#🍥|anime
1 messages · Page 109 of 1
Very good
Vae is what transforms the latents into images/pixels
a) don't ask for the leak. b) it's been superceded by better models by now
So it's only the last part of the generation
But isn't nai still best base to train on for 1.5 anime models?
It is the most general
damn, this prompt is hard for the model 😮
So your lora will apply to most anime models
And even non anime models
Ok wait I send guide
Just follow the guide on the github page, it should be easier
This is focused on beginners. The training settings probably aren't 100% optimal but should do well.
Github repositories used in this guide:
1: Kohya Script: https://github.com/kohya-ss/sd-scripts
2: GUI to use Kohya Script : https://github.com/bmaltais/kohya_ss
3: Scrapper: https://github.com/sA...
ye sure im doing it rn
Dragon maid
Bmaltai's
What
Derrian's GUI is the way to go 
Ok then, send them guide for that ui
it's lighter than auto and comfy?
Haha
It's UI for training
oh
https://github.com/derrian-distro/LoRA_Easy_Training_Scripts It has a installer and pretty UI 
These?
Never thought about using add detail at -1 for some reason 
What is that first part
Never?
ye nvm i just ran it to it btw can you tell me how do i get the a111 file the thing that makes it applied
That's the first thing I did with it
HOLYYyey8feiudfweoidiwenfowetgfyuwemf
like when you find low-fat milk
I GAYTTTTRTRRRTTTRRTTR SIMETRTHNTIKNNGGGGG IN NNNNN NMMTYYTYYY& MYTY EYEEEEE
Are you okay lol
that was the first thing i did 😮
You gay?
-1 detail best of the detailer lora
I mean, its one of the examples xD
Bro you copy what I said
Its new to me too tbh
Oh
didn't read it, we just think alike xD
lol
The what 
wait is this ai?
the file..
ye?
No
which one
and where
Except me
NO WAY
you are ai? i'm an ai too, greetings
like where do i put it
Where you put like
I am Aristotle incarnate
The LoRA?
No no
Open user bat file
Go to models folder
Then you find lora folder
As a language model, I'm not programmed to provide human interactions.
Put LoRAs in there
Then there is button (someone please screenshot to show) to show all the LoRAs you have
Then you click to choose
You put LoRAs here
Then you pick lora strength
ik that.
/weight
Good
im asking how to apply a1111(im dumbfounded and dumb)
H-
Wha
You
Got A1111
Now
scratch that im extremely dumb
Just put lora file in models/loras folder
You don't need to apply it
A1111 is the person that made the WebUI, you downloaded it, and installed it
You got it
Yes
You got A1111, trapped in your computer, torturing that man inside your PC to make AI art for you
Yes
aight dont get mad or annoyed but i was tryna get the png info and failed..
What
yes im stupid
??
Did you make sure to drag it into the space that says to drag images there
And download the image from Discord?
???
would you stop assisting me each damn step i take if i say it?
I am confuse
Did you make sure to click "Open Original"?
Man we need to get on a call or something
OHHH where.
When you open a image on discord, you get a little text at the bottom
You don't have to do that I am pretty sure
I mean, just to make sure
I don't think its big enough to get processed by discord, but maybe it is
it says open in browser?
that it
high slit
is the best prompt ever 
aight wait
wifi is tripping balls rn
@magic wave it finally worked( and im sorry for being so dumb for not understanding fr)
Should be pretty straightforward
I returned
btw which safetensor file of aurora should i download?
donzo
FP 16
nice okay i think ik how you get your style
is from the two loras you sent
cyancapsule?
send meta data dawg im quivering with excitement
Yeah, that's one of them
you use both of green and cyan?
