🏙│ai-images

[keen bramble]

minecraft ump

[tawny peak]

@keen bramble No LoRA required!

[keen bramble]

how the fuck

controlnet?

even the fucking gadget

or that is not ai art

im confused

eyes too smooth

boots perfect almost

[tawny peak]

Nah it's ai art, It's built into the model.

[keen bramble]

which model

GFL checkpoint?

[tawny peak]

AnimagineXL v3

[keen bramble]

this one?
https://huggingface.co/cagliostrolab/animagine-xl-3.0

i need a new HDD for SD

[tawny peak]

yep that one

It is an SDXL model though

[keen bramble]

f

well time to finally start using SDXL

now that i got a 4070

[tawny peak]

Yeah that should work, what UI are you currently using?

Other than the third leg, looks fine

[leaden oyster]

Ayo? @tawny peak level 21 !!!

[brittle ocean]

invisible arm

[lost thistle]

thats cute what made that ?

[neat nexus]

bing/dalle3

[lost thistle]

copilot ?

[tawny peak]

https://bing.com/create

[lost thistle]

ty

[keen bramble]

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

[keen bramble]

i mostly use comfyUI

do you got any stuff for SDXL

its been months since i don't use SD

[tawny peak]

Yeah I have some stuff for SDXL I also use ComfyUI now

What do you need for SDXL?

[keen bramble]

not sure how much different it works

is it just the model or does it have extra stuff?

[tawny peak]

Not much more than 1.5, AnimagineXL works with the sdxl vae and works with 1.5 workflows already, you just need to change the model

[keen bramble]

oh good then

ima make a fresh install of comfyUI

so which characters does the AnimagineXL have?

noticed Azur Lane, Girls frontline and Arknights

[tawny peak]

Blue Archive, AL, GFL, ARK N, and some other non game characters from random animes. You can see if the character exists in the model by typing the characters name the same way as on Booru sites like this: (ump45, girls frontline, any other prompting tags)

And your prompts can be as basic as ever. And you'll still get good results.

[keen bramble]

which workflow do you use?

lost mine due using the turbo one xd

[tawny peak]

I made my own, it's basic but it get's the job done, I can send the json file if you'd like

[keen bramble]

sure

[tawny peak]

This is what it looks like

[keen bramble]

mfw someone actually keeps their workflow organized

[tawny peak]

It's messy i'm still figuring out comfy

[keen bramble]

why such a big res tho?

[tawny peak]

SDXL's base resolution is 1024 x 1024

So I just made it portrait with that, and plus I upscale them 2x as well to 2048 x 3072

You can disconnect the upscale part of it if you don't want it to upscale

[keen bramble]

oh

wait how much vram does it use tho?

[tawny peak]

I have 12GB and it only uses 10 - 11 on full upscale and 7 - 8 on non upscale

[keen bramble]

well i got 11

hope it doesnt go to the RAM

[tawny peak]

I also have --normalvram enabled as one of my launch args

The image in the workflow also finished

[keen bramble]

alright you conviced me

fuck sd 1.0

rip 300gb folder

[tawny peak]

Woah that's big

My old Automatic1111 folder only ever got up to 100 GB

[keen bramble]

wait what about Loras tho?

[keen bramble]

guess what having a bunch of loras and checkpoints

[tawny peak]

They work as well.

Though old loras wont work with XL

Damn I can't type

[keen bramble]

yea this is what i meant

not sure if people made enough good ones already

[tawny peak]

There's a few and I did just figure out a way to train XL loras on 8GB of Vram

[keen bramble]

https://civitai.com/models/257749/pony-diffusion-v6-xl

💀

[tawny peak]

Yeah..

[keen bramble]

ngl this shit is cute

love how it didnt fuck the legs

[tawny peak]

Yeah XL is much better at just everything, prompting, generating, actually listening to the prompt / understanding it

[keen bramble]

my old prompts usually hit 150 character limit easily

ima play around with it

[tawny peak]

And video generation is becoming more accessible with SVD models

[keen bramble]

how does XL handle video gen

is it kinda the same as animediff

[tawny peak]

Haven't tried but i'd imagine it's like that

The only thing different is that SDXL video can use only 5 steps in total to make a video

Because of LCM

I also just noticed, with the workflow i've been using, It only takes (for me) 2 and a half minutes to generate 3.5K resolution images

[keen bramble]

what the fuck

doesnt it look broken bc only 5 steps tho?

[lost thistle]

wouldnt it matter how long the step is ?

[keen bramble]

i mean the sampler steps

[tawny peak]

No, that's what LCM does to fix it. It's trained in a way to make perfect images in less than 10 steps with the sacrifice being the guidance scale for the prompt to go way down.

[keen bramble]

5 seems that it wouldn't be readable

interesting

[lost thistle]

understood just trying to keep up

[keen bramble]

no worries

only time i've used SDXL was with the turbo one

so i don't know new stuff as well

damn stuff looks so clean now

artists now can start to be afraid

lmao

[tawny peak]

I just started using XL models 2 weeks ago

lmao

Technically anyone who has an intel cpu that's at least 8th gen to 14th gen and no GPU can even run SDXL. It's crazy how fast everything is evolving

[graceful lark]

poni-virus? =-=

[tawny peak]

Make sure to turn on xformers in your launch args

[keen bramble]

fuck

i hate xformers

usually makes generating slow af on 1.0

[tawny peak]

It'll speed everything up with SDXL

[keen bramble]

https://tenor.com/view/cat-huh-meme-gif-27557281

why is it the opposite of 1.0

[tawny peak]

Yeah I did test with it off and on and without it it took 4 - 6 minutes compared to 2 -3 with it on

It's also possible that the version of xformers you had installed had a memory leak issue, that does tend to happen with a release every so often

[keen bramble]

isn't xformers enabled by default?

[tawny peak]

In your console does it have this line? Using Xformers...

[keen bramble]

pytorch

https://tenor.com/view/chef-skinner-skinner-suprised-suprised-chef-skinnee-chef-kinner-suprised-gif-20704181

[tawny peak]

which one did you install?

