#🏙│ai-images
1 messages · Page 17 of 1
💀
YOO nice
Did anyone test out SD Cascade Versus SD XL?
Still trying to get used to realism in Cascade, its not the same as SDXL I realized lol
Ok the only thing Im not getting is the lightning blue eyes...
Alright I figured it out, though it seems so uncanny lmao
404 squad finished
no way, it works with only 7 images datasets
Yea... Stable XL has more of a good edge in terms of realism (Though not by much).
100+ Art that I made for the Video were not saved T-T

Heres Bob Ross eating a can of beans on the moon because why not. ||Before anyone asks, i have no idea why the AI that i used decided to put the moon in the background||
Ayo? @fringe owl level 2 !!! 
glitchy!
using dpm_2_ancestral on Stable Cascade and it keeps giving me these spiral like images
its magical
Ayo? @shadow bane level 32 !!! 
#🤖│bots ?

I be trying to make a image of Samus Aran in a Letterman jacket. I type the prompt in, and the result keeps giving it with her helmet
And even when I type "Without her helmet" It keeps giving me her helmet!
AI isn't perfect so don't expect it to give you perfect results
Maybe you can try adding it to the Negative prompt?
There's no negative prompt with bing ai
Oh sorry
SD3 though so promising
Bing AI can't use loras
No no I mean for him to create LoRA and use SD
Lora Turned out nice
True I guess, the colab could be a decent option
Yes
Bing Good but SD Better
Not really art but I thought it was funny. I was going to cancel the installation so I can just re download the installer.
and it just failed the install
My graphics drivers were out of date maybe it’s a fix 🤦♂️
how i can make a photo by ai
Do you have a pc?
Here use this: https://bing.com/create
what model u use
AnimagineXL v3
i have never used XL models
Totally worth giving a try at some point. XL is so much better at listening to your prompts and actually just flat out better at making images.
Ayo? @tawny peak level 26 !!! 
woo
Oh and I should mention that a lot of things SD 1.5 needed lora’s and textual inversions to do XL can do most of the time with pure prompting
so is XL a base model or just a model you download and put into ur models folder
lmao I tried to do the postal dude before
the sunglasses could help
U put it in models folder
But I think u need sdxl vae
ok
very nice
I like it background 
Looks a bit like Italy
is italty.
Guys you never guess who I saw the other day I saw Taylor Swift in Sydney
😮
and before you say: Yes its legit.
I saw this popped up and was geninuely surprised Mark was there.
Though I suppose, they had their shows within the same time period and location LOL
Btw anyone know where i can get a good ai art generator for free ,no limit ,and lets me generate femous people (cuz in bing image creator doesn't let me generate femous people)
Damn is so cool, which model did you use?
He
apple emoji kinda thing
It's taco time
Guy from fortnite
T-T
AAI
funny how the headdphones at least look right
0.4 weight
0.6
mfw that random ass ghost
o3o
Off-Brand Pomni
all my images have the model in the filename, so it's easy to find out 😉
SdxlNuclearGeneralPurposeV3Semi_v30
Thanks ;D
this image goes so fucking hard ⚠️ (Light GORE) ⚠️
open original and zoom in
its so detailed
whoa wtf its so detailed
what model di-
juggernautXL 8
lemme take a look ig
comfyui
my secret is applying sharpening to the upscaled image, which tricks it into thinking its super sharp and focused
its not organized lol
lancoz scaling
@strong cosmos
no tw
banned for not giving tw to sensitive people!!
but ye looks actually good
idfk if its realistic to real anatomy but who cares 
I want SD3 to do this shit and more
1000 yard stare
the fingers are killing me
😭
Chara headpats
i should get back into doing this but am lazy
Ayo? @robust hollow level 28 !!! 
