🍥｜anime

[rough oxide]

for single char

i have an easy way to get the char

i didn't implement a way to do it for multi char yet

[silk steppe]

they don't have EXACTLY the same

[rough oxide]

now i got right piccolo

but xeno goku

[silk steppe]

so there's that

[rough oxide]

ur fave goku lol

[silk steppe]

oh fuck yea

the peak form

[silk steppe]

doubt he had that in sayian saga tho lol

[rough oxide]

by base style i just mean

i keep rerolling the artists until i like it

then add all the scene details

like wut they're doing & bg and stuff

but what i mean is the script outputs the chars so i dun have to go and copy and paste from a txt file

nonstop xD

[silk steppe]

it sure keeps getting sjj4 damn

[rough oxide]

it's xeno goku

i told u

right piccolo wrong ogku now

xD

cuz evey son goku

has "son goku" in name

it's a problem with chars who have multiple forms/names

[silk steppe]

there's 500 gokus so ya know

[rough oxide]

now i got orange piccolo & xeno goku...

it's just gacha

lmao

my script is ai gacha

1girl

what happened here

oh

fio piccolo

from kurenai no buta

LOL

[silk steppe]

p5x devs would be proud

[rough oxide]

https://tenor.com/view/porco-rosso-fio-piccolo-gif-8555179522769796401

lmao

[silk steppe]

yea, that ain't the picolo chief

[rough oxide]

SO PICCOLO ALSO SHARES A NAME

WITH ANOTHER CHAR ENTIRELY

but his entire db entry is "piccolo"

so it's gacha

lmao

[silk steppe]

there's a whole monkey named son goku too

so ye know

and with gohan ya might just get the grandpa

[rough oxide]

i giv up and just random 3 chars from everything

i only recognize middle

is a pokemon char

[silk steppe]

idk any of hem

i'm too old school on animus ig lol

[rough oxide]

u dunno pokemon?

but idk the other 2

[silk steppe]

idk the char i mean

[rough oxide]

their designs r cool on left & right tho

[silk steppe]

ofc i know pokemon

[rough oxide]

i didn't even play her game

but i remember back in the day

[silk steppe]

they're fine ig

[rough oxide]

AI had a REALLY hard time with raifort's glasses

[silk steppe]

no clue man, i only played pkm emerald lol

i really don't like the franchise's gameplay

now that i think about it- i don't like any nintendo franchise

weird

[rough oxide]

she has unique eyewear

and the thing on her head

modern ai does her flawlessly now

she's from a pokemon game i didn't play

i think

[silk steppe]

it does look very pokemon

they got outrageous

char designs

[rough oxide]

this is my fave pokemon char

of all time

[silk steppe]

the e-girl ofc lmao

[rough oxide]

and the 2nd is fairy miku

[silk steppe]

last thing i saw that i liked was the robot pokemons

[rough oxide]

fairy miku was actually from pokemon

an official collab

lol

[silk steppe]

it's a fucking CRIME you don't go to the future

in that pokemon game

~like cmon bro, robot pokemons

is the collest shit since

well that's the only cool shit in pokemon

[rough oxide]

for designs gen 4-5 were best era

everything after feels a bit put together

still good jrpg tho

and they still have ideas for char designs clearly

they totally fit together so well tho

pokemon was ballsy enough to collab with vocaloid for the mikus but...

i wish they appeared in game

[silk steppe]

i hate the gameplay, and how childish the franchise is

but then again it is for kids

so wtv

[rough oxide]

it has many layers

it's beyond just for kids

gen 1-2 when it was still simple might've been

[silk steppe]

nah, it has 0 grit

no edge

need some real shit

[rough oxide]

the strategic jrpg combat i mean

in terms of online play anyways

the devs can NOT make good story experiences

but like compass mod for scarlet/violet?

[silk steppe]

, gameplay is ass too, last game i did try was the one where ya go to the future

[rough oxide]

the rebalance & stuff too

[silk steppe]

it's still the same as emerald in the end lol

except for that game's new mechanic or cryrtalize

or wtv

at the end of the day you beat the game by over leveling 1 pokemon

it's real lame when you compare to persona

or ff

or expedition

but i get it, it has to be easy

it's for kids

i accept that i outgrew it lol

[rough oxide]

das why i said u gotta play the modded games

like compass for scar/vio completely rebalances the world, changed every trainer fight

[silk steppe]

i played some

[rough oxide]

feels like a smt-lite

[silk steppe]

the rom hacks

and some nds hacks

[rough oxide]

the good ones.

there's like 1000s

u gotta find the actual good ones

[silk steppe]

nah, no smt thanks lmao, that turns shit got me spinning

[rough oxide]

i mean difficulty wise

[silk steppe]

it was ok in metaphor tho

[rough oxide]

cuz smt v is notoriously difficult

even on normal

[silk steppe]

i really doubt that, there's no such thing as hard atlus game

[rough oxide]

with smt v 80% of the time when u die it's cuz u ran into a bad mob pack lmao

wat

the smt games are difficult

persona r not

[silk steppe]

nocturne was more annoying than hard

[rough oxide]

i never played smt 3 tho

[silk steppe]

nocturne is like a sleeping pill

[rough oxide]

so i can't speak for nocturne

[silk steppe]

if you wanna fall asleep

you should try it

[rough oxide]

smt v has a lot of strategic depth/difficulty tho. everyone i tried to get into the game

like 70% couldn't get past first major boss

the lv 12 one

and other 30% couldn't pass first story arc

the final lv 20 boss of the first island

lol

[silk steppe]

depth don't exactly mean dificulty, so has persona, you can turbo min max but sadly game gives you no reason to

[rough oxide]

i can't get ppl into smt v because it's too difficult

[silk steppe]

boring too

[rough oxide]

not boring pfff

i played on hard and every encounter u have to think about

[silk steppe]

exp 33 fucks the balance too

[rough oxide]

even the random mobs

[silk steppe]

act 1 and 2? perfect

act 3 omg

balance goes out the window

you're doing millions of dmg

it's hard to NOT 1 shot

idk what the devs were smoking

on act 3

but ya know

[woven gust]

kino no

tabi!

[silk steppe]

hope neither of these is too much, if it is: my bad

[silk steppe]

Foinal foitosi

[rough bay]

hello, im new, i have questions can i use this for my training image? i made the character big to make sure the main focus of the AI when training is the character not the background - should i make him smaller ? also my character is kinda blurry should i upscale him more or is this enough ? also should I crop him better ? for stable diffusion 1.5

[rough oxide]

@autumn roost #🍥｜anime message phishing link

(er when i click that it overshoots, it's above these images tho)

[gleaming flame]

"man's dream" or "iron maiden" works better as a title?

