesting results

[hybrid coral]

Goodluck.

[summer crescent]

do you use the baddream or unrealistic dream antiprompts with this?

or neither

[hybrid coral]

None.

[summer crescent]

Thanks for your help

[summer crescent]

I did some messing around with the Lora, I think its pretty cool. Arcane has such a distinct and unique style

The shirt in the first picture is extra cursed

[restive shuttle]

My first ever manga is done!!! Chapter 1 from Echoes Across Time

[raw quiver]

everywhere I go, I see your face
nice gen

[opal warren]

god damm

you again

17 mutual servers

[raw quiver]

hehe

[opal warren]

you are in about 18% of my servers

thats crazy

[raw quiver]

Im in a lot of servers, would take too long to count my % lol

Im quite enjoying just spamming some random letters and symbols into SDXL and seeing what comes out

maybe you could use it as some kind of style as theyre often quite similar still

[opal warren]

why are some of my outputs like this?

or like this

and then some are

makes no sense

[raw quiver]

lol

¯_(ツ)_/¯

[opal warren]

wow, text is good

[raw quiver]

yeah text seems to have improved

[opal warren]

i need to try some text

[raw quiver]

dreams usually comes out as dreeams or dreaams though sadly

[opal warren]

how do you preview the output as its being generated in comfyUI?

[raw quiver]

no idea, Im using the bot

[opal warren]

it almost got the text right here, you can probably tell what im trying to replicate

[barren summit]

I no know

[subtle condor]

Youtube is your best bet. Tons of vids

[hybrid coral]

I like second one the best. Less flaw like one adam apple and purple eyes

Meh I've made similar in sdxl

[summer crescent]

They have different flaws for sure

I think the forehead has nicer definition in the first image

and the lips look a bit better

[vestal ore]

why whenever i put a prompt on photographic is creates a picture of a camera

[vestal ore]

like i want a realistic image without that

[nimble geyser]

put camera in the negative prompt

[winged olive]

Been loving SDXL, generating some desktop backgrounds

[unborn light]

This is interesting: A photorealistic image of a black hole, captured with the Event Horizon Telescope. The black hole is rendered in sharp detail, and the accretion disk is visible around it. The image is a scientific marvel, and it provides a glimpse into the most extreme object in the universe.

Made by BARD

[subtle condor]

Amazing

[subtle condor]

Prompt?

[wild sorrel]

Why do you even put "photographic" ?
If you'll google that word , you'll get tons of cameras too.

Just use "photo of..."

[grizzled sage]

what a token shows on google doesn't really matter

it's about how clip encoder understands the context of it

[wild sorrel]

Yea, but that's about what I'd expect anyway

[royal tangle]

nvm found it here

[vestal ore]

thats a pre made style 🤦‍♂️

[formal lily]

You know I am happy with my results.
I basically spended 4-5 hours to learn what I knew very well 9 months ago XD

[grizzled sage]

expectations and reality often waver. my subjective expectations are that photograph describes a photograph. almost all of my photographic generations have some derivative of this word in them.

"Camera" is usually what describes a camera

or "photographer" describes someone taking a photo

[unborn light]

This one is called, Iron Cat

[wild sorrel]

photograph is one thing...photographic is another.

Photographic camera => photo camera => camera
We just shortened phrase to a single word, but that word is still a word that I'd expect to be used to describe camera (maybe few other related things, like photographic film)

[grizzled sage]

all cameras are photographic. That's not a phrase people use.

photographic is an adjective and it has a range of contexts

[wild sorrel]

But I can't imagine single use describing photo with that word...

photographic photo? photographic human? photographic what?

[wild sorrel]

Yea, but term is still there
https://www.vocabulary.com/dictionary/photographic camera
https://cool.culturalheritage.org/albumen/library/monographs/monckh/chap07.html
https://www.taylorfrancis.com/chapters/edit/10.4324/9780080926803-11/photographic-camera-systems-sidney-ray

[grizzled sage]

you'll 100% find it when you google it. that's called cherry picking a field of SEO terms. Google searches are not how people actually use languge. CLIP is a natural language engine. It's just not a common phrasing and clip wouldn't understand it that way. All cameras are photographic.

same seed. one difference

[unborn light]

Rasta Rat

[steel apex]

A lonely beach ball

[royal tangle]

i just trained my first lora on b1 - is this quality good enough? it's without upscalers whatsoever just base model and lora

for some weird reason the safetensors only has 980kb but it works really good. it's like when you could fit a demo int 40kb

[magic depot]

SURPASS THE NAENAE WHIP YOU CANNOT, this isnt a sdxl model and the text is fire clean

[sterile kiln]

Guys, is there a performance gain with Torch or whatever if we use Nvidia's Studio drivers instead of GRDs?

[subtle condor]

This one is called amazing! Prompt info plz...

[obtuse estuary]

💀

[smoky yoke]

Is that AI-made?

[wispy nest]

Looks like it is, some details on the clothes/horse armor warp around like AI would do

also the shape of the skull

[smoky yoke]

So you're telling me there's an AI out there that can generate GIFs?

[wispy nest]

Yes.. idk how they work, my pc can't even handle the knowledge

Some people post them on civtai, I see sometimes

[smoky yoke]

Dude

I said one time that one of the things I wished to exist was a SD-like software that produced animations instead of just static images

could this be it?

[muted mulch]

image is made by SDXL + lora, and animation with pikalabs
So yes, 100% prompt based result

[smoky yoke]

Is pikalabs an extension?

[muted mulch]

discord based app

like midjourney

[smoky yoke]

So I can't create animations using only local SD?

[muted mulch]

but you also can do it with Gen-2

here is a Modelscope extension for auto1111 who do txt2vid but not img2vid, and a lot lower quality

zeroscope model is better, you can try it here https://huggingface.co/spaces/fffiloni/zeroscope

[muted mulch]

You also have this extension to do 3d zoom
https://github.com/thygate/stable-diffusion-webui-depthmap-script

[hybrid coral]

Dude there is also an called runaway that can make your ai art into moving things. Here an example

[smoky yoke]

That is exactly what I was looking for all this time.

