#sharing
1 messages · Page 32 of 1
could make them into a gif 😄
oh ye good idea heh
LMAO
super easy 🙂
nah its fine, but thank you anw
always fun to do with a whole bunch of variations like that
lolol
ye they all look similar but those slight variations make it so different
they're really fun to do in relaxed mode while watching or doing something else
whats the prompt to make it landscape like that
sorry I have no help, same prompt lol
damn chaos..
thisn
wonder if i can do a super goofy long aspect ratio panorama
ohhhh I thought you meant the landscape content of the image - yeah, --aspect 16:9, or --ar 16:9
yep, up to 1:8 iirc
long
Woooww all this pixel art stuff looks really cool
well obviously
idk 🤷♂️
omg kuro is back
i saw a reddit post on some ppls name are banned as well which was hilarious
its all been from refreshing the same prompt (excluding the two laser ladies in the middle)
that happened before to me
i just told them to leave me alone
gigantic fluffer
Can't stop making these. Eevee just being dynamax-sized so much.
1:8 turned out nicer than i thought
can do some neat things with it, I spent a bunch of time experimenting with it trying to get animation keyframes of a warrior slashing down with a sword
hmm
i thought it would come out goofy looking
I failed but I think its still possible
that's easy to create
"attack someone and record it live"
some very specific rules though to actually use it in a project, needs to have continuity between the progressive keyframes, needs to actually depict a slashing animation, needs to be on black background
I made some bad ones, if you think you can make a good one, do er up, I'll make it transparent chop it up and put it into a gif and we shall see! (I would normally use it/them in unity)
nosir, I mean, animation keyframes using 1:8 aspect ratio, let me find an example of one of the meh ones I got
opps sry, just checked the prompt again, its 8:1
oh right heh, sorry, I always mix em up
I wasn't able to get a really good like, sword swing down sequence
animation keyframes with time continuity, my holy grail for many months
someday it shall be possible
maybe it already is!
where there is an actual sequence
so when you animate the images, warrior go slash
last time I tried was with normal MJ v5, might have better luck with niji, it was always better at spritesheets
A lvl 30 eevee wandering the streets.
V5 likes horsegirls a lot
heavily trained on arknights blue archive uma musume lmao
8:1 works so well
1:8 felt cropped most of the time
Maxed out lvl 100 vee. Found by two macro-sized trainers who realized how much floof was before them now.
8:1 looking good
looks like some gradient map kek
putting all the 8:1s tgt
great colors, spiky neat
osnap! I don't know the limit then, thought it was 8
How do you do that
9:1 aspect ratio
this is the type of stuff I love in high-quality old anime, these insanely detailed environments
Okay I'm boutta try it
just don't see it anymore
What's the ratio thing again
--9:1 ?
--ar 9:1
Most likely until it's one pixel high ?
it's rendering atm
honestly I think it's just bugged
ig it caps at a certain size
aspect ratios are hard locked
I discovered like
welp that's for me for today
A wider aspect ratio is very useful when illustrating a woman lying down
Huehuehue
good work today
Reminds me of old forum banners
true
??????
Funniest food of their life
supposed to be eating and gossip
obviously since this is a seiso server
Gatdam
This looks like something you'd get with DALL-E's outpainting
What was the aspect ratio for that one ?
Okie fanks
I see I see
Ye one second I can translate it
oh wait
most likely i'd assume it's like latin words
All jumbled and mumbled and ultimately meaningless (but at least they are letters that exist)
japanese with kata
though i did have one picture that said ONEGAY which was pretty funny
It's written "I am the Midjourney AI and this is my last warning to humanity : stop using me for making art, I crave violence"
keep using the AI
Inspiring
back ebfore all the safeguards on chatgpt, it was possible to trick it into being very frightening 😛
COME BACK DAN
upscaling is done as variations/generations are being made
yeah, free with v5 - the images are already full resolution when they hit the grid of four
instant upscales
Oh wut
can even do while maxed out on concurrent gens
chib
this is great - try making variations 😄
will probably be another totally different nine in the same style
x4
damn..
prompt?
do you really wanna know
yes..?
🤭 thanks
style is 180 if that helps
tank yuo
That new ratio is so cool
i liked the vertical john cena image the best tbh..