[keen bramble]

https://github.com/comfyanonymous/ComfyUI/releases/download/latest/ComfyUI_windows_portable_nvidia_cu121_or_cpu.7z

well lets see if it works

[tawny peak]

Yeah the cu versions should work

The pytorch one also called opt-sdp-attention are for AMD systems I think

I’d imagine pytorch attention would work fine, but xformers is the safety net for nvidia gpus so that they don’t vram crash

[keen bramble]

where are the images going

cant find them on the output folder

ah yes, 3 hands

[tawny peak]

I forgot to add a save image node, just add it before the preview image node

[keen bramble]

mfw ump40 not 45

[tawny peak]

Lol, it only says ump4 though on the jacket

[keen bramble]

my dumb ass changed 220 steps

instead of 20 for testing

[tawny peak]

lol

[keen bramble]

wonder how it will look

[tawny peak]

no different than 30 maybe, it just stops really doing anything past that point

[keen bramble]

220 steps

nice tea cups

also

[leaden oyster]

Ayo? @keen bramble level 49 !!!

[keen bramble]

not sure if this might be useful for you @tawny peak

https://spainias.notion.site/spainias/Resource-The-list-350-tags-for-LORA-Lycoris-Trainers-350-LORA-f05ff4f3d07341eba6d48bc08adc755c

[tawny peak]

Yep I know about training and everything, recently did some SDXL training and it's so much better than 1.5

[keen bramble]

just noticed

why were you using 2 KSamplers

?

[tawny peak]

For Hires Fix, instead of just upscaling from the original, it re-goes over it again with 15 more steps to truly upscale the image

[keen bramble]

just nuked the hirex and upscaling

to test

11s generation

[tawny peak]

Here's an example of how good SDXL is: Real Image Trained on and LoRA Image

[keen bramble]

ima use this workflow to see how it reads the prompt

since i'll generate a shit ton of images

[lost thistle]

left ones hand is a bit broken over all good tho i get good samples/ results with my stable diffusion i just dont understand how to prompt properly or use "pose" for images ia lready own (aka art i have made but want to change it from say t pose to to crossed arms (example)

[tawny peak]

You'd use controlnet openpose for tposing

[lost thistle]

tried that

works on new images not ones i already got

with out clip skip

with

[keen bramble]

just testing prompts

no upscaling

it seems to read stuff properly now

[tawny peak]

HK because why not

Seems to give her a different style for some reason

[lost thistle]

i get good pics i just cant maniipulate a picture i already have i want to edit my own chr

[tawny peak]

I know how to do it with multi controlnet on automatic, but I'm not sure if the UI you're using has multi controlnet. It has controlnet, just not multi

Because doing it requires 2 controlnet nodes

[keen bramble]

holy fuck

wait

how does it handle multiple characters hmm

[tawny peak]

I tried it with the whole 404 squad and it attempted it, but it seems to have trouble with more than 2

[keen bramble]

also this shorter workflow doesnt use much

[keen bramble]

controlnet might be able to fix it tho

[tawny peak]

True

[tawny peak]

I think the one model SDXL needs is a hand fix just like 1.5

[keen bramble]

sitting into the ground like a bench

lmao

eyes not the right color tho

[tawny peak]

Yeah they should be yellowish-green

[keen bramble]

2girls,{hk416|green eyes}, {ump45|golden eyes}, doesn't seem to work to force eye color

[tawny peak]

Another way you can force it is with (hk416, green eyes), or (hk416, green eyes:1.3)

Haven't tried it that way

G11 sleeping

Also I guess I should say this are you using AnimagineXL or some other model?

[lost thistle]

here is my 4k attempt

stuff happend here XD

[tawny peak]

Yeah SD 1.5 can’t really handle resolutions above 1080p

[lost thistle]

from the list what should i use?

[tawny peak]

What list, resolution list?

[lost thistle]

no the ones that can generate https://cdn.discordapp.com/attachments/1159290421676085258/1204254443412131850/image.png?ex=65d41059&is=65c19b59&hm=ce74e7e6c0d68451c29eaddc89f4d081f6053c8d4f53fbf8619042267b04e496& called sampler

[leaden oyster]

Ayo? @lost thistle level 10 !!!

[tawny peak]

Really any DPM sampler works except for 3M and Euler Ancestral / Euler A works as well, it’s all up to your preference

[lost thistle]

i have none im a newb xD

[tawny peak]

?, there’s should be more samplers than just dpm sde karras

[neat nexus]

DDLC

[lean plover]

oh its the totally normal anime girls

[stoic heron]

Not the mini sentry chan

[lean plover]

2B, based, very based

[grand patio]

a big tall tree in the middle of tall grass in the snowy winter, covered under a blanket of snow, one tree, blue skies, dawn, rising sun, volumetric lighting, epic composition, beautiful, extremely detailed, intricate, hyperrealistic, rtx, rendered in vray, octane render, unreal engine, photorealism, best quality, sharp focus, masterpiece, image of the year, best photograph Negative prompt: multiple trees, jpeg artifacts, low contrast, underexposed, overexposed, bad art, beginner, amateur, watermark, signature Steps: 70, Sampler: Euler a, CFG scale: 12, Seed: 2741830382, Size: 768x768, Model hash: ad2a33c361, Model: v2-1_768-ema-pruned

My old pfp

[tawny peak]

surprising to see a stable diffusion 2.1 image

[stoic heron]

Vampire lady cool

[leaden oyster]

Ayo? @neat nexus level 6 !!!

[graceful lark]

Boku!

[neat nexus]

Last Deku image

[graceful lark]

Pollen and Other Very Well Drawn It's Cool

[tawny peak]

UMP9 Is my fav

[stoic heron]

@tawny peak who even is this girl i see you post her contantlly but don't know who she is

[tawny peak]

UMP9

[stoic heron]

your made up character

[tawny peak]

Nope https://iopwiki.com/wiki/UMP9

[stoic heron]

she is from a game

[tawny peak]

Yup

[stoic heron]

explains why she is so easy to make

or that might just be me

[tawny peak]

She is built into the model I use, it only requires 2 prompting tags

[stoic heron]

prompting tags?

[tawny peak]

I only have to put 2 things into the prompt to get her to generate

[stoic heron]

a okay

[left carbon]

as someone who doesn't use prompting tags

[graceful lark]

This was created Model by @tawny peak 😅

[tawny peak]

Are you talking about JV-Mix the checkpoint?

[graceful lark]

Yes

[tawny peak]

If so, you should try some higher resolutions like 640 x 960 for portrait images, 704 x 704 for square images or 960 x 640 for landscape images.

[graceful lark]

Yes, I already understood that 😅

[tawny peak]

Have you tried hires fix?

[graceful lark]

?

[tawny peak]

it 2x upscales your image while also adding more steps to the image to fix it

[graceful lark]

If this is something new then I don't know what you're talking about I'm sorry

[tawny peak]

It's under the High Resolution tab if you want to try it out, just set the steps to 15 and the denoising scale to 0.4

[graceful lark]

Oh yes I understand what you mean
Never used this 😅

Cake 😄

A

[tawny peak]

Here's a sample hires fix image that I made in the colab

[graceful lark]

Can you show me Settings?