Sword sucks but it's something
this going very good rn
no shit bud
it’s called irony
obviously i didn’t expect the command to work
It won't work here
Then please don't use the Command here if you know
again. it was a joke.
beautiful thing called irony
@shadow bane generate gex
hot
aw cute fogx (im keeping it this way)
this is so goofy
Is that cascade?
sdxl (juggernautXL 8)
No wonder why it felt familiar
YO WHATS UP GUYS TODAY IM AT THE WHITE HOUSE ITS ME MAT PAT AND ITS JUST A THEORY
i don’t feel like there’s a major difference between 90% of those
there isnt
im just showing them off cause
@shadow bane make art based on @humble otter 
Didn't match the prompt super well but I think it came out really good ||The prompt was Mario, Ultra realistic||
GYAT DAMN!
||Not sure why his pecs have cracks though||
tick tock heavy like a brinks truck
Behemoth that I made for band project
training new lora :3
What’s it for
I just typed in bursting into the .gif search... 1st hit was perfect 😄
new lora test
@quaint tiger look :3
i trained with anylora this time
altho fuck i think i need to remove the little trans pin some of the arts have because ive noticed that it tries to generate the pin and fails
its cute :3
man what even is your comfy ui setup? this is some crazy quality stuff
mans being a god over here
@shadow bane it's cool what you make and all and I can see that you love gengar but it might be time to post something else just saying
i agree
its repeative asf
its starting to become bland
thanks for your understanding
Hm
Well I guess I’ll only post it in the other server then and I’ll see what else I can do other than him
Cause of course I’m addicted to his design
not gone lie but simp alert
a ghost who haunt's people willing to take there soul so it can stay with them for ever and ever
Nah he a chill guy
He don’t do that
my animal brother that is gengar for you
fredi faber
Yeah I know
i guess i can't make people/animals face the real deal but you guys can think whatever you want I guess
I guess that's a bit better
banned
I still think there shud be a seperate channel for SD creations (this includes stuff like comfy ui)
Like
The dall-e3 and sd stuff is way different
True
Feels as though I have one more in me
That’s pretty sick
TF2 MvM robot tank
LOL I just realized the resembleance
Cascade is something else
LOOK OUT ENGINEERS, ITS A SENTRY BUSTER 🔥🔥🗣️🗣️
In a world, where TF2 MvM Robot Tanks were mass produced across the world - Armed with Nuclear warheads capable of destroying the world: The entire world was nothingmore but hell - but remains is one girl, whose music may have touched billions (in more ways than one) - her name: Swift. As she roams of what was remaining in this wasteland, she will encounter some familiar to her early childhood and others she knew all too well. Now with her and her family, how will they do upon these dystopian wasteland, will they reproduce and restore SOME resemblence of Civilzation: Find out on the all new show, Breakout Wastelands. Now Streaming on ~GlobalStars for only $99.99 per month
tf2 but its after mvm

bowser
Neat
one more
petah
im covered in foam dawg
I wanna know how you can get that minimalist style too though
the same as aforementioned, you will set the detail tweaker lora into negatives
I dont know how to work it properly in ComfyUI though... every time I try to add it into the negative prompt it just adds detail anyway
@spring jungle
i dont use comfyui so idk
mk
One is not like the others
cooks
Ayo? @hollow sundial level 11 !!! 
samurott
Ayo? @spring jungle level 4 !!! 
AI art: Cyberpunk Matrix Butterfly (DALL-E 3) https://cdn.discordapp.com/attachments/1159290730200707153/1215268203266187274/OIG42.jpg?ex=65fc21b6&is=65e9acb6&hm=dff1a9b9d1f1a682cb7e1e3049229548d882c167ba746a05d94d9deb45d7c692&
Bro you could’ve just posted the image here again
No need to link it as it expires now
@keen bramble
Fursona (uh oh)
Here's Felix The Cat In Steve Purcells art style.
How can i create a ai image here?
go into #🤖│bots and ping @shadow bane and ask it to create an image
@shadow bane
If you care to see every image I've generated as of yesterday in a single video then heres that
https://stolen.shoes/embedVideo?video=https://files.catbox.moe/4c8kb1.mp4&image=https://media.tenor.com/mzCBk-mscLMAAAAM/help-me.gif
Controlnet is much more impressive than i remember
working off a single image
original
newly generated
original
newly generated on that single image
im impressed
no loras or whatever
o-o
⚠️ Light Zombie Ketchup Gore Alert ⚠️
Open Original and zoom in for the intricate details
@shadow bane What model are you using?
Ayo? @graceful lark level 36 !!! 
Still Bing Image Creator
T-T
I still rarely use SD anyways
Prompt?