Green for normal, and cyan for adetailer prompt
Prompt:
(1girl, one girl), (solo:1.3), (red eyes:1.3), (bright red hair:1.5), (ponytail:1.3), (wavy hair:1.4), (curly hair:1.5), (half-closed eyes, evil, light smile:1.3), (pale skin:1.3), (white skin:1.3), BREAK
((cowboy shot:1.4)),
(sitting, crossed legs, crossed arms:1.3), (thick thighs, curvy:1.3), (large breasts:1.3), BREAK
(black dress, elegant dress, ornate clothes, ornate dress:1.2), (fur trim:1.4), (white fur:1.2), (black coat:1.3), (black sweater dress:1.3), BREAK
(black background, simple background:1.5), BREAK
best quality, masterpiece, detailed background, highly detailed, intricate, detailed face, long eyelashes, detailed eyes,
<lora:Add Detail 1:-1>
Negative prompt: verybadimagenegative_v1.3, AuroraNegative, KHFB, negative_hand-neg, multicolored hair, (extra fingers, deformed hands, polydactyl:1.5), (nipples:1.3), (letterboxed, borders, white border, black border:1.2), (extra digit, fewer digits:1.2), bad anatomy, (worst quality, low quality:1.4), (text, logo, signature, artist signature:1.2), monochrome, worst face,
Steps: 30, Sampler: DPM++ 3M SDE Karras, CFG scale: 7, Seed: 3102167709, Size: 550x750, Model hash: 14879728af, Model: 絶望, VAE hash: f88e1cfb7c, VAE: Blessed kl-f8-Anime2.ckpt, Denoising strength: 0.52, Clip skip: 2
mostly the same thing
just changed the _ to spaces
u used a marged model?
ye
make sense why it looked so anime
https://github.com/Bing-su/adetailer
This is Adetailer if you didn't know btw
is it the same? same?
id rather ask what kind of hand embedding/lora u use for near perfect hanfs
I think so? I'm pretty sure I just removed the _
Aurora is really good for hands, lemme find the embedding
I don't even know why I'm using the aurora neg embeddings
I didn't merge aurora into my model
how do i apply the embedding which fiile?
You put the embedding in the embedding folder and then just type it out in the prompt
You can copy and paste my negatives
like Negative_hands?
Yeah
here right?
Yes
aight mk whatelse...
looks like upgrading my performance is done
aight i appreciate @magic wave @brittle jay the huge help im grateful and everybody else in this server too
and E face ofc
NAI is the best for training
Nice 
ill continue this maybe tmrow cuz its 1am in the morning and my aurora model is still downloading
I'd not dare say, i've never trained models, only mixed. i do know my best mixes were trying to get rid of the stuff that aom3 introduced
NAI is always good
Unless you're doing SDXL, NAI is the way to go
I don't think so
It don't think it's either of those
V2 it isn't, and I'm not sure what a parametrization model is, but probably isn't one
It is yeah
for 1.5 models, my mix darkalfa is pretty smart
i think i cant even make a mix of my models cuz of my storage or RAM neitrher the less i use whats best for me
They're asking for training tho
that i really can't help with, just say which models i regard as smart, which should make decent training bases
Try to achieve 1600 training steps, 2 repeats, 5e-4 learning rate, 768 res, batch 2, and 150 training pics, for maximum chance for it to be good
Those are the settings I go for and the amount of pics I usually get
but for training, i'd simply go with models with the most "brains"... that's all i'd go for.
Depends on your lr, but 1600 with 5e-4 lr is good
A good model needs good compatibility the most I'd say
I do 2 batch size, it's not related
But if you have too big of a batch size the AI can cheat at learning and not learn well
Derrian Distro on Project AI discord
Correct
Too little imo
OH, Yeah yeah
I read "25 or 40"
Do 2 repeats, and try to get to 1600
Yeah, and also low repeats are good for training
This cool
Guys be lazy
Don't use adamw
Use DAdaptation
With very ez settings
No need to mess with them
And you get fine results
Here config for kohya
With dadaptation
Too complicated
I myself don't understand
Linear algebra thingys
There is no real standard
Everyone uses what they see fit / what they like
AdamW good
Most popular
But you have to play with it a bit to get good settings for each dataset
With dadaptation you just throw in dataset and all good
Yes
Well if you use recommended settings with adamw you should also be fine
But personal preference of mine since I am lazy is dadaptation
You can do preview images on kohya to see the training process
It will output an image every let's say 100 steps (reasonable) so you can see how it changes over time
You just need to set it up so it does that
And input a testing prompt
This config good enough
Also to fix that just use the non 8bit one, should work
Yes
Either follow @magic wave or me though, not both, since there are different settings you need to use with dadaptation compared to adamw
Ok I go slep
Gn
@magic wave is it me or did aurora gradually decreased my waiting time to below an hour
Its no 20+ minutes
Im so glad i changed it thx to you
Anyways good night
Ty again
2B fron Nier: Automata there?