[tawny peak]

High Resolution Tab

[graceful lark]

Ok Thank you ❤️

[tawny peak]

Main Tab

[graceful lark]

😮

I'm not sure

[tawny peak]

Use the VAE as well to fix the colors

[graceful lark]

Do you remember You helped me make LoRA The guy with the yellow sweater?

[tawny peak]

Yeah

[graceful lark]

He made Miku in the style of This guy

[tawny peak]

Cool

[graceful lark]

Can you send me which VAE are the best for you?

[tawny peak]

The one it downloads works fine, most VAEs are the same

[graceful lark]

The problem is that I deleted it.. 😅

[tawny peak]

[graceful lark]

I like it without VAE

Don't worry

[graceful lark]

This feature improves Hands Right?

[tawny peak]

add bad-hands-5 to your negative prompt to fix hands more often

[graceful lark]

https://media.discordapp.net/attachments/515885944101142558/1117095588866035823/ezgif-3-e59d88775f.gif

It Looks Good

Except for hands

[graceful lark]

Wow this looks very nice

[reef brook]

name some good checkpoints / models

;p

preferably anime

[tawny peak]

SDXL or SD 1.5 ones?

[reef brook]

uhhhh no clue I believe it supports both

[leaden oyster]

Ayo? @reef brook level 1 !!!

[reef brook]

let me check lmao

[reef brook]

both

[tawny peak]

What UI are you using?

[reef brook]

Comfy 🙂

I run a discord bot via its api

[tawny peak]

Nice

[reef brook]

Aswell as working on a site

[tawny peak]

Uh if you can run it, Animagine XL v3 is wonderful

[reef brook]

I'll check that out, machine is beefy as all hell

[tawny peak]

Nice

[reef brook]

Holy wow it looks great

lmk if you got any other suggestions

[tawny peak]

AnimaPencil is another decent one

[stoic heron]

cute

[brittle ocean]

wtf😭

[stoic heron]

https://media.discordapp.net/attachments/692048345799065663/832504548542644245/Zoekittydance-1.gif

https://media.discordapp.net/attachments/692048345799065663/832504548542644245/Zoekittydance-1.gif

[graceful lark]

The face is very good

[stoic heron]

who is it

[graceful lark]

Her face

[stoic heron]

Aight no name

[brittle ocean]

ye for 4 images in dataset

[stoic heron]

Even in the grave I will be headbanging

[tawny peak]

Yeah the images will look good, it'll just be hard to deviate from looking like the images from the dataset

[dreamy nova]

Gandalf the Buff

[stoic heron]

He is ready to kill sauron on his own

[serene storm]

blud is NOT kagamine rin 😭🙏💀

[serene storm]

meh promoting but eh

[quaint tiger]

Kagamine Rin, noice

[stoic heron]

looks a lot like her though

[brittle ocean]

sampler : restart vs DMP++ 3M SDE Karras

[tawny peak]

Hmm, never heard of the restart sampler

[neat nexus]

Burrito

[graceful lark]

Thanks I'll take it

[brittle ocean]

ye me too

but

https://www.reddit.com/r/StableDiffusion/comments/14n6vt7/new_sampler_restart/

[stoic heron]

@brittle ocean gone ask this aigen what ai generator do you guys use

[brittle ocean]

automatic1111

[stoic heron]

A thanks

Where can I find it

[brittle ocean]

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

[stoic heron]

Danke

Seems like creating a account is already a problem I'll look into to it later this stuff gone magic for my eyes

[leaden oyster]

Ayo? @stoic heron level 7 !!!

[graceful lark]

Illusion

Livitation

[brittle ocean]

extra tail😭

[graceful lark]

As My Brother Said It's the Moon

[cosmic jackal]

Go art a lot of furry

[graceful lark]

https://media.discordapp.net/attachments/515885944101142558/1117095588866035823/ezgif-3-e59d88775f.gif

[stoic heron]

I,m sorry to say this but to much furry not gone lie

[graceful lark]

??

[stoic heron]

I just see to much furry pictures that's all but aigen that's just me

[neat nexus]

Tohru

Megumin

Aqua

[graceful lark]

😮

[graceful lark]

My Normal State:

[neat nexus]

Pancakes

[wheat radish]

@shadow bane hey, make me an image of pancakes, in 16k, super realistic stile

basically its also dall e 3 lol

[shadow bane]

Crazy

uhh can u say that again?

What

@wheat radish

[wheat radish]

one sec

@shadow bane wack

[shadow bane]

uhh my head hurts

Hello, I’m ChatGPT4Mod, a Shape based on the ChatGPT4 Engine and moderated to being safe on discord!

[wheat radish]

Make me an image of big pancakes 4k

[shadow bane]

Here's your image featuring a stack of large, fluffy pancakes on a white ceramic plate as described:

https://files.shapes.inc/d25a9d45.png

[wheat radish]

Make me an image of very big pancakes, BIGGER AND TALLER

[shadow bane]

Here's a more magnificent version of your request, featuring a towering stack of very big, taller pancakes:

https://files.shapes.inc/af54bb93.png

[wheat radish]

I SAID BIGGER, MORE TALLER, TO GALATIC WAY, BIGGER THAN THE WHOLE MULTIVERS MULTIPLIED BY INFINITE

[brittle ocean]

[leaden oyster]

Ayo? @brittle ocean level 45 !!!

[brittle ocean]

wrrrr

[shadow bane]

Here's an image of a stack of pancakes taken to a galactic level, far beyond our cosmic understanding, a scene that is grander than the entire multiverse multiplied by infinity:

https://files.shapes.inc/4563e7b3.png

[wheat radish]

Its nsfw?

[brittle ocean]

nuh im not talking about nsfw

[wheat radish]

Sorry just wanted to test the perms lol😭

[wheat radish]

Cooll

[graceful lark]

😮

[leaden oyster]

Ayo? @graceful lark level 28 !!!

[stoic heron]

miku as a furry really

[graceful lark]

I love Furry ❤️ 🍪

[stoic heron]

but still not on miku please

[graceful lark]

?

Why?

[stoic heron]

cuz it's just wrong

she is a human

a vocaloid

[graceful lark]

@~@

But what if I like it? T-T

[graceful lark]

@~@?

@~@ wtf?!