For which
Along the lines of
“Grainy/dirty found VHS footage of a Crayon/colored pencil scratchy style cute simplistic flat colored cartoon style doodle, big tall buff brown/tan colored male adult kangaroo, blue eyes, wearing yellow sweatpants, simple silver chain necklace, gold ear piercing, yellow vaporwave grid sunset background”
For my most recent ones it’s
“Grainy/dirty found VHS footage of a Crayon/colored pencil scratchy style cute simplistic flat colored cartoon style doodle, big tall skinny buff brown/tan colored male adult kangaroo, blue eyes, wearing yellow sweatpants, simple silver chain necklace, gold ear piercing, black fingerless gloves, dark brown feet, neutral expression, epic karate pose, Australian beach background”
Ooh yeah very detailed
thank you
good shi bro
I accidentally un-sorted my images in the process of making this so I’m in the middle of re-sorting them
Also I has 100 Folders of 5-6 Pictures 😄
Most AI image folders I make end up with around 100 or more images
It feels like pain trying to sort through them all again
Yes, but on the contrary, I have more folders, but in these folders there are 5-6 Images
Ok
The aftermath of only 1 day of SDXL video generation
Making a video is very difficult T-T
for me animatediff and hotshotxl automate everything for me
Free Sonic OC designs!
Also apparently one of the frames in this video includes an accidental frame of a screenshot I took to show something to someone
Should be fine, it’s just a discord screenshot with blocked details
Woah! What software you use?
just a website
here ill link it to ya
Discover featured stunning Outpainting of images at PromeAI. Explore our collection of images that will inspire and captivate your imagination.
Ok I found another, it was a cookie I ate too
Currently done with over 2/3 of the images, man this takes long but I’m trying to make the process faster
Okay I’m gonna stop providing updates to not clog this channel
Shleep
hi
Girl you got me blinded
Ayo? @uncut robin level 18 !!! 
Also shaddup
Ayo? @fluid escarp level 7 !!! 
Ty ❤️
yw
I’m gonna try and upload all my AI generations to deviantart
All 26,000 of them
Its gonna be a pain cause deviantart has request limits
If you wanna see what’s been uploaded so far go here https://www.deviantart.com/mrrfyai
Solution: Don’t

And it’s already happening lol
💀
A>! How?!
Low Res image -> High Res image with minimal vram impact.
yep
if you use Comfyui, you take the image, upscale it and process it again at like low denoising
in Forge/A1111 its like a builtin feature
pixel art test
Currently at around 1.8K uploaded
stop, you have a problem 
Catch me outside how bout dah
what checkpoint is dis
DO NOT
whab
"мда." И ПОТОМ ОНА ТАКАЯ: "ХАХАХА, нъет." короче 10/10
Would u buy this disk?
Happy almost 20th Anniversary X&Y album
Any remembered playing old Coldplay albums?
nope
mr mc muscleton
50 tigs
Cant you pick like 4 of the best ones and keep it at that?
Its kinda,annoying to flood half the chat space wtih so many of the same looking picture
Please don't flood the chat with a bunch of images it just takes away from other creators. So please try not to push too many
I love this, such an nice style :3

Those first 4 Coldplay albums are great
From Mylo Xyloto onwards they were still alright (my favorite album from that era is Ghost Stories), up to Music Of The Spheres
That album is kinda bad
The only song from that album that I sometimes feel like listening to is Coloratura
btw this could help yall to make better prompt for ai art https://discord.com/channels/1159260121998827560/1218902018673086524
Ok
Ok
Thanks. It's called monochrome spot color, it was generated without a lora and I just used the prompt: red background, (spot color, monochrome, greyscale with colored background, greyscale)
Yeah, i've used it, it's definitely not perfect but it's better than having to do it manually.
missed only a teensy bit between the paw and hoodie
and it did it in less than a sec
being able to configure the alpha mask sliders in A1111 does help it alot though
Oh I didn't know there was a rembg release for A1111
it works like a charm tbh
For comfy I just have to attach it to whatever my last image output node is.
want me to try for one of ya images?
i wonder how the a1111 compares to the comfy one
I'd say the A1111 one is better considering this hasn't been updated for a while
Or it could be that my images are too big, and it's causing issues with the removal
they are of very high quality yeh
high res
Yeah 4096 x 6144, i really just found out my computer could do it so all of them have been this size.

yeah okay this rembg does not like the size of your images
straight up fails XD
Only using a 3060 as well. It's weird because I've seen other people have issues with 40xx series cards with sdxl pushing 2k or 4k images.
idk, maybe you have somekinda setting that allows your gpu to work more efficient?