[gleaming flame]

i'm not an artist... i wouldn't mind...

i just make stuff

[dusky linden]

I wouldn't call myself an artist either lol

[gleaming flame]

even though I make music (not AI), I still wouldn't call it art... not yet at least 😛

[dusky linden]

Unless you whip out photoshop, edit the pictures etc i wouldn't call someone an artist

[wooden sail]

It's ok when you don't want to call yourself an artist. It's bad when other people decide that for you

[woven gust]

@wooden sail kino no tabi

[sharp imp]

Eh, I'm a prompt engineer. Just so happens that the AI makes art with my prompting. But I'm not the artist. It's a fair distinction. Cause while I put time and effort into my craft, it's not the same as a traditional artist does. So I believe it should have a different label.

[woven gust]

Good

[cold talon]

nice

[cold talon]

left is probably a bit much, right is good 👌

[silk steppe]

Didn't go as planned but they on cod

[silk steppe]

i have too much fun with these

for dbz fans

i can't edit for shit, it was so doomed i didn't even bother to upscale

[keen hollow]

evenin'

[wooden sail]

From here:
https://huggingface.co/spaces/ByteDance/DreamO

[wooden sail]

Left one

[near elbow]

succubus fun (SFWish enough i guess)

[sharp imp]

Id call that safe for work. A little cleavage is nothing nowadays.

[near elbow]

well i had to add "tank top", the version before was much more in your face 😉

[sharp imp]

AI do be like that. I actually made a song about it with Suno

[sharp imp]

hey even though it was made with suno, i can post it in Stable Audio right?

[tawny wasp]

work in progress 🙂

[keen hollow]

ultrawide is hard, there's always a jacked hand or something since you gotta upscale and rerun

goodnight

[near elbow]

anime-ified 🙂

[woven gust]

please teach me how to create such colorful images

similar images

😭 😭 😭

paimonpain

cryingkitten

[sharp imp]

Proper model + creative prompt and settings is all i use. That in particular was made with a wildcard pack I made.

Oh and a couple loras that fit.

[woven gust]

do you want to share them?

[steady grail]

usually, brilliant colors comes from pony and noob

[wooden sail]

"Put the corn down, lady"

[sharp imp]

I will, once I get home from work.

[near elbow]

use a good model (e.g. flux dev here) and prompt away... "anime style, girl sitting in darkness cowering legs up head down, black dress, eclipse behind her, flowing golden hair, exquisite intricate fractal"

[woven gust]

can you give me the model?

this looks pretty too

I think my models are bad

maybe I'm just bad at prompting

[near elbow]

personally, i wouldn't go with anything below flux or hidream (or maybe sd3). the older models just don't get it 🙂
this specifically is flux wigh ghibsky lora on replicate, less than a cent per image
https://replicate.com/aleksa-codes/flux-ghibsky-illustration

i usually prefer flux pro 1.1 there, really great results, and locally i run flux dev and hidream (fast, dev and full, but 90s for one image with full is cruel)

and these are not anime specific models, i never got into booru prompting and i don't like the pony hidden artist names and other intricacies. the same model can produce perfect photo quality as well, and i'm not that much into super specific anime figures or kinks 🙂

i might be a minority with that in this channel here though 😇

for fun, this is "anime style, girl sitting in darkness cowering legs up head down, black dress, eclipse behind her, flowing golden hair, exquisite intricate fractal, vibrant colors, contrast between dark and light" with flux pro 1.1

[silk steppe]

flux is indeed good, quite more hardware dependant tho

so i jsut stick to illu

[near elbow]

well, 4090 here, that spoils you quickly 🙂
i do recommend replicate for flux 1.1, really fast and 3ct per image, and the ghibsky is less than .3 ct... since the hit-to-miss ratio with flux is great, i just generate one image and it's already good enough, so my overall replicate costs per month are less than $10 usually. but i'm mostly using that mobile or when the rendering rig is sleeping

only downside besides the cost is that it's censored with replicate and many other web services (mage.space used to have many models and a paid NSFW-is-ok tier though, haven't used that for a while)

[dusky linden]

I like using high contrast

[woven gust]

you have to pay for flux? I can't run the same on my PC?

[near elbow]

if you have a beefy graphics card, you can run flux and hidream on your PC.
paid web services are just another way, trading convenience for a bit of money.
i do both. google for "run flux locally" 😉

[dusky linden]

Flux pro is cloud only iirc

[near elbow]

right, but dev is also quite good (and fusion and all derivates)

[dusky linden]

Yup its pretty neat but i prefer it for realism as illustrious beats it for anime style imo

[near elbow]

yeah, a dedicated anime model has its merits, especially if you want special characters or the booru tagging

[dusky linden]

Hmm illustrious v2 supports some Natural tagging and some booru tags

Illustriousv3 once it comes out is gonna be fully anime style but natural prompting

[near elbow]

is that sdxl based? good luck with character separation or good prompt understanding then 😕

SD3 would be fine i guess

[dusky linden]

Yep sdxl base deep down

It already struggles somewhat

[near elbow]

just for fun "miku next to groku next to totoro next to a pokemon" with flux pro
try that with SDXL and start crying 😉

[dusky linden]

It will give the miku haircut to tototo or the tototo build to miku

And ignore the male lma9

Still it works fine for me and the 5-8s gens are pretty nice

[near elbow]

yeah, and for solo characters it doesn't matter so much

[dusky linden]

Sd3.5 large is hella nice for styles

But 1min gens

Q8 gives me 30s gens (what i had on my old card with sdxl)

[near elbow]

i gave up on SD 3.x when flux came out 😇
flux fusion on the 4090 goes below 5s

[dusky linden]

Hmm thats schnell right?

Or schnell based

I get those times too lol

[near elbow]

it's schnell and dev mixed with some clever magic, and results are much better than dev

[dusky linden]

Dev is like 40s

16gb vram

[near elbow]

even with the 4090, hidream fast fp8 is 30s, hidream dev >45, hidream full >90s 😕

teaches you patience again

[dusky linden]

Oh dev gets me 50s ish with a 5080

Full is okay but gets a lot of jpeg artifacts

[dusky linden]

Biggest reason why i am still on xl lmao

[misty violet]

This is stunning work I can tell the prompting here was super intentional! The color harmony, from the deep space blues to the gold accents, shows a great eye for theme. Getting that heterochromia, the glossy outfit textures, and the perfect idol pose with both hands doing peace signs? That’s some precise prompting magic 👏

Curious how many iterations it took to lock in that balance of cute and confident!