[hybrid coral]

Its not perfect yet but eh. The ai called is runaway

Pika is free while runaway is paid

But i feel like runaway is a better for photorealistic while pikalabs is great for digital art

Here the image i used made from Midjourney

[smoky yoke]

@hybrid coral Tried to find Pikalabs in the extensions tab

to no avail.

[muted mulch]

its a discord but we cant post link here

[scenic umbra]

lemme fix that for you guys 😜 > https://www.pika.art/

[woven ether]

Hi everyone ! I'm wondering if anyone had issue about SDXL v1 rendering really bad stuff locally, while using the same prompt/parameters on a web implementation it performs well ?

[grizzled sage]

not i

[grizzled sage]

fox is just an image of a fox 😦

[nimble geyser]

which web implementation? lots of unknown variables

[woven ether]

for example i tried on clipdrop, but also some other in the past days that I don't remember the name. But for some reason, only locally do I get poor result, most of the time in 2d such as this example versus clipdrop. I forgot the prompt I used, but i used a prompt generator who suggested an artist close to what clipdroped generated. What is strange is that this only happens on sdxl 1.0, other models seems to behave normally

[nimble geyser]

looks like high cfg?

also, you might want to adjust noise offset and clip skip. basically have the clip skip the last few steps so the image comes together better

not sure about xl, but 1.5 models didn't really resolve the noise properly https://arxiv.org/abs/2305.08891

[woven ether]

no luck trying to change cfg or clip skip, i'll read the article tomorrow it's getting late here 😴 thanks for the help

[nimble geyser]

you don't really need to read the paper, but just understand the issue at least

[unborn light]

Watch OUT boy, that's a dangerous animal

[drifting fog]

By Salvador Dali

[unborn light]

What the..........?

[heady anchor]

wdf goin on in atlanta bro ☠️ ,aint no way carti coming back after this 🤣 💀

[unborn light]

Jejejejejeje He sure will loose his interest in the city

Its not what it looks like!

[hybrid coral]

Oof sorry

[grizzled sage]

yeahhh thats what i needed. ty

[nimble geyser]

is it?

[heady anchor]

best ai video

[nimble geyser]

that second one got me a little flushed

sorry

I hope that's not a repeat

https://tenor.com/view/abe-grandpa-the-simpsons-burlesque-hat-gif-5305508

[heady anchor]

no

https://tenor.com/view/steven-segal-serious-someone-is-going-to-get-hurt-ready-stare-gif-4954370

https://tenor.com/view/indian-drama-soap-opera-my-honest-reaction-my-honest-reaction-meme-my-honest-reaction-to-that-information-gif-25988696

[nimble geyser]

edgelord?

[heady anchor]

https://tenor.com/view/blackie-gif-25635440

https://tenor.com/view/no-omar-ahmed-gif-20403143

[nimble geyser]

https://tenor.com/view/my-gawd-gif-20286182

[timid kettle]

Guys, what kind of stable diffusion is she using and how can I use it?

[sterile kiln]

She's using Automatic1111, with a specific model (redhead.ckpt), she use the same seed on a img2img generation

[grizzled sage]

She has a full youtube tutorial channel on the subject doesn't she? i recogznie her but her channel name is eluding me

[sterile kiln]

To 🇫🇷 peoples, i did img and vocals x) :
https://www.youtube.com/watch?v=H5yoRHg1htM

[fair hornet]

4080 pallit
i9 13900k

and why they are chilln?

[steel yarrow]

type "home base" in prompt using epiCRealism... checks out

[sterile kiln]

Guys,
is there an argument launch command to clean the Vram at each generation? I don't understand why, for some time, as soon as I make a generation everything goes well, if I make a 2nd one, I have my max Vram loaded and it's dragging crazy. Forced to close/relaunch SD.

(A1111)

[gaunt heron]

that's so realistic

[gaunt heron]

ok spammy

[dense wind]

sorry uwu

ew that was cringe

[heady anchor]

not spammy

[dense wind]

what

[gaunt heron]

8 pics in a row isn't spam?

ok

[heady anchor]

depends on your definition and whether you are a mod or not

[gaunt heron]

My apologies then @dense wind

[dense wind]

ur all goodd

i can see why it’s annoying if you have pings on

not pings

i mean all message notifications

[gaunt heron]

It just completely moved my art off the screen

[dense wind]

PHH

[gaunt heron]

😛

[dense wind]

okay i get u

valid

[gaunt heron]

more posts the better though

[dense wind]

i should have sent them in one message lmao

[gaunt heron]

those are creepy pics haha

[dense wind]

i wanted to mark the blood ones as spoiler but i wasn’t sure if you can do that with multiple in one

wait and also

[dense wind]

what sampler?

[gaunt heron]

DreamshaperXL

[dense wind]

ph wow

never heard of it

[gaunt heron]

oh sorry, you said sampler, not checkpoint, uhh don't remember

[dense wind]

i was about to say

i thought u could import custom samplers

[gaunt heron]

It's mostly Cyborg Style lora and Wowifier loras doing the heavy lifting

oh DPM++ 2M Karras

[dense wind]

do u wanna move to #💬｜general-chat

[gaunt heron]

actually gtg haha

[dense wind]

and if u want u can send them again too because we pushed them back up haha

okay alg

[gaunt heron]

no worries

[wispy nest]

holy cows, this is the kind of images they are making with SDXL

my big issue with SDXL is that it takes me nearly 2 mins to load up a model

another thing i notice is that comfyui is more supported with SDXL than with A1111, that's disappointing cause im too used to a1111

[dense wind]

what checkpoint?

it looks great

!!

[wispy nest]

https://civitai.com/models/118114/sdvn6-realxl

well their discord server language is alien to me

[rare vapor]

I challenge anyone here to explain this to me in plain english.