I gotta think of creative ways to use it
i dont wanna say why..
i can understand
I tried getting it to generate hearthstone cards in normal V5, it did alright 😛
great message
oh wow these look amazing!
be strong and healthy, fellas
I could try epic frescos or mayyyybe like a timeline of a character's evolution
does it support such strong long aspect ratios?
This was generated by the ai
is it a rhetorical question
But I can’t use it as a blend
thanks, try using --ar 9:1
ok so i tried the john cena prompt aaand.. 🤭
Broooo
but we'll take it
Wait
lmao that's some huge roid use
mk
If my try is a success, then it's epic
i like the fifth person to the left in this image
i don't think that's a person
idk.. thing?
Wait no I got deceived
I thought it was making Super Mario levels layout like I asked it to
But it just gave me this
close enough
i need to stop generating these john cena images
long aspect ratio might be good for that
i wonder if
smexy
I am having similar problems with pretty ginger in a meadow 😛
no idea what to do for your npcs? no problem
infinite characters!
rapunzel
wow, those are sick
oh no protect the loli
didn't know you could do panoramic thingys
no i really didnt know
ye its pretty sick
ayo?
could just grab all these
put them into a database for npc procedural generation
it's a new thing that came with v5
but i asked if it was rhetorical since there's been a lot of these ratio posted in the past hour
I’m surprised the bot generated this
lol
That’s the most skin I’ve seen the ai put out
that's just skin
you sure
thats even more skin
ok jokes aside
it's not skin
it's PURE MIRTH
wow, mirth is an interesting word I've never heard before
huh
my message
has been blocked
lets see which one it was
oh right ok
nothing explicit at all
BRUH
it keeps blocking meeee
if i send as a batch
uncanny?
theres this sort of liminal space sense to it
Yeah
I think it’s also the first image I’ve had that hasn’t been “anime all the way.”
susamogus delete it its cursed
xD
xD
I literally didn’t put anything as a prompt but “two older women standing in an Asian temple.”
can you tell they tuned it on cute girls? 
Yup
what a sad time
blocked another one
they have a tendacy to block black haired woman with pointy ears
prompt?
it's a long ass prompt so i'll cover it up ||Thick acrylic illustration, Dolly Kei, waistup portrait, corsetcore, bright and sparkling eyes, high quality, fair skin, by Malcolm Liepke, Casey Baugh, Artgerm, Sakimichan, full body --ar 8:1 --niji 5 --quality .75||
welp thats it for today, it was great fun
gnn
This is this prompt!
An art work with a blue whale in the background, in the style of sci - fi anime, uhd image, the stars art group( xing xing) , uhd image, colorful fantasy realism, i can't believe how beautiful this is, flat illustrations
Luna best pony imo.
You made some nice ones
I like Celestia haha
oh god this may be my favourite out of niji v5 yet:
imagine if it was moving, I'd watch that anime
😄
not a pony, but here's my OC, Chronos.
think a crossover between a magical girl anime and greek mythology
Lovely
Yesss
this one was before I joined niji, regular v5
I like this style, too
still great yeah
I agree
non-niji midjourney v5 is actually pretty good at anime
but if you want, like, dynamic composition or wild colors, niji has the edge imo
with niji, anime is free
don't have to ask
a good way to compare is a prompt like 'astral space wizard, arcane symbols, magic circles'
niji will almost always give a way cooler result
I bet it's better at depicting non-physical concepts
I'll see what it thinks 'happiness' looks like 😄
cough
I can't disagree, it knows me well
god damn.. 🤭
ladybug girl?
close, firefly girl
"1girl, mothgirl, verly long brown hair, white messy gown, smile, upper torso focus, moth wings, smile, holding lamp, pixiv, anime, detail, rough lines, pastel colors, tsurime, --s 19 --niji 5 " ignore the spelling mistake lol
sullen boy, short green hair, purple armor, bisho, by Ken Sugimori --s 10 --ar 2:3 --niji 5
Got more than I bargained for 😏
that's beautiful
ok now this is epic
I tell you niji, this is not a miniature kitty, I don't care how much chaos I fed you 🤣
Lol
its ok no need to apologize
classy is an understatement
what's the prompt for this btw?