[keen hollow]

I wish I could say I'm an amazing prompter, but alas, I just grabbed a bunch of random tags and threw them together until something coherent came out. 😄

More examples

[woven gust]

okay ty, I have stability matrix

[silk steppe]

gotta say

daz a lotta animu

[sharp imp]

For this image (and others like it you've seen) I used an Illustrious model with locally run Stable Diffusion. The specific checkpoint is EvenlyMix_v10. Its on Civitai. As for the prompt, beings that it was made with wildcards it probably wont make sense, but thats half the fun of wildcards. This was the prompt:

<lora:StS-Illustrious-Detail-Slider-v1.0:3>,<lora:Dramatic Lighting Slider:3>,```

The only loras i used are Illustrious Detail Slider and Illustrious Dramatic Lighting Slider. Both also available on Civitai. As for settings, i just used the recommended settings from the checkpoint creator. I do vary sometimes, for example, with more abstract images ill use DDIM from time to time instead of DPM++ 2M/Euler A because DDIM renders some cool stuff with abstract prompts. But thats about it, nothing fancy, zero inpainting or post processing.

[sharp imp]

this is the only drawback with SDXL, but the new Illustrious 2.0 base does eventually get it right...sometimes lol

Daily dose of wildcard created cuteness

[woven gust]

Eclipse

[near elbow]

"anime cute waifu, solar eclipse background, backlit, hard contrast outlines, enticing" with flux pro

[woven gust]

I love the colors

please do more

its so cute

aaaaaaa

[sharp imp]

haha i have another 20 cooking right now. ill share the best/most sfw ones. These wildcards get well...wild sometimes.

[woven gust]

just dump all in here

its fine

[sharp imp]

lol, no trust me...its not haha

this one is risky lol latex kinda goes crazy at times

Thats about all i can share from that batch haha

This is all from a mega prompt i made (that i need to add to now since ive made more stuff) that randomly selects from several wildcard packs i made.

Each pack has a different theme/emotion. Ive got Eclipseborne, Bloom//Decay, Corrupted Goddess, Neural Collapse, Paint the World, Divine Temptation, Omega Kawaii, Primordial Wardens, and Mythomech. And then some other that are less colorful that were test projects to learn how to make recursive nested wildcard prompts.

[sharp imp]

Awesome as always @keen hollow is the last one a character from somewhere?

[keen hollow]

Literally random tags... And I keep forgeting to save them off somewhere in case people wanna see em

[sharp imp]

hahaha yeah i have my forge set up to save a copy of any prompt that makes an image lol just in case

[keen hollow]

i've been messing around with rented gpus and it destroys the instance when you're done

when i dump images here, I just use copy/paste which doesn't retain the meta

spinning up another one

[woven gust]

pretty dress

[woven gust]

that's very cool

I want to create images now too

but

my pc is turned off

and I'm at work

[misty violet]

yo this one's got such a clean vibe… lighting’s soft, pose feels natural, and the expression's got that calm confidence. honestly feels like a frame straight out of a high-budget anime. you nailed it fr 👏

[misty violet]

yo the prompting on this is crazy good… like every little detail just hits. the pose, the lighting, the outfit it’s all perfectly balanced between sweet and lowkey seductive. that sleepy expression with the kitten?? too smooth. whoever cooked this up knew exactly what they were doing🔥

[sharp imp]

well it was made completely at random lol. all i did was put tags and phrases in a neat little .txt file (or 5 in this case) and let the model pick at random from them to make the image.

Heyo i guess this model can do males when it wants. Didnt expect that.

[radiant oasis]

(masterpiece, best quality, official art, genshin impact style),
1girl, beautiful detailed face, long silver hair, blue eyes,
elegant white and blue knight armor, glowing sword,
dynamic pose, fantasy background, cel-shading, vibrant colors

[tawny wasp]

hi,which prompt did you use?

[sharp imp]

It was made with a wildcard. Once I'm off work I can look but it was randomly generated.

[tawny wasp]

really? it came out beautiful 🙂 . what program did you use?

[sharp imp]

Stable Diffusion with ForgeUI. Nothing special. I mentioned the specific checkpoint and loras earlier above that image. But the prompt is a wildcard prompt that you need to install the dynamic prompts extension to enable.

[sharp imp]

Also thank you! I spent a good chunk of time making my wildcard files, I'm glad people enjoy the creativity they can produce .

[tawny wasp]

Loona Super Saiyan 3 🙂

[keen hollow]

I need to fart around with a different model.

[keen hollow]

`1girl,

black hair, furry, white fur, multicolored fur, blush, fangs, cape, fur-trimmed sleeves, raccoon ears, sweat, black cape, black gloves, animal ear fluff, embarrassed, raccoon tail, yellow eyes, fur collar, cropped torso, claw pose, upper body, raccoon girl, purple bow, brown fur, white gloves, sidelocks, purple bowtie, high ponytail, claws, gao, furry female, hair bow, white hair, flying sweatdrops, medium hair, fur trim, solo, bowtie, white shirt, white mittens, bow, multicolored hair, smile, fur-trimmed gloves, two-tone hair, breasts, open mouth, :d, happy, white background, simple background, body fur, tail, gloves, index fingers together, nose blush, grey hair, ;d, long sleeves, animal ears, shirt, nervous smile, mittens, looking down, hands up, hair between eyes, striped tail, one eye closed,

masterpiece, newest, absurdres, safe`

someone was posting weird racoon people the other day, so ... here's a random one that came up

and a fish lady...

"monster mermich" is my new favorite term

`1girl,

monsterification, hat feather, cape, brown hair, hat ribbon, character name, star hair ornament, blue eyes, tate eboshi, feather hair ornament, fins, blue footwear, mermaid, blue cape, ribbon, monster girl, long hair, star (symbol), chibi, solo, english text, virtual youtuber, open mouth, wavy hair, simple background, red skirt, blue hat, personification, very long hair, looking at viewer, hair ornament, shirt, hat, feathers, skirt, blue shirt, blue ribbon, hair between eyes, white background,

masterpiece, newest, absurdres, safe`

[sharp imp]

the spear got wonky

[keen hollow]

evenin'

[near elbow]

"anime style weird raccoon girl" with flux pro 🙂

"flat 2d cel shaded studio ghibli anime style weird raccoon girl"

[woven gust]

what was the website name again for stable diffusion models

[dusky linden]

CivitAI

Huggingface also has some but civitAI is better

For the average user**

[woven gust]

thank you I forgot about this

[woven gust]

its not perfect

but its lookng cool

[hidden lotus]

swarmUI...I really like the way it looks

[misty violet]

This looks incredible the colors, lighting, and intensity are spot on. The prompt must’ve been super well thought out!