The only launch args I have is --normalvram and --enable-cors-header.
Also generating, my GPU doesn't go above 8GB of vram
i think you might be using gradient checking
cat
yes
I don't think I am though I dunno how to check with comfy.
Yeah on it's own Rembg doesn't really like high resolutions, though I can make it so that before going into the node I can make it into a mask then let it do it's removal
Though it is better than I expected
I really like that SDXL is bringing more life back into the SD 1.5 models I made that were pretty meh. The 1.5 version didn't look anything like this one here.
Goober
nicu
😮
Well yeah, uploading thousands of images would result in that, but what are the notifs? Comments, follows, view counts?
Ayo? @ebon dome level 42 !!! 
Mostly favorites and “added to collection” notifs
And around 71K+ deviation views
your and i use Model Cartoon ualalal xD
idk Model name
ah
and my white cat
This cat + Your LoRA
ohhhh
updated it too
No criticism against it, but it only generates prompts for realistic models. Which it does a good job at, Cartoon and Anime models use booru tags for their generation. So an update to it adding booru tags would be cool.
i think the most of its dataset was just realistic details as its only 77m parameters, maybe if in the prompt you tell the ai you want it to be like cartoonistic/anime style it should work
like:
expand this prompt in an anime style: short white haired girl
you could try but i cant really train it to be more inclusive for all type of ai arts tbh 😭
That's true, also the huggingface demo space has a runtime error
thx for telling me ima fix it rq
idk why it putted the same argument twice or i need new eyes fr
I'm not good with python, but the colab also is giving an error as well.
No UI
it was trained with natural language
you sure you runned the install cell first? maybe retry that cus it works fine
Ok, let me try again
fixed ?
Yea fixed, huggingface one is also still building.
fixed the hf space too 
added new parameters
The man, named nick, sits in a lush garden in the heart of the city. His face is bright and cheerful, and his eyes are filled with joy. He wears a flowing white shirt and blue jeans, and he gazes intently at the flowers that surround him. The sun casts a warm glow on the scene, making it seem as though he's been enjoying the peacefulness of the countryside. @wheat radish

@wheat radish
me fr
xD
Huh... Well at least you recognized the artstyle.
I was asking if the character looked familiar
Aah yep 😅
An Update!
magic
Looks like I'll update then, hopefully my models will still work
Oh man only 34 downloads we're the first testers.
Aw man, come on internet.
yummers
Oh boy
wowie
Woah
ahh bruhh this is booru type prompting or whatever its called
ye
but still high quality though 👍
you dont need detailed af prompts for these
depends on what you need
for Anime booru usually works best
@tawny peak pov: you're a zombie about to get killed by ump45
@ebon dome gimme a random ass prompt
both work
idk send me any prompt
no need to be booru
@tawny peak
1man, bald, furious, angry, fighting pose, in japanese dojo, firepits in the background
idk
epic
super clean outlines
just give me random ass prompts
wanna see something funny?
Ye crazy
This is what superprompt gave me lmao the 1man, dressed in a furious and angry suit, is fighting a pose in the japanese dojo firepits in the background. His arms are outstretched as he grunts and laughs at an opponent with fierce intensity. The scene is depicted in the Japanese style, with a fiery red flame in the distance
sure
wtf this is epic
mf became a demon
damn
FRIEREN ADDED
I have been experimenting with using Clip_G
idk if it's doing anything but I'll keep doing it
Blud is either high or does not want to be there 💀💀💀💀
Megumin is sleeping
The fuck is a blud color with stripes of green and white, with big brown eyes
Gojo rasengan
heh
I'm your basic average girl. And I'm here to save a world! You can't stop me cuz I'm...
Close enough
im still not understand mate wut key/ app do u use? cuz ur stuff looks so clean
Not exactly what I was going for but I thought it was pretty funny looking 😂
The detailed body shading then the simple face 😂
SDXL using comfyui
sdxl
Saitama be like
im using comfyUI
OH THAT'S COOL, was that photoshopped or all AI?
Nice, I'm trying to modify Sayano's RVC-Chat that uses comfyui as it's image gen backend to use SDXL
SEGGS
im using comfyUI with stability matrix
makes it easy af to use
and you can change your settings manually as well
Yeah, same.