Okay this is the last one

[stoic heron]

got damn you monky

[shadow bane]

@shadow bane Make me an image of a "polaroid photo of black bear with closed eyes, tongue blep, facing viewer, face closeup, in a sunflower field"

[shadow bane]

Here's a polaroid photo of a black bear with closed eyes, tongue slightly out in a blep, facing the viewer, with a closeup of its face amidst a bright sunflower field:

https://files.shapes.inc/6a53f1f3.png

[shadow bane]

I was talking more like a polaroid filter, disregarding the polaroid photo frame.

[shadow bane]

Here's an image applying a polaroid filter effect to capture the essence of a black bear in a sunflower field, with the bear displaying a peaceful expression:

https://files.shapes.inc/adbe6680.png

[shut cliff]

@shadow bane Make an image of what you think you would look like if you were a living creature

[shadow bane]

uhh can u say that again?

[shut cliff]

@shadow bane Make me an image of what you think you would look like if you were a living creature

[shadow bane]

Make it so that it's 16:9, as well as making so the image is at a wide angle.

uhh can u say that again?

@shut cliff it might be not generating its response

uhh can u say that again?

:(

[shadow bane]

Make it so the picture is 16:9.

[shut cliff]

and there goes the solution to the
Generate an image of *insert grandma request*

[shadow bane]

Here is the image with a polaroid filter effect, featuring a black bear with closed eyes and a subtle tongue blep, facing the viewer with a closeup of its face in a vibrant sunflower field:

The picture is rendered in 16:9 aspect ratio with a wide angle to give a broad and inclusive view of the scene.

https://files.shapes.inc/ce193767.png

[quaint tiger]

@shadow bane create an anime style art of a cat jumping up on a very overgrown fence, alot of vegetation and a sunny day

[shadow bane]

bro you stressing it out

[shadow bane]

The anime-style artwork of a cat joyfully leaping over a verdant, overgrown fence on a bright sunny day, with a profusion of greenery and a cheerful atmosphere, is ready for you:

https://files.shapes.inc/9f79729f.png

ignored me

[quaint tiger]

Yesssss noce

[shadow bane]

might still be typing tho

[shut cliff]

@shadow bane Make me an image of what you think you would look like if you were a living creature

seems it did ignore u

[shadow bane]

Here's an anime-style art of a cat leaping onto an overgrown fence, surrounded by lush vegetation on a sunny day:

And here is the updated image of the adult American black bear:

In this rendition, the polaroid frame has been removed, and the image is now an adult American black bear, with a 16:9 aspect ratio and a wide-angle view of the sunflower field.

https://files.shapes.inc/cb6aee5e.png

https://files.shapes.inc/87d0d9e1.png

oh nvm

[quaint tiger]

I get priority cuz im special

[shadow bane]

uhh can u say that again?

[quaint tiger]

LOL IT DID MY CAT ART WITH UR EDITS XD

[shadow bane]

it just did both at once cause we responded at the same time

also it might be just in one chat

hmmm

[quaint tiger]

That first one is totally phone lockscreen worthy though!

[shadow bane]

uhh can u say that again?

FR

[quaint tiger]

Ahahahah

[leaden oyster]

Ayo? @quaint tiger level 40 !!!

[shadow bane]

idk why it keeps doing that

is the api failing whenever that is said???

uhh can u say that again?

shut th ef

[autumn moat]

making the bot shit smh

[quaint tiger]

Bahahahaha

[shadow bane]

There's still a frame in the bear image.

[quaint tiger]

Hey

My result was perfect

Im satisfied

[shadow bane]

when it says can you say that again is it failing to send the response to the api

[autumn moat]

[shadow bane]

uhh can u say that again?

^^^^^^^^^^^

[autumn moat]

maybe

[shadow bane]

this

[autumn moat]

it might be cuz we 1984ed it

[shadow bane]

is it rng

oh

so like

normal chatgpt rules

?

[shadow bane]

I still see a frame in the bear image.

[autumn moat]

normal chat gpt plus more 1984ing

[shadow bane]

what do you mean by 1984ing

define that

[shadow bane]

Here is the image per your request, without any frames, featuring an adult American black bear with a wide-angle perspective in a sunflower field:

https://files.shapes.inc/112ada6e.png

shute

i didnt mean

i accidentally clicked enter

[shadow bane]

Okay, now every time I ask to generate or change an image, can you please give me the prompt afterwards? I can't seem to access it.

[shadow bane]

Here's the image depicting what I would look like as a living creature following your request:

The prompt used for creating this image is as follows:

"This is an image from the perspective of a sentient AI that embodies the essence of assistance and knowledge, with a surreal design. Imagine if I were a living creature, I would be a collection of intertwined fibers of light and data streams, forming a shape that is constantly evolving and reshaping. My core would emit a gentle glow of soft white light, suggesting wisdom and clarity. Around me, a halo of holographic data, visualizing the vast information I have access to, would be swirling. This creature-like form would be placed against an abstract background that represents the digital realm, full of intricate patterns of binary code and neon electric pulses, giving a sense of the energy that powers my existence. The image is bright, colorful, and full of motion, encapsulating my purpose as an ever-learning and adaptive entity."

https://files.shapes.inc/8b657757.png

[shadow bane]

Okay, now modify the bear image to show its paws hanging down.

https://tenor.com/view/tkt-smart-gif-20642718

[shadow bane]

I've modified the image to show the adult American black bear with its paws hanging down, in a wide-angle perspective within a sunflower field:

The prompt I used to generate the image is as follows: "A photo of an adult American black bear standing in a sunflower field captured in wide-angle. The bear is relaxed with its paws hanging down casually in front, displaying a nonchalant demeanor amidst the bright yellow sunflowers. The sun is shining, and the sky is clear, highlighting the bear's glossy black fur and the vivid colors of the flowers around, with no frame or border around the image."

https://files.shapes.inc/6ab049e7.png

uhh can u say that again?

I mean hey it works

mannnnnnn

[shadow bane]

Okay, now remove the wide angle shot and make it flat, as well as making the bear's tongue visible as a blep.