[woven gust]

Its like a six word prompt

try 'black background' and 'oshi no ko artstyle'

you will be surprised

make sure to share results

[silk steppe]

When no one shows up for dnd

[silk steppe]

NO, BIDEN! DAMN THE PAINTRESS. For all my expedition 33 fans

[tawny wasp]

Nico Wakatsuki - Witch Watch Illustrious Apex_Ai

[tawny wasp]

Beautiful,prompt? 🙂

[sharp imp]

Ok, but prepared for it not to make sense lol.

kitsune girl, fox ears, multiple fluffy tails, illusion shimmer, playful golden eyes,garb made from creation myths given form,
stance of creation with reality forming from hands, omniscient angle,dagger forged from shadow and silence,
moonlit glacier peak,aura sigil floating behind in a magic ring,glowing aura,elemental energy effects,intricate detailing,dynamic lighting,ambient fog,cinematic depth of field,mythic power,fierce elegance,swirling particles,magical residue,glowing highlights,<lora:Dramatic Lighting Slider:3>,<lora:StS-Illustrious-Detail-Slider-v1.0:3>,<lora:TrendCraft_The_Peoples_Style_Detailer-v1.0A-3_13_2025-illustrious:0.8>,```

playing with a new checkpoint. i like the darkness in it

[tawny wasp]

I see many of them that I don't know, and I don't have XD

[sharp imp]

just gotta download, plug in some prompts, see if you like the output. if not, delete it and move on to the next. ive got about 7 i rotate through regularly.

but im always testing new checkpoints that people post. and before civit changed its gen rules and stuff i was sharing more on there to give the checkpoint creators some love. kinda been slacking on that since they had their lil change of policy stuff.

[tawny wasp]

I knew they changed something. But what did they change in general?

I was able to find only 2 hours in the prompts you gave me, the rest I used the ones I already had. But the final result, I like it 🙂

[sharp imp]

it turned out pretty good! its always fun to see how other checkpoints interpret a prompt my wildcard packs make

[tawny wasp]

🙂

Which checkpoint did you use for your photos at the moment?

[sharp imp]

https://civitai.com/models/1552304/singlo-illustrious?modelVersionId=1782085

[sharp imp]

ear got a lil funky, but this one turned out cool

[woven gust]

nice lighting

what did you use for that?

for the lighting

[sharp imp]

Happy little accident, cause this was the prompt:

(masterpiece, best quality, newest, absurdres, highres,very aesthetic, very awa,ultra detailed),1girl,solo,hysterical sobbing ,cracked enamel armor inlays ,ribbons of light weaving around body ,jet black,Twists / Twist Out,amber eyes,middle w ,glitch-patched jacket ,tesseract bloom,tarnished royal seal pin ,mid-cast stance with arms wide, camera panning from side,looking at viewer,

But according to ChatGPT, this should get a similar effect:
Lighting: cool blue rim lighting, intense speculars

masterpiece,best quality,newest,absurdres,highres,amazing quality,very aesthetic,very awa,ultra-detailed,1girl,solo,cyberpunk,armored_chest,black_jacket,peace_sign,glowing_eyes,orange_eyes,long_hair,braid,cyber_camera,futuristic_camera,dark_background,blue_lighting,rim_lighting,digital_particles,high_detail,anime_realism,cinematic,masterpiece

kinda works, adding cyberpunk to a prompt changed the background a bit, which changed the end result, if i had kept solid black background it probably would have looked closer. this is why i enjoy wildcards though, you get some totally unexpected outputs

[dusky linden]

Just be sure to remove any and all _ because it just increases token count

[sharp imp]

true, i hadnt thought of that being counted lol

[dusky linden]

Its insane its like a token per word sometimes

Sometimes two

[sharp imp]

yeah removing the underscores dropped it from 150 to 119 lol

but still, 150 is two separate 75 token chunks which isnt terrible. just depends on whats in each chunk lol

same prompt, no underscores

[dusky linden]

Could improve it by adding a break somewhere so it doesn't split up a concept

[sharp imp]

Yeah so long and short of it @woven gust , "volumetric lighting, rim lighting, blue rim lighting, " any of those will match that effect, especially if your using a lighting LoRA like i am to make it really pop

[dusky linden]

Volumetric lighting is a hit or miss for me

[woven gust]

thank you for your fast response. I will try that. I dont have any loras for lighting. Only limbus

[sharp imp]

Ive had it work better when paired with another lighting tag as well, like volumetric + god rays etc.

[dusky linden]

Hmm probably, on my models it often adds a yellow light or red light and im not always a fan of that

[sharp imp]

huh, accordiing to google volumetric is synonymous with god rays. so my guess would be add one or both of those, and a lighting color and direction/origin, if you want to increase the likelihood it renders. it also does depend on the checkpoint and which tag was used to label training data, i assume for a lot of anime style ones it was probably god rays more than likely

[stone onyx]

What model did you use for this, it looks realllly nice
(png info would be cool :D)

[sharp imp]

At work at the moment so I can't check, but the model was the newest plant milk checkpoint. I used dynamic lighting slider and illustrious details slider loras as well.

[tawny wasp]

My new model is here: Blair 🙂

[near elbow]

mecha kitsune 🙂

[near elbow]

the redundant ears are a minor issue 😉

[wooden sail]

It's cold outside 🥶

[tawny wasp]

do you mean his 4 ears? I did it just like that, it's not a mistake

because she is an alien cat girl

[near elbow]

it's really hard to get TWO pairs of ear warmers. all i manage is furry ears...
challenge accepted?

[tawny wasp]

mmmmmmm

you try it, but in my opinion, a lot of mistakes will come out

[near elbow]

that's the challenging part 😉
i tried and all it did was inner-fur ears...

inpainting should work though

[near elbow]

cheers

[elfin valley]

Hi which model currently creates the best anime photos in this style?

[wooden sail]

arms at sides

[keen hollow]

Many of my more recent posts (including those) are with posILustrij.

[woven gust]

kino no tabi

[sharp imp]

There are a few. Hassaku has some solid ones. Illustrij( or any of reij models) do pretty good, though some tend toward more semirealism. Can't go wrong with Wainsfwillustrious.

[tawny wasp]

Amazing

[tawny wasp]

Beautiful

[wooden sail]

Ima jumping out of here

[tawny wasp]

prompt?

[tawny wasp]

prompt?

[sharp imp]

lol try to guess.