[wispy nest]

idk about others but the comfuyi wiring feels like a big hassle to get into the workflow of generating images

[fleet goblet]

Bing moment.

[wispy nest]

Bard moment

[heady anchor]

but it didnt explain the mathematical formula

https://tenor.com/view/nerd-emoji-nerd-emoji-avalon-play-avalon-gif-24241051

[steel yarrow]

for me it was the refiner addon that was making it brake, it doesn't release properly and even when not enabled it started leaking to the point I couldn't do same res on a second run

[inland minnow]

ai gave me this hot man

[dry crow]

What's in your webui-user.bat ?

[wispy nest]

i found that using "laser sky" as a prompt can give some really awesome skys

[inland minnow]

DAMN

also look at this

[sterile kiln]

Oh yeah, you are talking about SDXL refiner extension? I will kick out this shit. It correspond on time since i have problems

[steel yarrow]

I started having these weird runs where only the 1st actually went through, subsequent ones OOM at Hires.fix stage. Probably better report it on that extension github, cause they probably aren't releasing memory after it gets used

[solemn jay]

i dont talk here at all but i wanted to say im really proud of myself with getting better with prompting. i was able to create this nice one today. its supposed to be a elven forest at sunset.

experimenting with color and lighting

[sterile kiln]

Anyway, thank you, I would never have found where it came from

[hybrid coral]

@tropic shell i need help recreating this image i found on DeviantArt

This is my attempt

[tropic shell]

Well, what is your prompt, and are you doing it locally?

[hybrid coral]

Im doing it in TensertArt and using both dreamshaperXL and anime art diffusion XL

[sterile kiln]

https://tenor.com/view/baby-scream-yeah-hockey-kid-gif-24698394

@steel yarrow
Working back normally !

[hybrid coral]

a humanoid crow in a long dress pushing a baby carriage, cgsociety contest winner, gothic art, black feathers, vdragan bibin, ancestors, black crows, drawn with photoshop, mother, late morning, '20, sick, edited, cart, 3 0, mourning family, sakimi, cgsociety 9, dark, gothic

@tropic shell

Im sorry i meant the website tensor.art

@tropic shell can you help me out? I wasted 200 credits in 2 days to recreate it

[tropic shell]

#1101178378452750447 message something along these lines will probably work

[hybrid coral]

Those are already good than my 2 days attempt 💀

Its weird cause when i type humanoid Crow in black dress it works but if i put the other details its just made a maid

[tropic shell]

Read the prompt list for more info 🙂

[hybrid coral]

Hey one more question

If u dont mind? @tropic shell

[tropic shell]

You can just ask/reply

[hybrid coral]

@tropic shell I'm also trying to recreate Stable diffusion Discord and official SD website (ClipDrop) Prompt style

You know when there is option for multiple prompt like Anime style and digital art. Cause in TensorArt doesn't have any style option and you have to type it

Do u know what is their Anime style prompt is? For me to replicate in TensorArt website

This is TensorArt

And this are Anime style in this discord server

Or if u dont know can u recommend me a good anime prompt? My usual anime prompt is Anime key visual, cel-cheded and Cinematic

@tropic shell also your prompt works well on dreamshaper and anime diffusion

[tropic shell]

Looks great! There are plenty of ways you can adjust your prompt; as for the anime style, I suggest just using the bot, adding a few key words that you might be used to, etc! Just play around and don't be afraid of being creative.

There's plenty of key words in my document. You can add words like busy (for the background), illustration, a detailed painting, rim lighting, cinematic, etc. You might add stonework for the ground, like a cobblestone path, etc. Add additional game references, etc. Just do what works for you!

[hybrid coral]

Ya that's why i was trying to find out whats is the bot Prompt for Anime style. I've been also trying to replicate it and unsuccessful

The closest i got is when i use animification lora and felt like cheating

I'll read your document to see if my anime prompt will improve

@tropic shell i also tried adding manga creator like Makoto Shinkai and Hajime Isayama

[bold token]

tehran war

[icy lava]

Anyone know a way to face swap a human face into this image of tony the tiger?

[nimble geyser]

I'd suggest using face swap

[icy lava]

I've tried using Roop and insightfaceapp but it's not detecting a "face" 😦

[nimble geyser]

honestly with things like that it's probably easier to do it manually in something like photoshop and then using ai to clean it up

[wispy nest]

I finally did it, good looking avatar, or somewhat my gpu takes so long to gen

[bronze bramble]

I think this came out pretty awesome, like some kind of 80's horror film, what do you guys think? maybe too much blood?

[inland oar]

why am i getting deformed images in every prompt in automatic 1111?

[vocal cloud]

you're generating at 1024x1024.

Start with 512x512 and hires fix x2

[cloud mist]

I installed a1111 a week ago it was working fine but now I am getting this green image

[sterile kiln]

resolution?

and wich model are you using?

[digital root]

hello everyone, I would like to know if we can creat a earth that revolves around itself

[14:13]
animation

[sterile kiln]

❤️
https://huggingface.co/thibaud/controlnet-openpose-sdxl-1.0/tree/main

[wispy nest]

first 3d model i made with stable diffusion 😍

[frosty kiln]

which software or script you did use to make it 3D ?

[wispy nest]

Just used it as a reference with top down orthographic camera then mirrored it and UV Projected the textures from camera view.. and for the sides i manually select them while viewing from the side angle then UV Project from view and manually align the side UVs to fit pretty well, nothing perfect but good enough

generated a height to normal map with Krita and applied it as normal map for a bit more depth as well

[honest river]

i think im just bad at prompts

i said unicycle

[hybrid coral]

Did u use avatar lora?

[wispy nest]

Yes

[hybrid coral]

Try using arcane lora

[wispy nest]

Wish I could

PC broke

):

If you just wanna see it the Lora is available in civtai

Na'vi something

I like realistic gens more

[jagged garden]

Hi guys, I always get horrible hands. This is the best one i got. Can someone help me? Send me PN, i don't wanna spam this whole chat. Thank you

[wispy nest]

which base model did you use with this?