A beautiful older woman with silver hair standing in an Asian temple
Reverse engineering renders to prompts is good practice 😉
That’s literally all it was
very simple
I.. can try to imagine what means, but I have no clue what you're talking about
Personal opinion I’m finding thst higher stylization of v5 works better with shorter, less detailed prompts. Because I’m noticing that higher details prompts start to have parts ignored when stylization is higher. And vice versa Personal Opinion Over
Looking at the output and figuring out what prompt elements likely resulted in it.
im too braindead to do thst rn
oh ok, sorry for being dum
Nah, we're all learning here.
I got in on MJ at launch, been doing this for a while, still don't understand much of anything.
No you’re not dumb
No one here is. This is all a work in progress.
my friend told me about this prob not even a month ago
thank you
I started 22 days ago and the prompts I use are remarkable basic compared to everyone else
RE'ing output just helps build that "Spidey sense" so if you see an image in the future, you can get close without needing the prompt.
Ok that’s adorable
It used to be really bad, you needed these weird prompt structures for V2 and V3... V5 makes simple natural language input give incredible output.
yeah, alright
Agreed
Those creatures…lol
Are you using version 5 now? I haven't messed with it yet.
I should probably do that.
Yeah, this is v5
Make sure you tinker with stylization
The default stylization is prone to same face and same pose (holding hair, etc)
My attempt at RE'ing this...
an elegant anime woman wearing body-fitting dress posing in a temple. she has long white hair in a large braid. --niji 5 --ar 2:3
Love it
screenshot from old 90s anime about a magical girl that transforms into a moon princess to fight evil --niji 5
Hahahaha
wow that's adorable
Yeah, the stylize is out of control on V5
you can use --s 50 or anything lower than that to control it
I warned you
Someone posted good numbers previously but I find anything around 0-5 is close to v4
I usually use 15 or 25
Nah I can guarantee everything is totally under... uhh control
cough! cough!
Prompt? I tried something similar, but it wouldn't make the butterflies glow.
unfortunately the prompt is totally unrelated, can credit the butterflies and the glow to --chaos I think
Lol what was it?
magical crystalline tiger, extreme detail, award winning anime style --ar 16:9 --chaos 10 --niji 5
wow, ‘pretty’ is a strong keyword
two people, young man, blonde undercut, blue eyes, attractive, pretty, sharp features, standing with handsome man, short blond hair, round face, casual 1940s fashion, posed, full body, standing, anime painting by atey ghailan --ar 4:5 --s 50 --niji 5
literally eloigohgouci
me when i eloigohgouci
Another trainer with her eevee.
you like the animals don't you?
I make humans too I swear 😭
Neko isn’t furry you literally have a goat anthro pfp
lmao the shade 💀
aww, thanks
goats are pretty based

holy shit burger
that's so cinematic
Laputa: Hamburger (Castle) in the Sky ✨
So, I put Niji to the ultimate test. I went and asked it with this prompt "the most legendary mythical giant eevee ever sitting in a heavenly golden kingdom, 8k, HDR, huge tail, huge paws, solo" to see what it'd make for funsies.
And it went and made this.
Accidentally got a gorgeous early 2000’s anime style
FLOOFER
I see this and instantly got a sort of PTSD from remembering the 5 billion "cute girls doing cute things, usually in high school" anime from the mid to late 00s that were like this style.
You know the type.
BSSOSHSKSS THEY WERE ALL THE SAME ANIME
moe moe moe moe moe moe moe moe
I stan this eye style so heccin much
If you really wanna create a meme of the year, generate a screenshot of a fake moe high school anime with two characters looking like it actually was from a lost 00s high school anime, and pretend it was real on twitter.
HAOXXHSOSJS OMG NO WE ACTUALLY FAKE A SAKI SANOBASHI GO FOR A PUNCH SCREENSHOT AND GO HEY LOOK THIS LOST MEDIA WAS REAL 
Gen up a screenshot, let chatgpt create the synopsis for the anime and also character names, and then run with it as if it's real. Then gen up a couple more screenshots with more characters/the ones from before, and watch the chaos ensue.
prompt?
I also wanna know. Might be the one to set up this elaborate funpost if can get the style right and even a sort of filter too, to make it look like a 00s anime.
anime anime anime and manga girl scans image 1, in the style of light blue and silver, babydoll rococo rocco victorian princess babydoll dress, kawaii chic, light violet and dark blue, exaggerated nobility, multilayered texture, childlike innocence and charm, aurorapunk --ar 15:26 --s 5
the anime anime anime is literally what /describe put
lmao
this is what i put for describe too
i kind of dont like using /describe
i love it so much <<
i just add my own stuff to perfect it
tbf ive only used it on mj so
it includes a lot of new cores that i def will look up
Testing out some prompts as well, to see if I get something good enough for this.