[tawny wasp]

unfortunately my crystal ball doesn't work XDDD

[sharp imp]

copy it, put it into an interrogator or chatgpt and see what it says, itll probably be closer to the actual image than my wildcard prompt lol

since my wildcard is pretty random

[tawny wasp]

so i have to ask chatgpt what prompts are in the picture? i use chatgpt very little

[sharp imp]

you should ask it what prompt likely made the image, then ask it to give you 3 prompts that are 10x better in your preferred prompt format

[tawny wasp]

ok,i'll try

[sharp imp]

i did it with this image and it gave me this prompt:

masterpiece, ultra-detailed, very aesthetic, 1girl, solo, medium shot, futuristic anime goddess in liquid black vinyl bodysuit with reflective specular lighting, short brown bob cut with dynamic rim lighting, rainbow-colored dimensional aurora spiraling behind her, glowing miniature planets orbiting her chest, cosmic particle effects trailing behind her fingers, ‘rock on’ hand sign raised, nebula haze and prismatic light bursts in background, rich cityscape lights twinkling below, cinematic contrast, glossy skin reflections, space-pop elegance, sharp warm/cool split lighting, cosmic glow

which made this image with my loras etc

It doesnt copy the image exactly, which is better cause then its something uniquely yours!

I've even got a custom system prompt that gets chat gpt to make pretty decent prompts for you, and all you gotta tell it is a few tags to give it an idea of what you want (i usually give it 3 so that it has a subject, theme/artstyle, and a composition type; either cinematic or dynamic, which guides its choice of camera and pose a bit more)

Of course the more you give it, the better the prompt it will make. But its meant to be something to help when you have a creative block, otherwise if your giving it a full prompt yourself, youd be better off using the first method of sharing an image and asking it to improve it

[tawny wasp]

but in your opinion among the free ai programs, which is the best for creating images? and which for videos?

I would like to create something beautiful, and I would like it to be simple enough. Although, I would like to try something complex

forge? never heard of it honestly. could you send me the link where I can download it?

[sharp imp]

i use forge as well, its a newer version of A1111. Locally hosted Stable Diffusion. I dont bother with site generators cause even when using the same checkpoints i feel like i dont have as much control. Plus im not about to pay for it. That said, from what i understand Swarm and ComfyUI give a lot more control and are better once you learn to use them, theres just a bit more barrier to entry.

[tawny wasp]

ok,thx a lot 🙂

[tawny wasp]

I have always used only stable diffusion. I tried using comfyui this week, but honestly I didn't like what it created. How can I create amazing photos? What should I set to improve?

[sharp imp]

Its an iterative process. Your not gonna get what you want at first, and i spent the better part of 2 and 1/2 months just playing around with it til i understood what i was doing. Then i spent another month just testing checkpoints and concepts til i found a style i liked. Theres nothing thats gonna automatically make great stuff right away, theres a small learning curve.

I go back and look at my earlier stuff and its a world of difference.

As for comfyui, some people may be willing, or have already, shared their workflows on sites like Civitai or Huggingface, and that can help you get started. But unless you take time to understand WHY they did what they did to make it, you'll still feel like something is missing. Honestly, id ask chatgpt to teach you how to use it. Thats what i did for A1111

A1111 is just so slow, i had to upgrade. And i will probably make the move to comfy before long.

"Their after the cook recipe again!"

[tawny wasp]

so between forge and comfyui, which is better? so i can see which one to download and which one not 🙂

[sharp imp]

depends on how technical you are. comfy uses nodes in a workflow, which if your familiar with automation will be easier to understand. forge is just an optimized version of A1111. Its got built in extensions and is generally faster and easier on your GPU. Both are good.

[tawny wasp]

ah ok,thx a lot 🙂

[tawny wasp]

guys, I uploaded the folder with all the photos to sd to insert in all the prompts. But how do you do it on forge??? I downloaded it and installed it now

[sharp imp]

that sounds like a tech support question, cause i didnt upload any photos to forge. i did copy my stable diffusion models over, but thats files not photos.

[tawny wasp]

I meant, when I want to create a Lora, I created a folder with the photos of the character I wanted to create. Then I uploaded the folder to stable diffusion, and by doing so I inserted all the prompts in all the photos. and then I went to google collab to create the Lora

Dataset Tag Editor is the name

[sharp imp]

OHHHHH

That extension and its associated folders may not be baked into forge, but should be available to download. I know Checkpoint Merger is baked in but i dunno about training for LoRAs. Thats definitely out of my wheelhouse, i just make pretty pictures, i aint smart enough to train stuff

[near elbow]

no need for any of that with more modern models IMHO 😉
most LORAs are training wheels for SDXL models which can't do things that recent models like Flux or HiDream just get out of the box.
plus of course specialized application for specific characters or... poses and more specialized stuff

[dusky linden]

I have to say i wasn't very impressed with hi dream but i ran it on day 1 so maybe i was prompting wrong

[near elbow]

just for fun with hidream "semireal detailed 3d anime style, exquisite intricate, robed waifu standing in clouds, polychromatic aura around her head, delicate threads of reality, she's wearing a dark robe with red embroidery, cleavage, black shiny eye mask with an opalescent sphere on her forehead, holding an energy orb in her hands before her chest, smiling, showing legs, red hair in braids" (very first try, needs refinement, just as a POC)

another take

[sharp imp]

Flux and HiDream are good, but because their made to be so versatile theres a lot of style blend that occurs when prompting unless you get extremely specific about things like lineart, shading, etc. I will say their prompt adherence is really good though.

Now once we can tinker with and fine tune/merge flux and hi dream models, i expect they will outperform everything else. But out of the box their versatility means they fall short sometimes of the quality that Illustrious can provide, at least to me, specifically for anime. For realism and more abstract stuff, Flux has been crazy good. I havent used hi dream yet.

took off my lighting enhancer lora lol this is just the checkpoint and the prompt

[near elbow]

yeah, OOTB details with hidream could be better... increasing to "vivid dynamic background and style, highly detailed, semireal detailed 3d anime style, exquisite intricate, robed waifu standing in clouds, polychromatic aura around her head, delicate threads of reality, she's wearing a dark robe with red embroidery, cleavage, black shiny eye mask with an opalescent sphere on her forehead, holding an energy orb in her hands before her chest, smiling, showing legs, red hair floating, attention to detail, dynamic motion, action, background gradients and details, glowing reflections volumetric lighting" 🙂
still quite clean and symmetrical, and less dynamic

[sharp imp]

front row seat to the end of the universe

[near elbow]

same prompt with chroma 29.5 btw, definitely more anime-ish and more details

[sharp imp]

WAY more details lol

tried it with my new merge. this one understands natural language prompts better.