Yeah this isn't Stable Cascade or SD3 yet.
unless you use inpaint or controlnet
then you can fix stuf
yep
NO
I wonder how much more of Pandora's Box I can open up
Can't wait until that GFL spiritual successor comes in 3 days lmao
Ayo? @wise shard level 3 !!! 
thanks for the answer but after a while i think im gonna stay w my own style
(since im not found the server for design i guess im gonna put this pic on this server) "Modern Girl Old Issue" Simple design that i made by myself and use my tools for create tht girl, with Jnoize style
You can simply use style/artists Lora's if the model supports it and edit afterwar
IlariAI by @quaint tiger
Did he create LoRA for you?
Very Cool + 5 Fingers Respect
@quaint tiger Wow how did you get 5 beautiful fingers?
Man I knew I should had laid low not paying my 50M debt to MGM Casino
"All you young, wild girls"
ilaria model
Yis
pretty good
Riggght? X3
Well done! :3
😮
Ngl, I could had seen a Early Weeknd doing a rendition of Bruno Mar's "Moonshine"
Too bad Drake got to him first
how much vram does sdxl use? my poor 8gb laptop probably can't load both 7B llm and sdxl at the same time...
👀
It uses I think 10GB but that depends on if you use normal or lowvram
😮
Sorry, what prompt did you write?
i dont remember now, but you can just use png info that image
I collect many ai generated stuff, to compile everything in one ai that will generate all the artstyles that I appreciate here are some of them:
(for the vast majority, i did not made them and they are mostly -niji & stable diffusion generated)
Ayo? @maiden rune level 1 !!! 
Woaaah that looks sick
the blend of the lava x dark tissues, the colors does make it look good
(plus lava is hot rock ice cream 😋 )
Heheh yeah
done
thank you
Water
what did you use and what prompt if ya dont mind
DND character
Too bad the ice isn't floating
these r cool asf wtf
midjourney
a glass of water, dark background, product photography, ultra realistic --c 10
i know
Phishing link?
he's banned
Delete the message then
aight
Ty
back in 1.4 days
The spaghetti slayer
Its working 🙀
yea
I Just Disabled It Because This Feature Was Interfering With LoRA Creation
For Almost a Whole Week I Don’t Understand What’s Stopping Me from Making LoRA, and So I Started Disabling Everything and Making Minimal Settings And Only Because of This It Didn’t Work for Me
@quaint tiger / @uncut slate
doesn't this prevent overtraining? I think I used that for my negative embeddings
???
For 5-6 Days I Didn’t Understand Why I Can’t Create LoRA But When I Disabled It Everything Works for Me (I Think This Is Because I Use Colab)
🫶🫶
I realized what the mistake was
Anyone in need of a metal cover
gonna try making another lora with actual art + ai imagery as datasets
since just putting art in anylora (with veryyyy distinct artstyles) didnt work that well
Using AI images in a dataset can sometimes make or break a lora depending on the quality of the generations. If you’re training a character with ai generated images make sure most of the AI images have clean mostly natural looking hands
oh most of them look fine theres only one with totally ruined hands
the drawn art has pretty well drawn out hands i think thatll fix it
Nice then it sounds like you gotta pretty good dataset then, do you manually tag images or use an auto tagger?
auto tagger
but it also removes some bad tags
the generated images were made using prev loras of the same character im just trying to make a more general lora
Yeah, from my testing and creation of lora’s, i tag manually because you don’t need an insane amount of tags for a dataset, most every guide i read recommended that you put a negative prompt in to your dataset images but i haven’t for any of mine and all of mine turn out excellent.
My new realistic model:
just needs Negative prompt: (low quality:2), (worst quality:2)
I Just leaked the papa louie's film.
prompt?
a live action-comedy film poster called Taran Killam's Papa's Pizzeria, main character named Roy as Taran Killam He is the male worker at Papa's Pizzeria wears all red was a delivery boy, right side showing a old man with no beard wears glasses named wally happy, with all customers behind him, showing papa louie his uncle on left side thats bald wears an chef hat shirt and pants and chef playing as J.K. Simmons
new lora turned out suprisingly well, two different checkpoints used here
idk why theres a little white thing on the jacket in the second image if i completely removed any pins in the dataset there
@tawny peak lookie it worked just fine
To me it seems a bit oversaturated, but if you like it that's completely fine.
mayb its cuz of the vae idk
i only tried the 8 epoch ver, seems fine to me
but i think it looks gud
Yeah, If you ever get around to trying out SDXL I could help you get an XL model or train one for you.
i tyhink myyy hardware is too bad for xl lol
What do you have?