[hybrid coral]

Anime Art diffusion xl

[wispy nest]

awesome, thanks, been trying to get that look and having trouble 😦

[hybrid coral]

Don't forget to use arcane lora

[cyan helm]

how long?

[wispy nest]

another 3d model made with the help of stable diffusion ❤️

[boreal falcon]

Blender + SD?

[wispy nest]

4 minutes for each image on those settings

[wispy nest]

Those look awesome

[hearty violet]

hi guys

how can I choose where put a a painted wall in this house?

I'm architect

I wanna choose where put some material and change than too

I create from this

How can I put the same material of this draw in the final image? the firt whas by many atempts

[sterile kiln]

Samus Arran,
Realistic Vision 2.0 vs 5.1 :
Vote (left) 2.0 : 👍 | Vote (right) 5.1 : ❤️

[fleet goblet]

I wonder if there's a way to send images from the WebUI to a specific online URL

I want to find a way to streamline the output somehow so that I can just host images

[tiny cairn]

sent a FR 🙂

[gleaming flame]

Any different between those two ?

[stoic pivot]

I often hear and read that the CFG setting is to determine how well the AI shall follow the given prompt, but I have hard to see that, for me it mostly change the contrast. Between 5 and 9 do it give the image best contrast and colors; and at those values it also follow the prompt best, but at higher values it can flip out and just ignore the some aspect of the prompt. This is the result I see and observe, but when I asked in another channel they are sure I do not see this. I use Automatic1111 and SDXL, so is something wrong whit my settings or are the description of CFG not correct?
I want to know what CFG do, not what the result may be, I suspect it add contrast to the noise, but I have hard time find a real explanation for the CFG function, can some point to a place that showing how it work and not the common "it follow the prompt better".

[thin sparrow]

Anyone knows a model that can generate landscapes/arts with this kind of artstyle?

[fleet goblet]

A LORA could most definitely do that artstyle.

Probably something on the lines of Anime/Vintage Anime Art.

[stoic pivot]

animation background plate? That was came into my mind when I saw it. Try something like "studio ghibli background plate art, forest mountain", maybe

[thin sparrow]

thanks I will try

[thin sparrow]

maybe this sounds stupid but is LORA build-in in stable diffusion? is it a model or what?

[fleet goblet]

It's a different model type that helps nudge the style of your image a bit.

[thin sparrow]

yeah it looks great, almost what I want to achieve

[stoic pivot]

Loras are like style plugin/filter you can download, you activate them with a command in the prompt, you can find them on civitai.com. But as a warning, that site have some adult content, so close the door before surfing.

[fleet goblet]

dore

[stoic pivot]

Darn dysksplacia, and stroke donten helped. 🙂

[fleet goblet]

huh

[fleet goblet]

Seems the Studio Ghibli/Vintage Anime LORA are what you're looking for.

I made this with Crystal Clear XL + that lora.

I used Bing GPT to describe the image you posted as a prompt, and put that in.

Negative prompt: negative, ugly, broken, distorted, grainy, unpleasant, strange, repulsive, blotchy, unprofessional, pixilated, faded, low quality, poor resolution, inappropriate, unwanted, offensive, unappealing.
Steps: 20 | Seed: 3668096268 | Sampler: DDIM | CFG scale: 6 | Size: 1280x720 | Parser: Full parser | Model: crystalClearXL_ccxl | Model hash: cbc222b0 | VAE: sdxl-vae-fp16-fix | Version: 9fc70cb | Pipeline: Diffusers | Operations: txt2img```

With my typical negative prompt.

Also made by GPT.

[thin sparrow]

hmm, what if I have local installation of stable diffusion?

[fleet goblet]

Yeah, Studio Ghibli.

I'm running a local instance myself.

SD.Next Vladmantic WebUI

[stoic pivot]

I was inspired and I often like to test and work with only words and base model (SDXL), this was "studio ghibli background plate art, low angle grass forest mountain light blue sky light clouds".

[fleet goblet]

It's good, but LORAs definitely push the style much further out

It's really close, though.

More of a pastel paint.

[thin sparrow]

thanks guys, I think Studio Ghibli will be fine

[stoic pivot]

I agree with that.

[thorny lichen]

hi im new and trying to install it on my pc, but im not smart enough

could somebody please help me?

[stoic pivot]

There is so many version now, I use Automatic1111 and it was a bit complicated to install even folk say it is easy, but I manage to do so, I got help from Youtuber.
Matt Wofle is a fast talker that have many good videos https://www.youtube.com/@mreflow/videos
Olivio Sarikas also take a bit fast but clearly, he shall have a install video for a simple Sable Diffusion, look for videos newer than two weeks https://www.youtube.com/watch?v=uoUYYbDGi9w&ab_channel=OlivioSarikas
MattVidPro AI is a well known channel that often have good instructions, https://www.youtube.com/@MattVidPro/videos
Aitrepreneur is yet another fast talker but shall have a "one click" installation video, but it seldom that easy, but he often say what you need https://www.youtube.com/@Aitrepreneur/videos

Darn... I did not think of that the urls come with imagges and so that make the post look wild, I am sorry.

[thorny lichen]

thanks for the tips. looks like i got it to run on my browser.

still need alot of experience to improve my promts

[wispy nest]

Is 5.1 trained on SDXL?

[sterile kiln]

Nop, Realistic Vision is a 1.5 based model

[wispy nest]

good to know

[pastel mango]

what the fuck

how did I search that?

[sterile kiln]

Use one of my old Samus gen, from RV 2.0 to 5.1,
from 4:3 512x512 to
portrait 1305x1912, with ControlNet
Inpaint filling.

Unbelievable....

after img2img

[royal tangle]

on GPT-3.5-turbo:
testing prompt-based microposing for openpose (controlnet); output svg

[cedar seal]

/

[wispy nest]

girl with dogs

[lunar mantle]

Exciting news, everyone 🎉!