[near elbow]

chroma with "friendly necromancer waifu, night, summoning evil spirits, sardonic smirk" 🙂
nice mood, definitely evil fingers

[sharp imp]

Lol

Once I get home I'll remix that and see what I get. I'm at a cookout right now

[near elbow]

enjoy - here's a necromancer cookout for inspiration 🙂

[glacial coral]

Agreed

RayFlux 3

Been my goat

[wooden sail]

Day/night sometimes works with Forge Couple

[river shale]

OOOO Necromancer~ I'll make my own renditon too

[river shale]

This is what your code gave me :3

[sharp imp]

mine also had purple fire til i added "green fire, green spirits" to it lol still didnt create the spirits but i was feeling lazy so i didnt push it

masterpiece,best quality,amazing quality,absurdres,ultra-detailed,high res,very aesthetic,very awa,8k,1girl,cyber_ninja,drenched_clothes,glowing_katana,wet_environment,intense_stare,red_scarf,electric_rain,pink_lightning,magenta_hair,cinematic_lighting,battle_damage,glowing_water,low_angle,techno_warrior,post_apocalyptic_city,stormy_background,ultra_detailed

[river shale]

My own rendition

[sharp imp]

Or maybe you prefer something cute, in which case...
1girl,long_hair,brown_hair,hugging,bunny_plush,galaxy_outfit,sparkly_boots,frilly_skirt,blushing,soft_lighting,inside_vehicle,snowing_outside,cozy,winter_scene,pastel_palette,dreamy_atmosphere,ultra_detailed

[terse ore]

How do you make a pose similar to a backbend, or when a pole vaulter is at its highest point and sort of arches their back to get over the bar? I tried controlnet and it doesn't get the right pose. Hard to do those gymnastics type poses

[sharp imp]

mmmm thats a tough one. poses are always interesting, cause if it isnt also the right angle for the camera angle, it still looks wrong

[river shale]

poses are hard tbh

Angles are important too, from above, from below, side view, dutch angle.

[river shale]

Also things like

• hands behind head
• Hands on hips
• Squatting, sitting, standing, kneeling
• full body
• legs presenting

Also help with poses

[terse ore]

I did try prompts like backbend, hands behind, legs behind, etc. I'll look for lora

[river shale]

Curious what kind of pose you're looking to do; id like to try and make it, see if I can

[sharp imp]

just look up polevaulting images, its the position of the body when they are bending to clear the pole at the height of their jump

[river shale]

yeah good luck 😂 I tried and it hates me

[sharp imp]

lol upside down poses and contortionist type poses are tough, and most models arent trained on it so they fuck it up

[terse ore]

I wonder if I did a basic drawing/3d character pose and then img2img it with controlnet 🤔

[sharp imp]

that might work. its still gonna require multiple passes to get it right

[terse ore]

I tried before, it often makes the hands into feet and vice versa because that's where it expects it to be

[sharp imp]

yeah, its an odd pose that most models probably havent see much

[river shale]

A French baguette

[sharp imp]

is that a baguette wearing a french mime outfit

[river shale]

🤭

yes, yes it is

🤭

[sharp imp]

i cant get it to gen the FUCKING BAGUUETTEEE

[sharp imp]

SUCCESS

one second

apparently french mime and baguette this model has not been trained much on, and it fought me every step of the way on it

i had french generals, napoleon, and at once point a mime wielding two sabers, but it wouldnt do what i wanted for like 6 gens

[river shale]

I just said: Image of a baguette with a striped shirt, French hat and mustache

but then again I used Grok

[sharp imp]

1girl,(mime:1.3),mime makeup,French outfit,dual wielding,wielding bread,(wielding baguettes:1.5),dynamic pose,cinematic lighting,playful,elegant,black and white stripes,beret,detailed background,theatrical,long hair,intense expression,action,looking at viewer,

[river shale]

im gonna put that in grok see what it says

[sharp imp]

yeah im using my personal merge checkpoint on stable diffusion lol

[river shale]

your code in grok vs stable diffusion

BEHOLD THE GOLDEN BAGUETTE

[sharp imp]

its so majestic

[keen hollow]

evenin'

[river shale]

Do you create your codes on your own?

Good Evening OmniaBurger 🍔

[sharp imp]

you mean my prompts? yeah usually. Some are from wildcard files ive made, some are freely written. Some are made by chatgpt, but i still give it a prompt to start with and it enhances it

[river shale]

yeah prompts 🤭 huh nice 🤔

I use chatgpt for backgrounds for my codes somtimes; you just come up with very different gens every time and I like that 🙂 I suck at it 😂

[sharp imp]

I have a system prompt for use with LLMs and stuff i made that causes them to take an image or a simple prompt, and refine or enhance it into a better prompt

[river shale]

o0o

fancy

[sharp imp]

i can share it if you like. it isnt a jailbreak so it wont get you banned

[river shale]

🤔

Like what do you mean, share what

[sharp imp]

it should work with grok, i havent tried it yet but i dont imagine it would fail to work. though in the past grok and gemini tried generating images with my partial or simple prompt when i didnt want them to, so i use primarily chatgpt

[river shale]

https://tenor.com/view/rachel-green-jennifer-anniston-woosh-over-my-head-over-your-head-gif-15405360

oh ok you mean what you said to the AI to get it to make you the things?

[sharp imp]

copy that and start a new convo with grok or whatever llm and paste it in as your first prompt to them, they should reply with "Understood, what concept would you like to explore first?" or something similar

after that, you just give it a simple prompt

and it should make it better.

[sharp imp]

like "1girl, anime, elf, fantasy world."

[river shale]

fancy

ok ima see how it works

ooo it made a prompt

[keen hollow]

colorful, or dark errie...

not sure which I prefer

[sharp imp]

for that character, colorful lol

[sharp imp]

yeah i plugged it into grok and it immediately used the example prompt in the instructions and made a prompt as its reply

[river shale]

n_n yaay thank you

[sharp imp]

after that it should do the same thing with any prompt or image you share

[river shale]

👀

[sharp imp]

and then you can plug em into SD and see what you get

[sharp imp]

this was made with the prompt it gave me lol

[river shale]

oh youres is making a vamire girl too 🤭

[river shale]

This is so good

[sharp imp]

yeah the vampire girl was in the example in the instructions lol

[river shale]

OH i didnt read the whole thing 🤡

I just said "create a new Natural Language Prompt" and it gave me a whole new one, i think ill just use it like that n_n

[sharp imp]

lol i mean its for the model not for you, so its fine if you didnt....even though i put a lot of work into it

[river shale]

It shows 🙂 It was very thorough

Well cool now I can use these to change them a bit and use them for my OC >:D

You said I can send a image and have it make codes too?