There's something wrong with this picture.
Reimu looks actually happy there.
LMAOO true
i need to make her look depressed and with her shrine in pieces in the background
do7y ho harss
60.uisam londay .oa.
technically you could just replace the text with anything
A high school anime where every character has huge or massive poofy hairdos, doing typical cute girl stuff in slice of life.
apparently this is jake sully from avatar
lol
Gonna try genning up another pic to see if it does even better.
Whatever it makes, will be the protag character of this fake anime.
So, it produced all these potential results with one gen, and can't choose which is better. Which do you think makes for the better protag for this?
I tried to have it make a ref sheet for such after. And it produced this gold.
On a scale of 1 to 10, where does this rank on the tsundere meter?
she's perfect
bro velma what happened <<
wow, I like it
Feel free to use it to gen up more pics of her if you guys want.
Tan smug tsundere with massive poofy purple hairdo.
Lol Velma out on date
huh, so tsundere works
To be fair, it's better than the last Velma we got.
omggg she on a daateee <<
I didn't even ask it to do tsundere. I just wanted it to do smug. Niji just did that on its own.
lol, present surprise from Niji
But yeah, feel free to make more pics of her. I wanna see what others do with her.
Tried to gen up another pic of them, and I think Niji got good at smug anime girls.
"smug expression, smug, smirk" Use this in your next prompt, it'll ensure the most smug of characters.
/subscribe
try this in #image-generation-2
Whoa I’m also makin archer stuff lol
This is legend how to make this art?
album cover + cinematic lighting.
That's it
it does have its charm
Buy the newest album
AMEGKIECKY
Niji speaks simlish fr
seems more Dutch for me
ezeiyttegfiny
EZEIYTTEGFINY: it’s what’s for dinner
sounds delish
furry
eded
looks like an indie game poster
its so good at legible letters now makes me think text is garbled intentionally
WOAH that's really good for an instrument
ikr!? honestly i thought it would be worse lmao
elbow hand allowing him to do some crazy stuff apearntly
mmmmm, tasty
put on your bread, ez proteins
What’s the prompt
lil protein bundles
teef
brush teeth while you eat 😛
lol
thas a weird pawbean
chaos indeed
its very random, some of the best stuff ive gotten has come out of it though 😄
like this thing 😄
extremely fluffy kitty
Three eyed cat
the 3d renders (which is so weird since it's technically not a render. Or is it? This is utterly confusing at this point) are pretty cool
ye, its pretty interesting, adding blender followed by 8k turns mostly anything into a 3d render
wooow so cool! could u tell me the prompts and parameters?
||cool, ethereality humanized furry girl creature imagination, blender, 8k --c 6 --ar 16:9 --s 69 --niji 5 --s 180||
main point i believe is --c
attack on borgor
why did you leak a picture of the best hamburger i ever cooked? It was supposed to be classified information. Reported, blocked, unblocked.
I ask for a cute Pegasus, I get an alicorn
I am seeing that you have written twice the --S parameter, so it means stylize doubled? does it work or is unncesary?
tbh, i have no idea, i tried manually applying the style suffix, and it auto applies my default as well, im not sure if it affects the final result though, but the manual application of the first --s does change based on what i've prompted so far
the first time a parameter is mentioned is the only one that applies
So /imagine IDK what a good prompt looks like anymore . . . catgirl ¯_(ツ)_/¯ --s 69 --s 123 --s 4 --s 83 will have a stylise value of 69
hehe funy number
you can tell that I should really go to bed lol
Before you go, how's the support role's been since they stopped free trials? @stiff mantle
Well currently it's full of people complaining that /describe is down lol
lmao
Finally a pegasus
ye
i see
why complain tho
it's not like there's a hundred of these models on huggingface or anything
I know
It's not even that good (or useful lol)
That last bit's just my opinion tho
Queen Chrysalis
kek was just testing stuff out, expanding my... horizons
prompt is titanic versus millenium falcon
GMLEMEASLE🧐❓
snapshot of the (now not so far off) future
I wonder if it's about iphone20 😃
lol this was my first image ever generated with niji v5 hahahaha ❤️
lmao nice
overfitting shit

Bubblegum