Monster API SDXL now supports 20 amazing Styles!

Steer your image generation workflow towards a specific style.

Try it out here - https://lnkd.in/gH3j3p9F

[ripe bluff]

sharing some of my LORA pic

[wispy nest]

"What have we done?"

[fresh hound]

Fast 11 leaked!!

[sterile kiln]

In a hole in the ground there lived a hobbit. Not a nasty, dirty, wet hole, filled with the ends of worms and an oozy smell, nor yet a dry, bare, sandy hole with nothing in it to sit down on or to eat: it was a hobbit-hole, and that means comfort.

[kind quartz]

yes not perfect, probably too strong denoise filter, but it is working without instruct pix2pix model this way.

[abstract anvil]

gm

[fresh hound]

WOAH i am so impressed!!! look at the numbers

wow and pika just did it in video but even clearer/better. what a time to be a live!

[marble wasp]

4 images I made with Stable Diffusion Easy UI, epiCRealism modeler, and retouching/photo editing using Gimp 2.0.

[pastel moth]

Finally getting more comfortable with SDXL and Comfy.

[lime lotus]

Can a few of you try "armor made out of dragons" every model ive tried has just put the character in generic plate armor.

[marble wasp]

these look sick!

[loud notch]

It's insane how good cars are for SDXL out of the box without any Hires. fix or any negative prompts nor LoRa wizzardry

[toxic moss]

👀

[ripe bluff]

@icy jasper this. it is chopped because it is used for a purpose.

the generate button is red thought

[pastel moth]

Those are without prompting...been trying to make a consistent 'realism' style that will produce good images without any prompting.

[dreamy otter]

Trying to learn how to condition on sketch
The results are so good I fear for humanity

Prompt:
realistic fierce green basilisk snake open mouth fangs big teeth shiny scales big yellow eyes , dark fantasy, cave environment

But seriously what am I doing wrong? 😆
SDXL 1.0, eular a, 20 steps, 512x512, CFG scale 7, Denoising strengh 0.7... basically default params

[kind quartz]

sdxl native resolution is 1024x1024

[wispy nest]

does anyone know the prompt for this

please someone. 🙏

[pastel moth]

Baby penguins taking a selfie together in the Arctic.

Penguins, Baby, Cute, Fluffy, Selfie, Antarctica. ;p

[faint skiff]

Do you guys know what causes this when upscaling?

[lilac pilot]

Nice, are you able to share model?

[lilac pilot]

Who specifically created this?

[wild sorrel]

try lowering how much it changes...forgot how it's called lol

or maybe upscale in few steps with lower resolution

[wind umbra]

can i get ur workflow

[pastel moth]

I am just using searge with my own style I’ve been trying to get down. I’ll shoot it over when I am behind a computer.

I basically used empty prompts though, I am just trying to create a baseline style that creates coherent realistic images prior to prompting.

[worn minnow]

Does this sever have channel of SDXL

?

Im not really good at ai image generator but i wanna try that SDXL thing

[pastel moth]

Yes, I’d use SDXL with Automatic1111 if you want something user friendly.

Bunch of bots to try for free though already setup

[rain gazelle]

Does Auto have decent sdxl support now? Last I saw there wasn't great support for the model + refiner.

[pastel moth]

No, still an issue if you want to use the refiner properly. It’s not setup to easily support the workflow, but likely someone may have released an extension to fix the issue, not sure as I’ve been moving away from automatic.

[pastel moth]

Here's the PNGs. This is literally just a style I was experimenting with searge's workflow. It really isn't anything, I am just trying to establish a baseline that produces good results without prompting.

[stoic pivot]

Sign that you stay up to late, this is your sleep paralysis demon.

[bronze bramble]

true, it's almost 5am here and I'm just playing with the bots

[stoic pivot]

Good night my friend, you have my premission to go to bed. So will I.

[dense tapir]

Looks a bit like Adam Gomez in that one.

[mild tusk]

john

[pastel moth]

Made my first LoRA in an attempt to generate a consistent character at the highest resolution I could in 1.5. Got to 5120x7680 before it deformed past the point of usefulness.

[mortal schooner]

Juggernaut is insane. Who needs SDXL?

[nimble geyser]

I'd say that currently if a person uses 1.5 properly they can get results comparable to xl. xl just has a higher win rate if that makes sense

[knotty crater]

工业以太网交换机

indusrtral ethernet switch design

[low badge]

SDXL

[sterile temple]

love the blue wolf in the red forest

[nimble geyser]

yeah, those are pretty top notch

[flat niche]

Try to flip it upside down. Amazing.

[simple flame]

Can anyone tell me what check point and lora are like in this picture?

[sterile steeple]

guys i have a question , what is the best model to create logos ?

[kind quartz]

there are not best. But i probably would try gameIconInstitute_v30 or something like this name.

[sterile steeple]

thank you , gonna test this model out 🙂

[kind quartz]

np @sterile steeple hope it will work for you

[silk coral]

You can always interrogate and get SD to come up with a prompt for you

[acoustic moth]

whats everyone’s preferred upscaling method for photorealism

[wispy nest]

i still prefer 1.5 for most stuff just because I don't do realism much and XL has a -strong- lean towards that, even with loras and fine tuned models, but XL's text encoder is outstanding. i went back and tried a few of my more complicated old prompts and XL got them right on the first try most of the time. yes, you can achieve similar results with controlnet on 1.5, but it's much MUCH easier to get compositional things in XL by just prompting

it's still far from perfect, but the difference is night and day

[nimble geyser]

I realize I'm not the average user, but I make stuff like this with xl. don't really even go for realism normally

that's not like my top 10 btw. just grabbed 10 images I've made recently

but I will say, I could do things with 1.5 that I still can't come close to with xl

never really understood the draw to complete photorealism of things I could potentially see in real life. I can just look at real life or pictures for that. I want weird stuff

[wispy nest]

some pretty cool stuff there, i was meaning more cartoony/anime style stuff. XL can do animation, but it struggles get something like these easily