Anmitsumimimi + internet yamero lora
I don’t have the metadata now since I’m not at computer
Day's hard work done
Is that a pun?
Yes
should add blue eyes to the prompt to give it just a little nudge towards looking more like 2B. other than that it's pretty nice.
I always appreciate 2Bestgirl gens.
I have never seen any of the nier games or anime 
My friend plays with her skin in naraka I make funny prompts with the lora :p
You're missing out. The games are masterpieces and the anime is quite good too. I'd personally recommend starting with Automata, then Replicant, and last the anime.
This thing is trolling me
I'm trying to make her hold the glasses while wearing it
AI decides to get another pair and put over the camera 
She is technically holding her glasses and wearing it but... pls bro
try with (holding glasses : 1.9)
Thanks

adjusting glasses is the best tag for what you're trying to get
I gave up
ded
I tried that too 
lets go all in (holding glasses : 4.5)
haha
model?
Hmmm one of KizukiAi's models, don't remember which one was that
There's only 2 they are pretty similar
very cute
@pure iris do devs have mod powers here? no mods online

things escalated
I'd prefer not to have the server shut down by Discord
reminds me I need to finish re:zero
oh no u gonna watch season 2?
ayo someone spaning something thats not anime woman get rid of it
yeah, finish it
too much to do
how do I fit in waifu generation amongst it all
thats the season with 85% flashbacks and 15% real progress in plot so gl 🫡
rip
hows he not banned yet???
cuz @gritty spruce and others are sleepin
least insane naruto fan
least hygenic naruto fan
id prefer he would get a life but damn i think he doesnt have one thats sad
thank you pinging and everyone for keeping an eye! the ban hammer has cometh
thank you for keeping this server clean
and VERY kind fr
I gift you guys an anime comfy setup I was doing back in sdxl 0.9, just drop the image into comfy and it should load. Back then, I was testing with prompting different stuff to clip-G and L, it was kinda interesting for anime
what i missed? banned for nsfw?
a bunch of pooorn
some dude with all the rage in the world inside, shitposted every channel and a bunch of porn images here
damn, must have been funny to read XD
I was having my morning coffee and saw the ping, it was pretty bad
we are better without it lol
pretty sad to get a bad morning
ty for the good work, man
you are our man
please always ping any of us if you see this again, even if we're not online, that type of content can be really bad for the server
i really like the clean feeling of this server and the creative use of ai
i'm all in for preserving a clean environment
wow, that's injection, using ai to generate the image of an ai lol
I'm curious what model is this too 
very detailed, maybe an sdxl?
@lament osprey , @magic wave and Shaikorth,
Model is RealVisXL V1.0 probably the best ultra photo realistic model available for SDXL right now, by the same creator of Realistic Vision V5.1 for SD 1.5. Link: https://civitai.com/models/139562/realvisxl-v10
I just copied and pasted and then changed, the freckles girl prompt from the example and gave it an anime twist
as expected!
A few more:
i knew it!
yep, SDXL
the details are magnificent. what's the max resolution of an sdxl?
Ohhh fair then :o
too bad i can't run them even with comfy
I don't know the max resolution for SDXL, I know there is a list, I think in the paper SAI lists all the resolutions. These images are not upscaled, as I don't have an upscaler in my setup yet, these are in the basic 1024 x 1024 so they do run direct in comfy (I'm not using comfy though)
you can generate at 3840x2160 without too much trouble if you're careful
How's Japan @brittle jay?