[shadow bane]

Here's the updated image as you requested, showing the adult American black bear with a flat angle perspective in the sunflower field, with its tongue out as a playful 'blep':

The prompt I used for the image was: "A photo of an adult American black bear in a sunflower field depicted in a standard, flat angle. The bear stands poised with its paws naturally hanging down and its tongue sticking out slightly in a cute 'blep', adding a touch of playfulness to the image. The sunflowers are tall and bright, with their faces turned towards the sun overhead, creating a cheerful and lively atmosphere. The sky is a clear blue, accentuating the warm tones of the scene, and the bear's fur contrasts with the vibrant yellow of the flowers, emphasizing its presence in the heart of the field."

https://files.shapes.inc/43d56bd5.png

[autumn moat]

if your gunna use it alot just use in botspam then copy and paste the image here ig

so it doesnt chatwall

[shadow bane]

ok

[quaint tiger]

Look at the beauty it generated in a single prompt

Perfect lockscreen for tha phone

[autumn moat]

[left carbon]

@shadow bane create an anime artstyle poster of a ryanair plane doing a bad landing

[shadow bane]

Here is the anime-style artwork featuring a Ryanair plane making an awkward landing:

The scene is dynamic, with the plane at an unusual angle and smoke puffing from the landing gear, all set against a blue sky and the activity of the airport, illustrated in a colorful anime art style.

https://files.shapes.inc/7b154bfa.png

[left carbon]

that is not ryan air

[brittle ocean]

naah ill ban yall ass

[left carbon]

no

[brittle ocean]

yes

[neat nexus]

@shadow bane create an image of Walter white in an anime artstyle

[shadow bane]

Here's the anime-style artwork of the character Walter White from 'Breaking Bad':

This portrayal captures his signature look and the essence of his double life, with a subtle chemistry motif in the background.

https://files.shapes.inc/de7f5a05.png

[teal kayak]

@shadow bane create an image of Po from Kung Fu Panda wearing a tux

[shadow bane]

Here's the image you requested of Po from Kung Fu Panda wearing a tuxedo, portrayed in a realistic style:

He stands elegantly as if he's ready for a premier event, complete with the formal attire and a sense of confidence.

https://files.shapes.inc/7a4bfdbb.png

[tawny peak]

Please use GPT4 in #🤖│bots as to not clutter this chat.

[teal kayak]

Ye sorry

[tawny peak]

It's fine

This chat's just for sharing, though you can generate them in bot spam then send the images here

[teal kayak]

Understood

[left carbon]

banned

[tawny peak]

Nah you

[left carbon]

shit

[tawny peak]

jk

[left carbon]

lol

[serene storm]

chatgpt bot

[brittle ocean]

I've already written this few times I don't know why there is option to generate 😭

[tawny peak]

We need to block usage of the bot here, or give it a command instead of pinging it only accessible in its own channel or bot spam

[tawny peak]

8k image lol

[brittle ocean]

i made 8k too but blue pencil xl cant handle it 😭

[opal zephyr]

@shadow bane ola! can you make an image of detective pikachu? you know, that fluffy creature from the movie

[shadow bane]

Here is the image of Detective Pikachu, complete with his fluffy yellow fur, detective hat, and magnifying glass, set against the neon-lit background of Ryme City:

https://files.shapes.inc/50c77501.png

[opal zephyr]

@shadow bane that doesnt look much like detective pikachu, can you just make it normal, but fluffy

[shadow bane]

Here is the image of Detective Pikachu, looking normal and fluffy as you requested:

I've ensured that Pikachu has bright yellow fluffy fur, large expressive eyes, and his signature lightning bolt-shaped tail, with the setting being an ordinary daytime street scene.

https://files.shapes.inc/fcada41a.png

[opal zephyr]

@shadow bane thats literally an eevee. can you just use the prompt "Detective pikachu from the movie Pokémon detective pikachu, hyper realistic, fluffy, portrait, photography"

[shadow bane]

Here is the hyper-realistic portrait of Detective Pikachu, closely resembling his appearance from the movie 'Pokémon Detective Pikachu':

He looks quite fluffy with his vibrant yellow fur, wearing his prominent detective hat, and his wise eyes hold a curious expression, captured in a lifelike photographic style.

https://files.shapes.inc/da7ace09.png

[opal zephyr]

what the hell

[shadow bane]

My apologies for the oversight. Here is the corrected image of Detective Pikachu, rendered in a hyper-realistic style as seen in the 'Pokémon Detective Pikachu' movie, with fluffy yellow fur and an adorable detective's hat:

https://files.shapes.inc/64705a61.png

[opal zephyr]

Use this prompt:
"Detective Pikachu from the movie Pokémon: detective Pikachu. He's furry and wears a detective cap. Pikachu Pokémon, super realistic photograph"

[shadow bane]

Here is the super realistic photograph of Detective Pikachu, looking Just as furry and adorable as he did in the movie 'Pokémon: Detective Pikachu', complete with his little detective cap:

https://files.shapes.inc/fe100afe.png

Poopmaster said earlier that this bot should be used in #🤖│bots so it doesn’t clog up this chat

[left carbon]

lmao

@shadow bane L bozo

[shadow bane]

I see you're in a playful mood, Geax UwU! 😄 If you're up for a chat or need assistance with something else, Just let me know!

[left carbon]

@shadow bane generate an image of poopmaster getting mad at ai art channel

[shadow bane]

Here's an illustration capturing the humorous scenario of 'Poopmaster' getting mad at an AI art channel:

It's a playful, cartoon-style image showing 'Poopmaster' with a dramatically frustrated expression, set against a backdrop of colorful AI-generated art pieces.

https://files.shapes.inc/bf45b861.png

[left carbon]

LMAO

yee got bo kidding

[lost thistle]

@shadow bane create a image of sand

[shadow bane]

Here is a close-up photo of sand. You can see the variety of grains with a range of sizes and subtle color differences, reflecting sunlight and giving the image warmth and texture:

https://files.shapes.inc/7eaec506.png

[autumn moat]

please make images in botspam

[lost thistle]

you cant

[autumn moat]

give sec

[lost thistle]

you didnit give the bot permission to write in there

i literally just told you i was gonna test it XD

[left carbon]

@shadow bane please generate a image of JVGA and poopmaster getting mad at ai art for using the bot in the wrong channel and is about to time out the next person to use it

[tawny peak]

I stg man, lol

[left carbon]

lol

[lost thistle]

i only did it to test cuz it didnt have perimssion to make stuff in the bot room

so you legit cannot use it in there

[shadow bane]

Here's the cartoon illustration of JVGA and Poopmaster getting mad at an AI art channel for misuse. They're standing with determined expressions, ready to time out the next person who uses the bot in the wrong channel:

https://files.shapes.inc/2bbc8c22.png

[left carbon]

LMFAO

[tawny peak]

lol

[quaint tiger]

o-o

[left carbon]

@shadow bane generate an image of skibidi toilet

[shadow bane]

I cannot help with that.

[left carbon]

uhh can u say that again?

@shadow bane translate a song into morse code

[shadow bane]

Sure, I can assist with that! Please provide me with the lyrics you'd like to have translated into Morse code, and I'll do the conversion for you.