[sharp imp]

Yep! i shared the image @keen hollow made just now and it gave me this:

masterpiece, best quality, solo, 1girl, mage, long_hair, brown_hair, braid, green_eyes, glowing_eyes, fair_skin, red_jacket, leather_jacket, white_skirt, pink_trim, combat_boots, metal_buckles, staff, glowing_staff, emerald_crystal, summoning_circle, forest, moonlit, ancient_trees, bioluminescent_moss, floating_wisps, low_angle, cinematic, volumetric_moonlight, glowing_particles, chromatic_aberration, semi-realistic, glossy, dynamic_pose, mystical, defiant, awe-inspiring

i added the appropriate character name to it, and it spat out this image when i used it:

[river shale]

noice! Just as nice as his 🙂

[keen hollow]

yesssss

more Aerith

[sharp imp]

of course, the differences in our model of choice for making the image is the biggest reason for their different aesthetic, but if you plug images in it should consistently give you a prompt that is similar, but better.

[keen hollow]

I hate when vader sneaks up behind you...

[river shale]

makes me think of fortnite

crazy you got them both in the same image

[sharp imp]

with illustrious v2.0 models you can get it, though its not PERFECTLY consistent yet

[keen hollow]

I am definitely NOT the person to ask about multi-character generations. I am terrible at it.

[sharp imp]

a dash of tifa for flavor

[keen hollow]

yessss

[river shale]

I tried makin Aerith and Yuna and it didnt understand 😂

[sharp imp]

close enough

[river shale]

also tried to make cloud and sephiroth 🤡

[sharp imp]

lol

motherfucker running around with frostmourne and shit

"did i do good?" ~smile

[river shale]

LOL WOW

I give up on the final fantasy stuff I can't get stable to do it right 😂

🤡

[keen hollow]

you got this

[river shale]

WHELL

I'll try again later 😂

Ima give my hubs the PC back for now

[sharp imp]

huh?

[river shale]

I'll be back on maybe tomorrow

[sharp imp]

theres a limit for you?

[river shale]

Grok has a limit- yeah

[sharp imp]

OH your on a free plan i take it

[river shale]

mhm

[sharp imp]

that explains it

[river shale]

ok be back tonight or tmorrow

its been nice

yall have a good evening/morning

[sharp imp]

you too!

[river shale]

oh and thanks for the help today ret

[sharp imp]

no problem, glad i could actually help lol

[sharp imp]

sure you dont

[rough oxide]

tifa has so many uniformss

[keen hollow]

try her honeybee one 😛

[rough oxide]

full danbooru name?

oic

it works but

i feel like i'd have to dm this one xD

or maybe i can use an anime style to tone it down or something

nyo i'll just send u a pic xD

this one is probably safe enough xD

[keen hollow]

haha yep that's it

[rough oxide]

ssi is best harley

that anime was fun

seemed to have low ratings tho

[near elbow]

playing with chroma, but that has too many issues for my taste
"harley quinn anime style sexy evil smirk view from below wielding baseball bat"

"semireal 3d style" instead

[dusky linden]

What anime is this from

Watching it rn

Wtf kirito buying meth

[near elbow]

now you're trying to taint the reputation of Walter, right? 🤪

he gets away with anything anyway

[glacial coral]

i kinda lost the link for SIH model could i get a civit link from some1 please

[lyric zealot]

https://civitai.com/models/1217645?modelVersionId=1371790

[lyric zealot]

real answer: ||it's from a game called "Clair Obscur: Expedition 33"||

omg, never mind

you replied to the top image and not the bottom

[glacial coral]

Ty my good friend hope ur doing well btw I’ve improved my outputs with the help of u lot

[keen hollow]

Name that fandom.

[rough oxide]

The character on the right appears in the anime series "Death Note", however it's a one-shot manga spinoff and he doesn't appear in the main series.

Pretty good 3d representation of that character though who just appeared for a moment in a few manga panels.

You can read it here: https://www.viz.com/shonenjump/death-note-special-one-shot-chapter-1/chapter/19985

need a sub tho

as for the left character, can't say I've seen him in anime.

might just be generic bald guy.

[tawny wasp]

Fie Claussell From (Trails of Cold Steel) Illustrious Apex_Ai
https://civitai.com/models/1597778

[woven gust]

kino...

no

tabi

[river shale]

Made my OC but with a mohawk and different clothing style

yes yes you inspired it ret 🤭

[sharp imp]

😄

[river shale]

Im dinking around with making her in differnt themes or whatever you wanna call them, aesthetics? whatever

[sharp imp]

i merged my previous model with Illus Stable v 2.0. It has had....interesting results. Its subdued some of the linework and softened the contrast and colors a bit. But it also has a sort of...cinematic feel

[sharp imp]

it recognizes different art styles 😄

[river shale]

eehehe chibi :3

[sharp imp]

im enjoying this, most checkpoint merges dont handle artist or art style tags unless a lora trained for it is merged with them.

[keen hollow]

evenin'

[sharp imp]

Lol evening indeed

[keen hollow]

oh my gosh

this animation fail

im dying

[lethal rain]

bilibili's open-sourced this anime video gen model called Anisora. The V2 version, which runs on wan2.1, kills it for anime videos—like, it's really good. But right now, it doesn't work with ComfyUI. I'm really hoping someone can make it compatible!
thanks!!
GitHub address:https://github.com/bilibili/Index-anisora

[lyric zealot]

@finite herald

@distant socket did you delete your message?

[dusky linden]

It just needs to release safetensors or gguf before anyone bothers

[sharp imp]

OC character in a P5 inspired style.

[wooden sail]

https://tenor.com/view/howard-stark-tony-stark-the-avengers-avengers-science-gif-6569892173967199721

[near elbow]

just generate at 1024x1024 natively or higher and forget about upscaling? 🙂

bad model 😉

with flux it's usually enough to add "beautiful detailed face"

if that doesn't do the job, i hit "generate" again - too lazy for postprocessing 🙂

strange

voodoo and black magic 🙂

[sharp imp]

Once I get home i can pull the exact prompt, but nah if there was resemblance it was unintentional lol

Bet. But also, I gen at 1024x1024 and upscale to 3072x3072 for funsies, cause I like to see my GPU sweat.

Someone put me on to 1536x567 and now it's all i do for landscape wide shots lol

[near elbow]

3k 🙂

[twilit portal]

🫧

[woven gust]

cute pose

[gleaming flame]

I came back from holiday trip... and almost forgot how to prompt... trying to get back to shape 😛

(by using the same prompt in different styles xD)

[woven gust]

anime?

[gleaming flame]

Ilya Kuvshinov is strong with this one.

[misty violet]

Looks like a cool little Santa helper from a fantasy world. Love the outfit and expression!