Very cool image I like
That halftone hair is really nice. Is the image edited @lyric zealot?
it's really just the work of a lora
nice bullet trains sound like a fun experience
i was spending a lot of time to get the fingers to generate nicely
Those came out very nice. I assumed at least the hand was probably redrawn.
still no prompt inspector in this server 
embedding and adetailer are doing miracles nowadays

i get perfect hands more often than else
protip: you can send images to Prompt-Dog in DM and right click the image > Apps > View Parameters to get prompts for images sent in here.
(masterpiece, best quality:1.2), detailed background,
1girl, pink hair, high ponytail, yellow eyes, from side, mask pull, jirai kei, bandaid, stud earrings, piercing, open mouth, smile,
<lora:Vampire:0.9> vampire
Negative prompt: (low quality, worst quality:1.4), (bad anatomy), extra digit, fewer digits, (extra arms:1.2), bad hands, by (bad-artist:0.6), bad-image-v2-39000, negative_hand-neg
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 22, Seed: 2717316180, Size: 512x768, Model hash: 931aaff934, Model: expmixLine_v3, Denoising strength: 0.45, Clip skip: 2, ADetailer model: face_yolov8n.pt, ADetailer confidence: 0.3, ADetailer dilate/erode: 32, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer version: 23.7.8, Hires upscale: 2, Hires steps: 10, Hires upscaler: 4x_fatal_Anime_500000_G, Lora hashes: "Vampire: 5a5da562cc0b", Dynamic thresholding enabled: True, Mimic scale: 7, Threshold percentile: 96, Mimic mode: Half Cosine Up, Mimic scale minimum: 4, CFG mode: Half Cosine Up, CFG scale minimum: 4, TI hashes: "bad-artist: 2d356134903e, bad-image-v2-39000: 5b9281d7c6ad, negative_hand-neg: 73b524a2da12", Version: v1.5.1
Template: (masterpiece, best quality:1.2), detailed background,
1girl, pink hair, high ponytail, yellow eyes, from side, mask pull, jirai kei, bandaid, stud earrings, piercing, open mouth, smile,
<lora:Vampire:0.9> vampire
Negative Template: (low quality, worst quality:1.4), (bad anatomy), extra digit, fewer digits, (extra arms:1.2), bad hands, by (bad-artist:0.6), bad-image-v2-39000, negative_hand-neg```
there you go 

You can see much beauty
I only like milfs if it's mother earth
Man I requested it like such a long time ago
I'd like to go to Japan some day. I've got a lot of studying to do first to be confident enough in Japanese to actually go though.
I just only speak in english
I am that one annoying foreigner
lol
at least you know how to speak well a lenguage lol
i'm half a**ing even my native lenguage lol
I'm the type that doesn't like to stick to the typical tourist areas when I travel so I'd like to know enough of the language to get by before going.
sounds like most americans with english XD
I mean I guess I am pretty good in both my native language and in enlgiski
Oh yeah if you do that in japan
You gotta be better than Japanese people
lol
Oh yeah btw @magic wave
Luna fire Luna water and now Luna earth
short hair suits her
Did you notice how noone refers to hair as medium length almost at all
Like wtf is medium length anymore
It's either long or short
Even when in sd
I do medium 
If you do medium in sd it just gives you short
Or long
Never actually medium
At least from my experience
I get medium :o
The one I'm really disappointed in here is absurdly long hair
exactly like my true set up!
Not a single cable is visible for more than 2" under my desk
Medium here is literally just long hair
Where's very long hair 
Oh yeah where is it
comparing side by side, it's a little shorter https://cdn.discordapp.com/emojis/1087599703128031313.webp?size=64&name=ehehe&quality=lossless
My PC setup be like
The waifu too
Cause I have stickers on PC with waifus on them
I make stickers for a living... I think the only stickers I have on things I own are my business decals on my truck 🤣
Oh shit fr?
yesh
Dayum
Why I only hear about this now
I'm sure I've said it before
Holy ahit
Actually the hair lengths specified in the prompt
This is a first
Gotta be
Idk
heres a shirt that I made for myself to celebrate the FFXIV Endwalker launch
the visible rectangle around the design is just because I took the pic right after pulling it off the press. that goes away after the first wash.