[boreal falcon]

or simply going to stock

[nimble geyser]

ahh. well I don't know if you've looked through them much, but there are some really top notch checkpoints and loras coming out for xl. it'll take some time, but I believe it'll give people the option to do just as many different styles and things as 1.5 does. 1.5 was just king for quite a while, and sooooo many great tools and things for it

I like to think of things that it's unlikely anyone else has really thought of together and throw that out there. doesn't always work out, but sometimes

[wispy nest]

I agree it'll get there eventually! i tried a couple of them about a week ago and some were already ok, but it's just easier with 1.5 right now. amazing tech

[nimble geyser]

1.5 is also way less resource intensive so you can really pump out the images

[wispy nest]

yeah, i've got a 3060 so it's not terrible, but i feel the pinch on xl

[nimble geyser]

man, I have a 6gb 3060, lol

[wispy nest]

1.5 you can output a bunch of 512xNNN or NNNx512 images to later hires/upscale, but doing that on XL will frequently just not give great results, so your test images start at 1024x1024. generation times are pretty much the same, but since you can test seeds a lot quicker with 1.5 the workflow as a whole is faster

and i'm still running 16gb ddr3 ram on an 8 year old processor on a not super great ssd, so my system doesn't really help with anything that uses more than the gpu...which is everything to some degree lol

[wispy nest]

i was running a 6gb 1060 before i bought the (12gb)3060. was about as much as i can afford for what amounts to a hobby, but the pc is definitely the bottleneck for anything not relying on gpu vram and cuda

[nimble geyser]

well I was dumb and bought a high end laptop a couple years ago, so haven't even looked into the idea of swapping out the gpu

I can do xl just fine though. and actually made my own node for comfy so I can hit all the proper aspect ratios so I don't get extra limbs and things

usually get over 2it/s with xl as long as i'm not doing anything too crazy. nothing to brag about, but tolerable

[brisk dagger]

Is the StableDreamer bot capable of private messaging or do we have to be in a bot room in order to generate images?

[nimble geyser]

https://tenor.com/view/confused-ehh-not-sure-idk-gif-5328954

[pastel mango]

I think he needs a dentist and a surgeon

maybe more

this is Becky, she's big and beautiful and just wanted you to know that

[kind quartz]

Nice collection @nimble geyser

[little vault]

功夫熊猫

[shut sinew]

#1072220168534642768 Also server language is english, thanks

[fathom violet]

#1072220168534642768

[sage ermine]

Sometimes, you just stay there expressionless with no idea what to say...

[sage ermine]

that's how I feel sometimes, so I can relate to him there

[sage ermine]

and you? do you have times where you don't show any emotion and just stand there?

[still flax]

[young crescent]

I am having issues with my generations actually being images in the bot channels. Maybe it’s only like this on my end on my phone, but I’m not on my PC.

[winter carbon]

any tips on how to generate image of this quality?

source; https://www.pixiv.net/en/artworks/110487506

[forest furnace]

how do you feel about girls with 6 toes ?

[nimble geyser]

lol

gross

never understood the feet thing. don't hate feet, but they're just functional things that help us walk around and all that. and they usually don't smell very good

[nimble geyser]

check when you get on your pc. I looked in that channel and didn't see any activity around the time you posted. never actually used the bots, so I don't actually know how it works. but I"m assuming they'd show up in the channel

[loud halo]

Shout out to Youtube "AI Marketing Teacher". You got yourself a new subscriber and this is my first share from self hosted stable diffusion

[frozen jasper]

which of these do you think is the best looking one

[echo garden]

has anyone has any clue why my controlnet is doing a clow fiesta in the face?

[stoic pivot]

Who need photographers and designers when we have SDXL now.

[rugged dagger]

We will always need their knowledge and experience, even with AI. They just got better at their jobs. And now they save a lot of money in travel costs. And who needs to buy a Blackmagic or Sigma when you can just prompt one in? 😂

[boreal falcon]

Everyone

[boreal falcon]

If it was that simple 😄

[rugged dagger]

I know.

[boreal falcon]

I mean you are right for some niche of consumers

I heard about cameras with in built generative AI btw

[rugged dagger]

Blackmagic cameras have some AI edit, background removal and tracking toys built in. It's meant to be compatible with Davinci Resolve. That's the only one I've heard of. And it's only on their 6k and 12k models (last I heard).

It's cool tho. I'm glad to see tech moving in that direction.

And ofcoarse cell phone software.

[boreal falcon]

Maybe that one, but im not sure someone has built in such a software in Nikon camera

[stray kelp]

"Commander Shepard!"

[tulip yew]

Can I move the root folder from drive C to drive D?

[stray kelp]

i am 90% sure you can without a hitch

[tulip yew]

it works perfectly

[manic mist]

click on img -> open in browser -> click on img again to enlarge/zoomin

made with SD 100% ai (other than txt)

[prisma stone]

Maybe I shouldn't use 1024x1024 for this

[kind quartz]

@prisma stone in comfui, this cant work for you?

[prisma stone]

I have never use Upscale Latent By

What is that???

[kind quartz]

i think it wont work...
just double click somewhere and type it

[loud halo]

Would we get banned for sharing trump in jail images? I'm new here, just intrguiged because I can do this with Stable Diffusion while others can't with Midjourney.

[nimble geyser]

dude, that's a win

[stray kelp]

No least not that ive seen, would be weird if they did, there's stuff of trump fighting with steven seagal and stuff like that so you're prob fine

i'd assume most of us just came for image gen stuffs so images be images

[spark finch]

how about some 8k images

all of those are 8k or near 8k

[prisma stone]

Idk what kind of prompt do I need to add
to make my checkpoint generate the black side of his body... it's like a virus.

Could someone help me?

[spark finch]

here's a closeup of that little house on the left and a person walking around. I wish it didn't take 2-3 hours

[spark finch]

think like a model not a people. Black Virus, doesn't make sense in model context.

describe what ou want it to look like, not what you generally would think it looks like to a human or yourself.