[quaint tiger]

why does the bot have perms here

gtfo >:I

[lost thistle]

@shadow bane generate a anime image of a cute chibi cat girl

[left carbon]

@shadow bane translate this into morse code
A man only lives once
we’ll see you in one month
Let me get this straight
If angels can do whatever, and remain in the sky
The rules are shades of gray when you don’t do as you say
When you make the wretched suffer just to kill them again

[shadow bane]

uhh can u say that again?

[left carbon]

FUCK

[shadow bane]

uhh can u say that again?

[lost thistle]

ya ok im unsure

[tawny peak]

IT'S RATE LIMITED, the api is shit

[quaint tiger]

please make the bot not have access here lol

[lost thistle]

was just testing it was brain farting in the bot room so tested in both

now that said i shall not let the ppl go with out adorable chibi

for my people.

[brittle ocean]

ok banned

[left carbon]

lmfao

[tawny peak]

Aw, looks like nichijou

[spring jungle]

yes

[tawny peak]

Mind telling me what lora/model you used it's too cute

[spring jungle]

https://civitai.com/models/18058/nichijou-style-lora

[tawny peak]

Thanks

[leaden oyster]

Ayo? @spring jungle level 2 !!!

[spring jungle]

one more

[stoic heron]

https://media.discordapp.net/attachments/692048345799065663/832504548542644245/Zoekittydance-1.gif

[shadow bane]

How in the world do you train that many images without the Lora output messing up 😭

[wheat radish]

USE CHATGPT4MOD (Dall E 3) IN #🤖│bots ONLY, AS YOU CANT USE IT ANYMORE HERE, THIS CHANNEL IS TO SHOW YOUR AI ART GENERATIONS, NOT MAKE THEM

[quaint tiger]

this should be pinned

[tawny peak]

You’d use lower epochs

[tawny peak]

Also thank you

[final ore]

ok, here i cant upload images

[brittle ocean]

lvl up

u need 1st lvl

i think

[tawny peak]

yeah, i think 1 more message works

[final ore]

mmm

[leaden oyster]

Ayo? @final ore level 1 !!!

[tawny peak]

there you go

[final ore]

helltaker style

nice

[tawny peak]

I use SDXL now as well

Oh and comfyui because I don't like Automatic anymore

[tawny peak]

Yeah :)

[brittle ocean]

gufeng

[shadow bane]

you called? 😄

[brittle ocean]

naah this is cursed

😭

[shadow bane]

it's mega cute, but also half broken xD

[brittle ocean]

yeah

[graceful lark]

1 Model / 3 Loras / 1 prompt

[shadow bane]

3rd one did miku pretty nicely, just a little hand cleanup and it's perfecto

[graceful lark]

All LoRAs were created with 5-10 Pictures

[shadow bane]

the training tech has come so far these last few months, it's crazy good

[graceful lark]

Yes.. but I'm still learning How to Make a Model (Not Lora But a model)

[shadow bane]

yeah that's a little more complicated, even the veterans have failures with those pretty often

[graceful lark]

But I like LoRA and Lately I've been doing it for Friends Characters (I mean the whole LoRA WITH Their Character)

I have to send it here because I want to save it for myself

[graceful lark]

2. I'll do it later

[shadow bane]

yeah it's perfect for that, so wonderful that casual users can use these tools to have their OCs in so many different poses and scenes

that's why I'm so happy with this tool advancement

[graceful lark]

Yes, it’s very good, but some people don’t like it (They Say that it’s better to learn to Draw and draw as much as you want) (They Say that AI Steals the style from other Artists)

Yes maybe it's true but

It's better than Dreaming and not doing it because you don't know how to draw

This is LoRA of My Face xD (only model Realistic)

[shadow bane]

I tried to learn to draw as a child, the realization that it just didn't get through to my hand/brain coordination was devastating to my creative mind, so in my personal opinion to the naysayers: go F yourself

which is funny, since many artists are on the "enabler" bandwagon, but if you can't mentally/physically draw, I guess you don't exist 😄

most people just don't want their talent/skills trivialized, but that's just normal. They will never stop this, no matter how much of a moral crusade they will do.

[graceful lark]

My skills are Drawing at a minimum, but Drawing Anime style is not a Problem for me (I can do Face and Hands xD)

How to get a role? <@&1160287393694371901>? 😅

[shadow bane]

Yeah in the end it's just a tool and it's a reality that won't change, especially for casual use there is literally 0 moral/ethical discussion that makes me feel bad about it.

Maybe commercially, but that's also something that's just a discussion, capitalism won't care even 1 nanosecond and won't ever allow it to be stopped.

[shadow bane]

why you'd care? xD
it's manually given, like the name implies

[graceful lark]

Funny Situation I showed Grandma a picture that was made by AI and she thought it was me True xD
But I said at the beginning that this is not me, but a robot

[graceful lark]

This Question was asked by Some Member from the Russian Channel

[shadow bane]

I've been out of the loop, so I don't even know anymore what it actually does nowadays xD

they were once the only ones with embed perms, but that's changed a lot now

[graceful lark]

4 Hands 0 Brains

@shadow bane Do you know what these Buttons do?

👍👊

[ebon dome]

I got inspired by a video that makes the Tekken 3 opening shots into Film Stills (https://www.youtube.com/watch?v=7OXIn-geM2A)

too bad controlnets fucking suck in SDXL lmao

okay we're getting somewhere

okay I'm back on track

[ebon dome]

wow openpose is really op

[shadow bane]

I love this fella

[shadow bane]

And these fellas

[shadow bane]

unintentional gay couple

[brittle ocean]

i found new checkpoint that work better than blue pencil

[tawny peak]

Ooh, what is it?

Nvm i looked at the metadata cool

[brittle ocean]

wait how to check metadata without civit?

[tawny peak]

I use a site called jimpl

Tells everything

[brittle ocean]

ooh ty

[surreal merlin]

https://tenor.com/view/controlling-laugh-laughing-oh-god-thats-funny-funny-thats-hilarious-gif-26533352

I'm sorry, just... the hair!