[sharp imp]

Jester clothing looks fun

[wooden sail]

NoobAI is good when the image isn't broken. Very wide stylistic range

[sharp imp]

Illustrious 2.0 does alright. Probably a little less artistic than noob, but it does recognize quite a variety of styles

[wooden sail]

And this is what happens if you add elden ring to the prompt:

[viscid sapphire]

I want to combine this two loras
https://civitai.com/models/203155?modelVersionId=228721
https://civitai.com/models/1538231/tung-tung-tung-sahur-cosplay
How can i do it? I am getting this

[stable crane]

you prob want the clip from the 2nd one

imagine a pipeline

this is approximately correct though

[lyric zealot]

I am surprised this image stay up for so long ngl

[sharp imp]

i mean ill take it down if you want. its definitely pushing it. probably should have spoiler tagged it to be safe lol

[lyric zealot]

i mean, i'm not a mod/staff, so you can do whatever you want. It's just surprising to me that nobody deleted it yet

or maybe because they don't care that much anymore

[sharp imp]

its not really explicit. but it is suggestive. shes clothed XD

it could be that its artistic and not raunchy?

[lyric zealot]

i mean, will you show that to someone that's not an adult?

[sharp imp]

i mean, i watched anime that was not much better than that as an adolescent. so probably i would. but then my idea of what is nsfw is probably less strict than others.

i wouldnt show it to a small child, but then a small child shouldnt be on discord or generating images with stable diffusion

[lyric zealot]

well yeah. People need to be at least 13 to be on discord, so if you think it's fine, just leave it up. My opinion is just my opinion anyway

[sharp imp]

something less risky

[glacial coral]

I saw these prompts and liked it quite a bit so I just grabbed a Lora combo my friend gave me goes real nice

[sharp imp]

Oooh pretty

[languid yoke]

I need to find a way to improve my work so that it turns out like above.😅

[woven gust]

kino no tabi

[tawny wasp]

My Female Orc: Erzoka 🙂

[tawny wasp]

ssj4

[glacial coral]

👀 I like the style also

[gleaming flame]

which one of the 3?

[river shale]

1st and 2nd imo

[gleaming flame]

3rd one too cartoonish?

[river shale]

na, just not as detailed as the other 2

more flat

[gleaming flame]

that's the style... but I get it

[river shale]

It does look nice for the style, I still like the 1st and 2nd more 🤷‍♂️ But that's just my opinion

[gleaming flame]

ok

[sharp imp]

I know what you meant by more flat. It's ok

[river shale]

I'd love some suggestions for background for my OC

i can't think of where to make her 🤔
This is an abandoned cathedral.

I usually just do a bedroom.

[wooden sail]

This one

[river shale]

Autumns a cute theme, I like the bridge 🤔 Wonder what I could come up with, maybe somthing more springy, with flowers and stuff.

[wooden sail]

Also
👷‍♀️

[river shale]

o0o those pants have me thinking something too.

ty for the ideas

[river shale]

super Saiyan cat girl 😄

https://tenor.com/view/caulifla-dragon-ball-z-super-saiyan-gif-15035166

I approve

[glacial coral]

added new samplers to forge so i can use res multistep with KL Optimal

[sharp imp]

new samplers? from where if i may ask lol

[dusky linden]

Ahh they are built in into SwarmUI and i think comfy

[dusky linden]

Forge had a lot pulled due licence infringement and obfuscated code to make it work iirc so you have to manually install then instead

[sharp imp]

mmm. sounds unnecessarily complicated

[dusky linden]

Obfuscated code was to hide that it used comfy code iirc

[sharp imp]

also sounds like something im not smart enough to do without fuckin it up lol

[dusky linden]

https://github.com/google/blockly/issues/8712
If you wanna read the drama lol

Meaning i cant trust a ui that does this

[sharp imp]

mmm i see

[glacial coral]

a extention u can download

here

https://github.com/MisterChief95/sd-forge-extra-samplers

https://github.com/DenOfEquity/webUI_ExtraSchedulers

these 2

enjoy

[sharp imp]

thanks!

[glacial coral]

got u remember if i can do it and im still "new" to AI so can u XD

[tawny wasp]

Hela Demon Girl Illustrious Apex_Ai

https://civitai.com/models/1614976?modelVersionId=1827713

[glacial coral]

Do people use more illus over NoobAI here I still think V-Pred NoobAI is very good

[sharp imp]

i prefer illustrious, but its personal preference

[glacial coral]

Ah I see does NoobAI have advantages over illus and does illus have one over Noob?

[unkempt vapor]

what is noobai? :3

[glacial coral]

https://civitai.com/models/833294/noobai-xl-nai-xl

[wooden sail]

From my experience
NoobAI finetunes have more style variety than Illustrious finetunes.
Also don't recommend using their base models. You'll need very detailed prompts for a generation to look good.

[glacial coral]

What NoobAI model would u recommend?

[wooden sail]

Wai-SHUFFLE-noobai

This is illustrious ☝️ pic non related

[glacial coral]

ok ty does v-pred supported in current forge i cant seem to switch to dev branch

[glacial coral]

also very nice how u get that sketch style?

[wooden sail]

I use Epsilon version. And bout the style - the prompt is simple: Look, mom, I'm an aritst!

[glacial coral]

xd

ty is v pred better than eps or is eps better?

[wooden sail]

V-pred doesn't work. So eps is technically better

[glacial coral]

ah i see i thought it did on forge or maybe it did but i cant switch to dev

[unkempt vapor]

thnx!

[tawny wasp]

My first Lora Noobai. is Kiriko 🙂

[glacial coral]

ah very nice

[tawny wasp]

Thx a lot 🙂

[glacial coral]

anytime

[woven gust]

kino no tabi

[tawny wasp]

Gojo Senpai 🙂

[glacial coral]

yes

i still need to find a good lora for Alucard

[tawny wasp]

mmm

[glacial coral]

rn stock prompt for bro is like 40% of him

💀

[tawny wasp]

try to make the sky another color

[glacial coral]

what u got in mind

i remember playing apex a lot i did ash to masters only

was fun

[tawny wasp]

Green sky 🙂

[glacial coral]

gotcha

[tawny wasp]

or dark green 🙂

[glacial coral]

ill do both

[tawny wasp]

🙂

[glacial coral]

im on hi-res rn (its a little slow cuz im on macbook XD)

[tawny wasp]

take it easy, calm down 🙂

[tawny wasp]

ohhhh 👌

what a beauty

[glacial coral]

thank you

[tawny wasp]

Green makes red more vibrant 🙂

Your welcome

[glacial coral]

ty for the colour combo

[tawny wasp]

🙂

[glacial coral]

@wooden sail v-pred seems fine

on forge

[tawny wasp]

My Legendary Super Saiyan Kiriko 🙂

[glacial coral]

hell yeah cooking

[crystal vigil]

Cat

[sharp imp]

what are you buyin?