Now I'm sad I'm working in someone's garden, no time to SD
When an assembly of regulars is here
Wait, I can, I left my pc on public SD lol
Remote SD gooo
Strange axe
They got the Shovaxe and the Halbaxe
Bruh

😳😳
You will never guess
Me and my step brother are supposed to get to our hotel today
And we just now found out that it's a love hotel
Bruh
I'm 17
And a pure boy
And also it will be extremely sus to be with my step brother there
We are in the same room
We were sure it was a normal hotel
maybe more of a doujin plot
test
True 
lol
Kinda true ig
*yet

Not of laughter
I thought it was a reference, and I just found out Reimu and Marisa are best friends, I thought they was sisters
This reminded me that I could try to gen some more yuri
Ohhh, those two are sisters, I didn't know that either :o
yep komeiji koishi and komeiji satori
Now you have to share tho
||warning:hand holding||
quick question guys whats the average size yall use?
default 512x512
only 512x512 for u?
sometime 960x540
yet you generate high resulutions?
depend on the mood
yes I multiply them all by 2
so 512x512 become 1024x1024
512x512, 768x512 and 512*768 usually
oh btw uhh my SD said Could not find (my sd VAe)
WHARR i doo
i now use aurora
does it not use vae or is it not compatible?
boba
your too addicted to this chibi
why don't you look at my pfp and tell me
Sung Jin woo
so your a boba addict AHA im calling the copss
boba isn't illegal
he looks like hes wearing stones
ik im being sarcastic
Ik
Ok I just got to the hotel
And there was a tv in front of the bed (only 1 bed for me and my brother)
So I opened the tv to see what is there
And you could guess what was there
Also there are "toys" and various stuff here wtf
Well only to be expected
Cool
Me when each one here has created at least one character a ton of times
My sd vae isn't vaeing
Like nubby has that one coffee drinking girl
IT GIVES LIKE A PISS FILTER
Aurora
At least there is background music
Its Kl anime4k (Vae) or something
You can just use the WD 1.4 vae
4k?
What's that.
Ye That's the VAE.
I mean show the image you were talking about
hehe i was copyin it last night
This
Show example
Bruh
I sent it
Btw i can't send more because thats just the only one i generated
I got knocked out after waiting a bit
And woke up
What are you taking about?
Are you just saying the white is not white enough?
Done it's fixed.
Go to settings and change vae to yours
Oh ok nice
Another day another dumb me
Ez
But thx for the help like always@brittle jay
”So What” by Miles Davis
Listen to Miles Davis: https://MilesDavis.lnk.to/listenYD
Watch more Miles Davis videos: https://MilesDavis.lnk.to/listenYD/youtube
Subscribe to the official Miles Davis YouTube channel: https://MilesDavis.lnk.to/subscribeYD
Follow Miles Davis:
Facebook: https://MilesDavis.lnk.to/followFI
Instagram: https://MilesDavi...
Sounds so good with my new IEMs

Somehow an entry level IEM has almost better if not actually better sound separation than my shp9500s
Now make her drown in boba
Honestly this could make good stickers for server's
This is so OG I forgot who made it lol
Done chopping wood. Now it's bbq and relax time
Nerfgun
Oh yeah
The one and only
I like this one
Now with bbq
Damn
@brittle jay dm 🙂
Did anyone tackle the idea of Kaguya Yamai from Date-a-Live fame with a sword, before?