[spark finch]

also , have you considered splitting your prompt into L and G encoders? CLIP G uses natural language while L uses the SD1.5 style word salad as you use above. But maybe if you feed both in using two prompt inputs, one you describe your image in natural language, and the other implement very concise and exacting descriptors (removing natural language descriptors like Black Virus, replacing it with specifically what that looks like, as black virus is subjectively understood, but not so objectively as L prefers) I'm no expert in it, but, that's what I was playing with the other day with SDXL

[prisma stone]

L and G? hmmm

Idk what are these ;-;

[spark finch]

CLIP/L and CLIP/G I forget the full name, but they're different encoders

[spark finch]

here this is CLIP encoder.

as you see.

The default text encode node passes the same prompt to clip-g and clip-l.

[sterile temple]

That's for SDXL, they might be using 1.5

[spark finch]

he showed comfy so I assumed likely sdxl

but you are right

either way, if 1.5 , still avoid subjective expressions, like "black virus" when you say that instead of seeing half a man with little wierd protein barbs sticking out of him or whatever you want, think about what a black virus is objectively speaking, it is not what you want, objectively. Even if I can guess what you want, subjectively.

but you're right @sterile temple that latter half of hiis model name doesn't look like something I'd expect to see just yet from XL

is this how comfyui should look?

[sterile temple]

could be as easy as 'splashed with black paint' or something

[spark finch]

I have a problem I think 😦

I just keep adding int switching to turn things on or off 😆 no one would be able to use it, cos no one would have any idea why 1,3,2,2,5 works well and 1,2,2,2,4 makes really bad looking stuff :\

[spark finch]

and yeah, that, it's like when trying to make "steel", just say brushed aluminum, it almost ALWAYS works better than actual specific name of metal texture.

[sterile temple]

I find it's better to just make small workflows for a specific purpose, it's really easy to have multiple tabs open with different workflows

Plus I made a bunch of nodes to help consolidate multiple nodes into 1 node 😄

[spark finch]

true

I use webui mostly, comfy i just go pel with and keep one.

I labeled all my stuff recently, which was annoying

but it does its job

[sterile temple]

I went a bit crazy at first having heaps of nodes and previews

After you use it for a while it starts to get annoying having all the clutter

And since you kinda know what to expect there's no point having 5 previews looking at each stage of the process

If you find something that works, try making a new simple workflow with just the necessary components to do the same job

[prisma stone]

I used SD 1.5 sadly.

[spark finch]

I don't really have it as cluttered as it looks, most of it is hidden and my prompt and controls all riight there in front

[spark finch]

Its not so bad. I usually use 1.5 until I get a better handle on how to talk to this one

[prisma stone]

https://civitai.com/models/35960/flat-2d-animerge

[spark finch]

@prisma stone can it generate scuba tanks? this is the threshold for is it truly a good model, though art models sometimes can, most photorealistic cannot, none actually thus far LOL

[prisma stone]

Idk, I asked it to generate bottles of potion before

For my Dungeon and Dragon campaign

[spark finch]

that's okay

not air tank or scuba tank lol

a male scuba diver sitting on the edge of a boat, the diver wears a brushed aluminium scuba tank on his back, from behind

that';s just a prediff, but still, that's about how it always works one time the diver had a neoprene tank wrap... empty and flat :\

it DOES not like scuba tanks, it is a red ball in a blue box 😆

this is closest I've seen but, it's very... not really. I've bee trying to get multiple 1.5 and XL models to produce one lol

[prisma stone]

I prefer to generate anime 2d stuff

[prisma stone]

But these are cool models indeed

[spark finch]

yeah, I was kidding actually. I was just in the throes of the air tank debacle I've found myself in

I just want a normal diver... not a norman rockwell scuba diver, which works fine, with watercolor / painting, but never photo

it sure will generate a coleman gas camping stove fuel tank for the diver instead of air

lol, still not the scuba tank I asked for. That's literally a camping stove fuel tank...

yeah XL cannot do scuba tanks,

it is even hit or miss with non photographic styles lol

I just asked for a watercolor painting of scuba tanks :

it gave me 9 NOT scuba tanks

the closest I got was this thing, which looks good, but oddly enough, it ignored "photograph of" lol

anyone know why scuba tanks are impossible to generate/

I'm done, i give up, it is literally a new red ball blue box

[young crescent]

The bot is only broken for me 😦

I've been waiting for 10+ minutes and the bot hasn't spit out an image at me for 2 of the same prompts in limbo.

[sterile temple]

probably not enough training images

[prisma stone]

I could generate anyone oc

but I'm bad at english

so prompting will be difficult for me

[young crescent]

The bot literally still thinks I’m dreaming. I can’t do any image generations in the bot channels. 😦

[stray kelp]

I just thought i'd share this with you all cx

[cold sentinel]

I would like to ask you, why I use controlnet to select my local image, will always generate no image, and then in the generated command can also see "width:0 height:1", regardless of whether I manually select the image ratio is so, the complete command I generated is "/dream prompt:UHD,highly detailed,professional,bokeh,chain saw man,senior high school student,asian,fighting_stance,white shirt,sneakers,tie,black jeans,blonde hair,tokyo,street,cinematic angle,(((full body))),rain style:Comic Book width:0 Height: 1 seed: 192892105 ", I downloaded to the local image, the url is https://i02piccdn.sogoucdn.com/a77f145ac4c1576a

[spark finch]

I can't imagine that. it will do them, lol but only as norman rockwell looking watercolors

it's conspiratorial

[sterile temple]

sawadeekup

[prisma stone]

Omg sawadeekup

[prisma stone]

Anyway what is this app?

The fading red A icon...

[sterile temple]

do you have an image without text overlaid?

[prisma stone]

I don't.

[sterile temple]

ow

[prisma stone]

Yeah

[sterile temple]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/2901

It's an icon someone made for A1111

[stable creek]

Need help! When I tried to inpaint the cloth, the area highlighted in red rectangles became transparent, how can I fix this?