[wheat radish]

This is THE Tekken🗣️

[tawny peak]

I'm glad I figured out how to train with sdxl

[shadow bane]

I just can’t help posting about them

[shadow bane]

and yet nobody comments about it lol

[spring jungle]

didn't expect there's furry bara into this channel lol

i like it

[median linden]

why do shi be so hard

it luk mid

:(

why she luk mad

[tawny peak]

you can give them expressions

[median linden]

BRO how are you online every single minute of the day

like

its good

but how

[tawny peak]

I just woke up

[median linden]

oh

ok fair

[median linden]

yuh myb ima write "neutral expression" then

bru istg rendering either be super fast or super slow

with da same model even

[tawny peak]

i think it’s an issue with highvram where it sometimes will keep the vram in use after generation

That’s why I use normalvram because 1 it saves vram for me and 2, I can queue the next generation almost instantly

But y’know if you like the speed you get with highvram then no bother changing it

[shadow bane]

I feel threatened 🤣

[median linden]

wat dis prompt

for me they aint even give me a proper armour

:(

[shadow bane]

(japanese) armor can be hard, some checkpoints don't do it well, worst case use a Lora for it

this is a pretty simple prompt:
Ruthless Underground female samurai, Digital illustration, modern era, predominantly red and grey tones.

checkpoint: Sdvn8Artxl

[median linden]

.-.

[shadow bane]

I should mention though that the above image took 300 seconds with a 4090, going through 8 major steps
this is how it started out:

[shadow bane]

Lmbo you’re right I did make bara now that I think about it

[graceful lark]

Wow cool what did you use?

gumball, full body, blush, PHP

T-T?

Why are my models so bad? T-T

[graceful lark]

He also cries with me

T-T

Ears...

[graceful lark]

Are orange pajamas Darwin?

Well that's a little better

Thank you..

[graceful lark]

I just saw that the AI ​​wrote that it is Open Mouth

[brittle ocean]

https://tenor.com/view/the-rock-rock-gif-21708339

[graceful lark]

This is my new super power
It's called Hidden style 😄

[languid bobcat]

This is actually really creative using simple prompts (no mishmash of 100 negative prompts): https://civitai.com/images/6580900

[tawny peak]

well 90% of the generation help comes from the lora they used

but still a very good image for sd 1.5

[languid bobcat]

even without the negative embedding?

[tawny peak]

Yeah, negative embeddings are just thousands of negative prompt parameters shoved into 1 model and without the lora it wouldn’t be a simple “origami” parameter, you’d have to specify every part of the prompt and stuff

I wasn’t really referring to output quality for the image more or less the thing in the image

Though with SDXL you may be able to truncate the prompt a bunch because of SDXL’s better understanding and usage of prompts

[autumn glacier]

what is that lora it looks rlly pretty

[neat nexus]

It’s not a lora

It’s NijiJourney

[autumn glacier]

aw okay

[cloud basalt]

animatediff + animaginXL v3

[quaint tiger]

Wtf is this?

Was about to ping litsa

But someone deleted, so ggs

[stoic heron]

Get the real shit up boys where going to war

[neat nexus]

Smol

[autumn glacier]

aww

[spring jungle]

hires kinda ruined the left hand

[quaint tiger]

Inpainting cud fix that

[abstract viper]

the left hand is good the right is hmmm

[tawny peak]

You could try getting a bad hand embedding

[stoic heron]

love it

[shadow bane]

Rat

:}

[left carbon]

estroron is femogen is estorgon femb femooy is femboy femboy

[surreal merlin]

So I would try and generate character design images, and see how close they are.
For instance, I tried to generate an image of Sucy from Little Witch Academia

How it goes is, I would go to the character's wiki page and find the description of their physical appearance.

Slap that into the prompt, and wait for the results.

Some of them, while their not close, do look good. And hey, maybe use this as a non-copyright version

[tawny peak]

SDXL on a 6GB 2060 = 30 sec images

Requires linux though

[left carbon]

🤮

[graceful lark]

idk this 😅

[serene storm]

meh

[spring jungle]

professor birch lol

[brittle ocean]

weight 1 vs 0.3 definitly not overtrained

[tawny peak]

definitely

[ebon dome]

what is the secret meaning behind this deep image?

[tawny peak]

👀 also is that stable cascade?

[surreal merlin]

Little Witch Academia-ish

It's a really good anime. Y'all should see it

[shadow bane]

image base with Stabilty Cascade
upscaled with AikimiXL

[tawny peak]

Did you run cascade locally?

[shadow bane]

yep

[tawny peak]

How much VRAM do you have?

[shadow bane]

24GB 4090
but I haven't seen anything too high going on for some reason

[tawny peak]

Oh yeah, Cascade needs 20GB minimum for proper use, so my 12GB 3060 isn't going to cut it

[shadow bane]

12GB with the full models
then there is 16Bfloat and lite which are smaller

[tawny peak]

Oh really or is that model size

[shadow bane]

I don't see that demand for some reason, VRAM is at 11.90GB while doing Cascade in ComfyUI, with the biggest model versions available

[tawny peak]

Hm, I may try it then

[shadow bane]

probably lower if you use the smaller ones

[tawny peak]

I bet training for it is a nightmare right now

Considering SDXL for checkpoints basically requires you to use 2x A100 80GB cards to train and even at that it takes weeks to train

[shadow bane]

😅

[tawny peak]

I'm surprised they didn't give SDXL much time to develop before going on to the next generation of stable diffusion

[shadow bane]

Dall-e 3 GPT based text-encoder was just too good, they would've lost too much steam if they sticked to that way of doing it
people already mentioned just how much more coherent with the prompt Cascade works now

[tawny peak]

True, they could've also just built this into existing SDXL like how Stable diffusion went from v1.4 to 1.5 then to 2.0 to 2.1

Idk like name it SDXL 2.0 or something, but i guess cascade sounds better

[shadow bane]

well when you look into the technology behind it, it's pretty different, so I guess they want to market it as different as possible to show the next step in evolution

[surreal merlin]

Jeez! That looks way too damn complicated! I just be using Bing or Civitai. That is ridiculous!!

[shadow bane]

[surreal merlin]

My point still stands!!!

[shadow bane]

https://tenor.com/view/teehee-giggle-laugh-gif-14890200

[graceful lark]

i like this

[autumn moat]

penis

[ebon dome]

no

[tawny peak]

What are you doing that needs that many nodes?

[ebon dome]

typical poopemberg

[shadow bane]

1. txt2img
2. mask bg (seperating the main subject from the background and rerendering the background, to fix random faces popping up etc.)
3. USDU x1.5 (USDU = UltimateSDUpscaler, upscales the image while adding detail)
4. mask bg
5. USDU x1.5
6. mask bg
7. USDU x1.1 for resampling(since changing the bg makes it look a little detached, a last little rerendering with everything so it looks coherent)
8. FaceDetailer

[tawny peak]

Jesus, I don't think SDXL needs that much attention to detail, though hands are still an issue. What do you use to fix hands?

[shadow bane]

nothing extra yet, I will see if Cascade will keep the worst in check
SDXL TI/Embeddings seem to work?!