Gave him acces to my public SD instance, as I can't use it atm
Yeah, was helping friends chop and clean up wood piles
so cute ❤️
ty
my generic model has a really hard time getting that kind of chibis
maybe it's time for you to switch?
no, i can't get anime models to work ever
really? every prompt I took from you work splendidly
must be a me problem, or not using the right models
maybe you just haven't found the right model yet
the only 1.5 anime model that satisfied me was dark alfa
which my only gripe with it was, that the hands were worse than on alfahentai, which had absurdly good hands for a 1.5 model
but i managed to get it outfitted with a lot of general knowledge
no chibi's with boba tea for me
armored chibi
why he look so mad lmao
because he was not allowed to get boba tea
he felt denied
I see more 'sdxl' models popping up on civit. Is there a trick to using these with Auto?
'holds boba tea hostage'
1.6 can use them natively
some models don't even need the refiner anymore
you just need to meet the huge vram requirement, as a1111 is less vram friendly than comfyui
Right on, thanks @wispy canopy 🙂
I messed with the SDXL release on Comfy a little bit but didn't get super into it.
i'm on an SDXL model which doesn't need refiner
Yeah?
as ever, generic model
yeah, but sdxl is like HUGE smart
so you can get away with normal models now
fine you win here you can have it just let go of the boba
👀
boba tea mecha!
tengen toppa guren lagann power!
More boba-bots, plz
speaking of boba with face
ahh
seems half of it is
I'm gonna need a cloud storage service to start storing these dang models... 😛
that's why i loved my model: it got it right every time -> intelligence in a model simply trumps every other thing
you can make fancy output with loras, but you can't replace internal intelligence
rainbow hair and bismuth armor!
my eyes. she's magnificent!
shiny
yeah ik boba just doesn't have robot body so I didn't realized at first
damn I made typo to his name by going along accidentally

boba power
wait what if we made chibi soldier fight
full modern military style
introduce the chibi soldier
@wispy canopy lmfao is she just holding jpeg image of dirt???
🤣
I'm dying never had AI do this before
i was tryign to get my prompts done, i'm still stuck at 30k out of 50k images
why you making that much all at once?
nah, just going through all my old prompts
i've been doing that for the last... 2 weeks? XD
yeah
weird af
they are attacking!
ai just gave up
don't let anyone get away shoot all the runner!
I mean we both using star rn lmao
your look like allies and the one on my shoulder kinda look like soviet star

I didn't even put star in the prompt
and we can use make up symbol as emblem instead
botched enough. have the bad guys shooters
get them push back the front line
I also have a few where AI make perfect emblem of those guys lmao
I decided to skip them
corporate allies have arrived
is that....toyota?
or traitors... mercedes being germany based lol
nearly correct toyota
they can't hold the line use the SPEED
dunno why last one kinda look like police helmet
speed u say?
eyo that's not chibi
here we go proper bad guy emblem
can't mistaken this

wholesome bad guys
final fantasy got involved too, they grabbed their suitcases: "we're out boys"
lol, oh man... good luck getting this image out of your head. 🙂
which model are you on?
are you generating at least at 768x768?
sdxl models don't work well under that resolution
always time for boba tea
yeah i'm just messing around with all kinds of stuff. I think they work best at 1024
new objective pillage enemy's boba
"Reusing loaded model sdxlNijiSpecial_sdxlNijiSE.safetensors [fa892e66c0] to load sdxlFaetastic_v16.safetensors2.00it/s]" -- I get this when swapping models, what does this mean exactly?
refiner?
gib boba or we'll drink your boblood
you don't scare me, once i find the magic words to counter you, i will fight back!
this one does interesting maps...
texan pirate map?
like... loot the whole state?
the coolaaa
that's where the coke originate
Here's the model: https://civitai.com/models/129681/sdxl-faetastic
ok. imma go from this war, moving back to my history 😄
i mean, this is important too
what vae are people using with these sdxl models?
i'm on protovision -> built in vae
damn my soldier quarrel among themselves due to lack of boba supply guess we can't fight on either
built in vae
they seriously went mad, props ❤️
the right one is just about to stab with a barrel, is n't she?
interesting style...