What's your current gpu, or is it just integrated ryzen graphics
i think my godtier specs is good enough
integrated
one of these things is not like the other

We're back after a little mishap
Apparently the DeviantArt API did an oopsie and I forgot to unconfigure a setting
um well
How do we get too post on here
Lvl up
what leve do you have to be?
I think it’s the same level as getting #✦│chat
Got it now !
Thanks
Now I can share
I'm having a lot of fun
Nah 1st lvl
Mk
Monika got the chat squidward build
ermmm.... akshuwelly it's "chad".... get it right... 🤓🤓☝️
agrjkjakjsadkhkskj
😮
Ilaria? 😮
Nope owo
das very nice
That kinda looks like her avatar tho lol
XD
XDD
As a RX 480 owner, I don't recommend it if you wanna do this stuff
We're basically forced to have Nvidia
Yes I'm answering a message from 1 week ago
youre late... i alr have it...
what do you want to make stickers of poyo?
My fursona
Is it possible to do this IN colab Without Pony Model?
hmm
i specifically use pony to make the stickers
so i wouldnt know x.x
@graceful lark how much art do you have of your sona?
i can try and make some stickers for ya
???
wdym?
how much pictures
you have of your sona
if you can send me like 5+ then i can make you a lora and make stickers for you o-o
no no, i need actual images
like this
dm :3
Ayo? @little copper level 2 !!! 
Gadget Hackwrench as a Sonic Character!
That does not look like the Sonic style but ok
It was a mix using the SonicDiffusion model on civitai, while using a character model of Gadget someone already made. Then turned the volume up for a better effect.
These were the other results I got.
I see
plausible, depends on what kind of character and image data you have
Trying out a vpred model. Does dark colours super well.
vpred is good
i personally dont like vpred a lot, but nice generation
😮
????
/j idk what that means
Just pulled this off the description of a vpred model:
This is an experimental model that uses v-prediction to fix Stable Diffusions 1.5 poor noise scheduling and sample steps.
Experimentation with the model has showed a variety of possible improvements, including but not excluded to.
Improved understanding of prompts
More accurate colours
Significantly enhanced contrast
That’s just what SDXL does now
I see
Sandwich
Heres a picture of George W Bush running from an army of Garfields in a Kmart parking lot because why not
stickers i made for a streamer i watch :P
how
SDXL
just grab some images of him
then train on thos images on pony diffusion SDXL
and just prompt correctly
like its fairly straight forward
That’s pretty cool
XD
played around with dall-e3
oooh SD3 will be so powerful for these
SD3 trained on Pony dataset, just think of the possibilities
hell will break loose
made a Ralsei LoRA that allows both black and white version to be specified :3
soon
Nature
looking awesome as expected
Squishy?! OwO
also my god
why so many of the same picture
wb btw squishy x'3
my dude. not to be rude, but cmon.... thats 36 images that are almost pretty much the same in 1 go, you've been talked to before about this x.x.....
7/8 images is more than enough no???
Been uploaded in #1159496755218501692
Sorry
They’re all just too good I can’t pick that small of a sample size
Warn kinda feels deserved in this situation tbh 😔
i get you are passionate about it and that you love all the results you get, im the same way. i also dont dislike the stuff you share, i just dont feel like flooding the chat with 36 images which are extremely similar in nature is the way to go. hence why i said it would have been fine with just 7/8, that way you still get to share your results that people can enjoy. but you also dont flood the chat :3
shii- I should retrain my tf2 map lora
fuck I need to find the files 💀
YESS
@strong cosmos new tf2 lora
same dataset but retrained
theoretically better
ah alr
Aight
welcome back
workin on leh tom & Jerry model
Looks good
tnx ❤️
what would happen if you tried to mashed them into one character?
It happens actually
You get either
Huge jerry
Tiny tom
Or tom with Jerry's colors
Or jerry with toms colors
The cons of multi character loras
Yup
oh hi :333
hey
woof
Your banner is cursed :3