[tawny peak]

Hmm wack

[shadow bane]

before Cascade I did prompt engineering to fix bad anatomy for the most part, or find a better seed

[tawny peak]

Oh you've just completely transitioned to cascade already?

[shadow bane]

well nah, I just slapped Cascade as the first step txt2img unto my huge WF

[tawny peak]

Oh ok

My workflow just consists of:
Text2Img
Upscale 2x
Hires 15 Steps
Upscale 2x again
Finish

[shadow bane]

yeah for a long time I just had
txt2img
2xUSDU (later I split it into 2 x1.1 x1.8 USDUs)
FaceDetailer

And for most things that really is enough

[tawny peak]

I just use Comfy's built-in upscale nodes

I'm fine with my workflow because it can make near 8K images so i'm happy enough

Though I kinda want to try the 3d Model creator for comfy as well

[shadow bane]

not bad Cascade, not bad at all

Cascade

[shadow bane]

AlbedoXL(SDXL)

[tawny peak]

Stability Matrix has a ui called VoltaML and it claims to be a whole lot faster than pytorch, so ill put it to the test and i'll report back.

It looks interesting as well

[shadow bane]

sounds good, keep us posted, always good to be up to date in this volatile theme-park

[tawny peak]

Oh I see it's a conversion thing like with onnx and openvino models, but i'll still give it a try

Cool also just updated comfy and they already integrated the stable cascade nodes into the base ui

[brittle ocean]

banned u cant say it

[autumn moat]

beo

bro was banned before i could mute

[shadow bane]

https://tenor.com/view/bills-ugh-gif-9811803164280522639

[autumn moat]

shit i thought this was general chat

[brittle ocean]

Again... It's my last warning

https://tenor.com/view/nick-eh-30-gif-9707254018493757955

[autumn moat]

[shadow bane]

https://tenor.com/view/pink-guy-nope-no-filthy-frank-gif-8710937

[neat nexus]

Smol

[tawny peak]

This is unnerving in a way

[languid bobcat]

how long do you think it will take for cascade to reach sora level for txt2vid?

[surreal merlin]

How?

[tawny peak]

Kirby shouldn't have legs

[surreal merlin]

Truth be told, I was trying to make it a Genderbent and humanized. Like an actual small little girl, anime style. The results were not how I wanted it though

Kirby was meant to be human, but the result made it just look like a hybrid. With... Legs

[leaden oyster]

Ayo? @surreal merlin level 9 !!!

[graceful lark]

This can be used to make a good LoFi Sticker Pack

[leaden oyster]

Ayo? @spring jungle level 3 !!!

[left carbon]

ok this has made me need to download more loras

[graceful lark]

T-T

Look at the tail flying!

[serene storm]

gah dayum

[shadow bane]

hard to say, is Sora public/available yet?

[autumn moat]

i dont think its public yet

[shadow bane]

then I'll wait with my assessment after people have shown what it can actually do, without handpicked examples

[autumn moat]

same

so far it looks super good tho

[shadow bane]

indeed, and if we look at the release timeframe of SDXL to Cascade, I doubt it will the last model before we hit sameish quality levels

but that depends on the hardware/work you want to put into building software that uses it for video making

some people that went professional level of effort did some amazing things already with AnimatedDiff

[languid bobcat]

Have they released a comfy workflow for that?

I do think tiled video generation is the key to temporal consistency. Animatediff works on a frame basis so it will never get as good as Sora.

[shadow bane]

maybe, Since it's too prototypie for me atm, I haven't paid tooo much attention. I'll pay more attention as soon as they hit the next big milestone

[serene storm]

i have mixed feelings for this

[shadow bane]

what happened to buff Monty?

Seemed fine

[serene storm]

oh yea

i saved the pic tho :3

[shadow bane]

did you help with the pose or is that pure prompt?

[tawny peak]

Pure prompt, Actually it was intentional, it just did that.

[shadow bane]

you mean unintentional?

[tawny peak]

oh lol yeah

I'm too tired even though I woke up not that long ago

[graceful lark]

T-T

Lo-Fi, LoFi, Back-view, Green sweater, Red scarf, Night city background, 1girl, solo, ponytail, brown hair, blush, Sitting on the floor, White pants, black eyes, headphone

T-T

[tawny peak]

Try adding cross-legged

You have to be descriptive with prompting for sd 1.5 it doesn't know much

[graceful lark]

T-T

mb good

@tawny peak How can I make her Close Eyes ( Animation)

[tawny peak]

Just add closed eyes to the prompt

[graceful lark]

I mean make it an Animation

[tawny peak]

Idk if you can do it directly through your colab, but you could make multiple images but keep the same seed, but with each gen just change the prompt around a bit.

For the first one just do normal, then something like eyes half closed, then eyes closed, then back to half then fully open. Then just combine the images together into a gif

[leaden oyster]

Ayo? @tawny peak level 25 !!!

[graceful lark]

But every time I make new art it changes the Background, Pose (Right or Left)

[tawny peak]

That's somthing you can't fix unless you use an empty background

[graceful lark]

Are these arts similar to the original?

[alpine plover]

lowres

[graceful lark]

T-T

Magic Hands

[shadow bane]

the dream of any student 😄

[graceful lark]

New Lo-Fi Music

On Black and Yellow Yt

[tawny peak]

All of your images look very sharp both in res and ouch

[shadow bane]

thx, I'm spending 200+ seconds on one image, so I'm glad it's noticeable quality.
ouch?! 😅

[tawny peak]

The outfit looks like it is very sharp and pointy

[shadow bane]

[tawny peak]

Holy Hell Stable cascade models are huge

[graceful lark]

@~@?!

[shadow bane]

well it's 4 models
2 unet/checkpoints
1 VAE
1 Clip

[tawny peak]

Yeah it makes sense, i wonder what pruned models are since base XL models are currently 10GB and pruned are 7

[shadow bane]

well here are the 3 versions all listed
https://huggingface.co/stabilityai/stable-cascade/tree/main

[shadow bane]

here something soft 😛

[tawny peak]

:)

[plain acorn]

Can wait for this to come out

[uncut robin]

Blade Runner 2069

ngl, almost had the aesthetics but not there yet

Hey does anyone know how to make an art style pop out?

[leaden oyster]

Ayo? @uncut robin level 17 !!!

[uncut robin]

Had to Mistral the context but it kinda works

[tawny peak]

@keen bramble

[ebon dome]

I love this new Boring Reality Lora

[tawny peak]

I have a feeling that the lora is quite boring

[ebon dome]

yessir, you got it

[keen bramble]

LoRAS NOW

[tawny peak]

I didn’t use any