right one use a lot of face powder
#2 even mre
niice
do you know why i feel limited by anime models now? 😮
if i can't pull off shit like that, i'm a sad boi
early anime models simply couldn't do it
at all
but this's literally your prompt
🤣
nowadays anime model are good
you should give them a try
nah, having too much fun with this 😄
plus i can switch to doing alternative stuff like gundam rallying national geographic photos the moment i want it
😛
that's a weirdly specific thing
by now, you should've realized i'm doing weirdly specific prompts 😛
quite a different result lmao
yeah, same for me, the og model just couldn't do it
retrying my winning pow randomizer prompt. (not the actual winning prompt) (theme was chinese newyear, year of the rabbit)
uhh... 1989 wants their iconic tank back... i think
that look like m60 turret on a t55 chassis lol
man these sdxl models are tough to use.. I'm bad at this. 😄
they're easy to use ❤️
dang, salt in the wound
try this model
if you want anime output, just start your prompt with 'anime cell shading'
no refiner needed
i'll scope it out
this pic give so many diff atmosphere lmao
is that the half-dead prompt?
yes yes it is
damn, sdxl would've SLAYED PoW 6 months ago
also I'm rolling out a tank too
u tried the rabbit prompt on non sdxl model?
no diff prompt
the actual prompt is __mediums__ by (__manga_artists__:1.2), __movements__ , rabbit, chinese newyear, (ultra detailed:1.1), anime {vector|line|sketch|oil|water|ink} {splash|splatter|stroke} art, inspired by ( __artists__:1.2) style, , {__locations_buildings__|__locations_other__|__locations_indoors_main__|__locations_outdoors_man_made__|__locations_outdoors_natural__}, {fire|ice|wind|life|death|crystal|heart|light|sun|solar|lunar|celestial|moon|split|duality|reverse} themed , (best quality:1.3),(highres :1.1), surreal,
only part that's the same is "tank interior"
so uhh.. yeah.... lots of random going on 😛
I'll try that and we'll see what I got lol
until this batch done here have the new russian tank
i mean... that was my previous tactic. keep the random rolling and wait until something really cool pops up, and then keep generation variations of those
this one's burning
this indicates performance
holy F... i did over 700 random attempts of a good random output 😮
here's the 1st result
but then again, some are truly amazing
aight. lets try the winning prompt
lets see if sdxl can make it magic too
oh no. i think this is the first time sdxl fucks it up, because... it's better lol
it actually knows what i asked so it gives me... instead of glitching out
the winning prompt... now it's just an illustration. nice illustration, but not PoW winning by any stretch
yea SDXL can't give the same glitch effect the 1.5 got
this was the winning image on sd 1.5 based model
lemme try my award winning prompt too
I highly doubt it'd be as good for the same reason
ooooh... i had a wrong artist... not the EXACT prompt
this one gets a bit more abstract ❤️
not a lot, but a bit better
these aren't bad at all, actually
oh shit... an encounter with my first sd prompt ever 😮
yeah it look nothing alike
I'm waiting for other 3 to finished generate
glad I still kept the award winning model around

❤️
corona screwed up facial masks for a long time in a lot of models lol
medical masks every where
sad
it's the cut off tail of her enemies
🤣
I couldn't get SDXL to give black sclera
that's like 1 main downside for me
well it might be possible in rly short prompt
but if it's long everything else will overpowered it
just "black sclera" prompt really I put that in lot of thing
now it just don't really work with anything really
nope, doesn't want to
oh.
oh fuc.
i... uhh...
this prompt is amazingly stupid lol
warning, non anime
👀
but no black sclera
cyclops!!
real cursed lol
at least 2nd one is trying to give black sclera I feels like
🤣
but more like a cut out contact lense
to get your pupils that big... you need a LOT of drugs
that's why i hate realism XD
here, have another of my earliest prompts ever ported to sdxl
such curses are hard to epurate from my mind XD
so fresh
oh look it's the bard
nice
have a random selection of waifus
ya'll tip for carryout pizza?
No idea, but you can probably try generating yourself
no
is that your honest reaction
hm?