[prisma stone]

A.I. is so fucking stupid

I SAID 1 FUCKING HUMAN

NOT 2 HUMAN YOU BASTARD

[crystal delta]

FAQ: What is Stability AI?

Our vibrant communities consist of experts, leaders and partners across the globe. They are developing cutting-edge open AI models for Image, Language, Audio, Video, 3D and Biology.. AI by the people, for the people. Learn more here

[stray kelp]

clocking technology spotted

[wispy nest]

good morning

[stray kelp]

it 5pm here but yee guten morgen xwx

[wispy nest]

8 morning here

8am*

[stray kelp]

https://tenor.com/view/anime-food-gif-10121861

mm bekfest time xwx

[subtle condor]

Gadget from Paragon the Overprime

[stray kelp]

i actually enjoyed paragon

i think what got me was it being 3rd person rather then top down like lol
I like to be in the cockpit

[subtle condor]

Greetings!

[buoyant lagoon]

i need music!

https://www.youtube.com/watch?v=BTQ6zBKybro

[stray kelp]

i always need music (will spin this up rn xwx)

[buoyant lagoon]

https://www.youtube.com/watch?v=wTR-ScwmOXI

https://www.youtube.com/watch?v=gk-HyNjZimI

[buoyant lagoon]

https://www.youtube.com/watch?v=87Wv_e2FrjE

https://www.youtube.com/watch?v=OwkkBVQB_3U

https://www.youtube.com/watch?v=imBlPXbAv6E

https://www.youtube.com/watch?v=T6a3Ch8GwOQ

im going now, have a good one, cya 🫂

[stray kelp]

Average male-identifing sd xl user:

[heady anchor]

weakest sd user

[stray kelp]

has killer shades cx

[rocky quail]

classic SD

[zinc patrol]

@wispy nest use https://colab.research.google.com/github/camenduru/sdxl-colab/blob/main/sdxl_v1.0_webui_colab.ipynb

Magnus carlson according to sdxl

click this icon

now wait till a blue link pops up in the logs

that is code

[rocky quail]

chess 2.0

[zinc patrol]

12 hours per day

downloading the models

scroll down

so the very end

[rocky quail]

i bet they use it to train their own AI with user provided prompts and stuff

[zinc patrol]

rate this image

[zinc patrol]

no smart ass

gta 5 illustration art go brrrr

[rocky quail]

6 finger/5

[zinc patrol]

conquering europe

[rocky quail]

they correct AI art

be proompter or die

[zinc patrol]

All fun and games till sdxl learns to prompt itself

I work as a Machine learning engineer at netflix

[rocky quail]

funny, i asked bing to write me a prompt, was actually decent

[zinc patrol]

and I can confirm as a junior my life is fucked

120k to fix netflix's shitty algorithm

5 mins

Rate this one

[rocky quail]

a trip to the ER out of 5

[zinc patrol]

Me be leaking my 12tb on sdxl wizardy be like:

[rocky quail]

fix your hands prompt bro, hand looks gross

[zinc patrol]

I havent mastered inpainting

[rocky quail]

you can do perfect hand without inpainting

[zinc patrol]

@wispy nest

4 links should appear

click the second one

[rocky quail]

my pc took 6:34 min to make this

[zinc patrol]

click the link

@wispy nest

select these settings

[rocky quail]

that is just realistic vision v5, somehow, it got perfect hands

[zinc patrol]

the gradio one is the best

no

ok click it dont click the remote moe

[rocky quail]

my negative prompts

[zinc patrol]

gradio is "software builder"

a1111 is the webui

[rocky quail]

the middle 3rd paragraph part is written by bing

[zinc patrol]

bro hates nsfw

and I love it

ye

The most buggy ass ui possible

[rocky quail]

i hate it when a naked woman appear instead of my nicely clothed girl

[zinc patrol]

Set your settings

[zinc patrol]

sdxl knows shit

unlike 1.5

Sdxl is pretty hesitant to nsfw

ok also dont try to use the refiner its broken

[rocky quail]

realistic vision got all bodypart really nicely, even the correct anatomy, youshould wait until it got a sdxl version

nah

you can always load another one

[zinc patrol]

yup

[rocky quail]

but in my experience, most of the time when SD is broken, it can be fixed by deleting a certain file

[zinc patrol]

Average comfy user be like

Its my workflow btw

[rocky quail]

what do you even do with it

[zinc patrol]

this @rocky quail

[rocky quail]

you make anime art? cool

[zinc patrol]

HENTAI

[rocky quail]

you make hentai games?

[zinc patrol]

Hentai images

[rocky quail]

to sell?

[zinc patrol]

I mainly work as a ml engineer at netflix

[rocky quail]

oh, ok, thats cool

[zinc patrol]

no that aint

Netflix is sinking

even if you just put a pixel at image its yours

[rocky quail]

i read the license thing, you can own it if you change it a bit

[zinc patrol]

Just bring the image to photoshop and put a pixel somewhere

[rocky quail]

realistic vision comparision

[zinc patrol]

google's gpus are pretty good

https://stablehorde.net/

[rocky quail]

i gotta try this colab thing, 12hrs free perday, no strings attatched?

[zinc patrol]

if you wana get your balls deep in image gen

[zinc patrol]

g̶o̶o̶g̶l̶e̶ i̶n̶v̶a̶d̶i̶n̶g̶ y̶o̶u̶r̶ p̶r̶ivacy̶?̶

I perfer https://stablehorde.net/

they are sponsered by stability.ai

also make sure to share your art

cant here

but I smap where I can

My nfsw folder is 12tb

ye

[rocky quail]

my man cooming for AI images, bro

[zinc patrol]

Nah I dont in my workflow for 1 supreme image the image does through 600 re gens

means 601 image for 1 result

[zinc patrol]

Like this 🔥

[rocky quail]

but why

[zinc patrol]

to make good images

make some

or tell chatgpt