🍥｜anime

[steady grail]

lower the steps and the cfg

[glacial coral]

Alr cheers ❤️

[steady grail]

on the setings of the hires too

[glacial coral]

I seen with 1 CFG negatives are ignored as far as I remember I’ll assume I can use 2 CFG Ik going higher may cause output issues with the Lora

[steady grail]

i use it at 1.7

[glacial coral]

Gotcha

[steady grail]

between 1.5 and 2

[tough matrix]

whats the difference with painting with krita inpainting/ outpainting vs doing it with comfy ui?

[steady grail]

more tools, and impainting with nodes is really unconfortable. saying by someone that loves to play with nodes, means a lot lol

[tough matrix]

Awesome I have a new workflow now to help with poses

Extract existing poses from prexistig pics and create with depth map and open pose

[tough matrix]

Can you share some workflows?

[fickle oyster]

@autumn roost Sorry to bug you but can you take out some trash?

...yes?

[near elbow]

winter wide

[steady grail]

inpainting with comfy gives me head hake

[tough matrix]

I mean your other workflows. Trying to see what others are doing..I am after max quality

[gleaming flame]

still discovering new styles inside illustrious... love this one

[fickle oyster]

https://tenor.com/view/john-travolta-gif-25290651

[keen hollow]

Vertlain needs to setup a website to showcase some favorite generations.

[gleaming flame]

Actually I need to stop making them

[steady grail]

more experimenting

[keen hollow]

but why? 😬

[terse ore]

How do I know one of those $1000 rtx3090 PCs on ebay doesnt have a worn down mining card? I want to upgrade my 3060 one to one of those but idk...

[steady pumice]

You don't it's a risk you take when buying used GPUs

[gleaming flame]

because it consumes my free time and is generally pointless, isn't it?

[gleaming flame]

fat?

[gleaming flame]

...and still making them cute girls...

[gleaming flame]

...and weekend is almost gone... sigh

[twilit portal]

🤍

[desert island]

So many good generations! Been gone for a bit but just came back and caught back up. Good stuff!

[tawny wasp]

Rumi From (KPop Demon Hunters) illustrious Apex_Ai
https://civitai.com/models/1861678?modelVersionId=2107015

[fickle oyster]

I do mean roughly everywhere.

[twilit portal]

💀 🙏

well maybe u just so happened to join those servers after me so

[fickle oyster]

Highly highly unlikely sir

[terse jacinth]

my good sir, what did you use to animate this ?

[fickle oyster]

Wan 2.1 LCM

[tawny dew]

Hello

I'm new to Ai generated art

I am trying to set up my own AI generator. Stable Diffusion Automatic 1111 to be specific

But I encounter a bunch of error and do not know how to fix it.

[ocean stirrup]

I tried to generate an image but kept getting grey boxes insatead of my images

[wheat falcon]

Come to #🤝｜tech-support with more information

[empty pelican]

Hello you still need help?

[keen hollow]

I like this one.

Makes me think of an adventure movie character.

[bright totem]

Hi

[terse ore]

I have a screenshot image made in a 3d program of a character. But the hair is in place. How can I use AI to make her hair fly upward as though she is falling from a height, while keeping the hairstyle itself the same?

[keen hollow]

ugh.. been trying to do image2video for a while and just.. woo.. I'm garbage at this.

Terra VonTwitchyMouth 😐

this one is a bit better

well.. it just takes a lot of resources, fiddling, tweaking, time, and... just isn't as fun as image generation

feels like i'm just fighting with it

[tough matrix]

wow great colors

[tough matrix]

wow what software? wan?

[tough matrix]

wow really creative

[steady pumice]

thanks

[keen hollow]

Yes wan 2.1 fusion.

[near elbow]

"stereotypical 1980s manga cover" 🙂

[tough matrix]

world tree!

[tough matrix]

mushroom city

[fickle oyster]

We just live in our own little art bubbles and I guess that's cool too

[gleaming flame]

we all live in bubbles, some smaller than others

[wooden sail]

Give him an alike-looking sister wearing the same outfit, standing next to him. Keep the man unchanged, maintain his original pose, maintain original framing

Nano banana again

[near elbow]

for comparison 🤓

(both with Qwen 😉 )

[glass citrus]

gudshit

[steady pumice]

thanks

[cold talon]

@steady pumice great stuff, can I ask what model you're using? all the images look very crisp

[steady pumice]

wai nsfw illustrious v14

[past junco]

Honest question, which is better? The original or the upscale? (Note, the original looks more cleaner but the upscale has style)

There's something extremely aesthetic about this particular model but I can't for the life of me figure out what it's composed of (it's a merger from my old model merging experiments phase, and I believe this one was a very complex one a la final product)

[terse jasper]

idk which onee is upscale but left one better

actually no upon closer look right one

[past junco]

Left upscale, right original

And it seems it can go either way

Obviously left here is original and right is upscale

[past junco]

So much personality, but still has flaws. Upscale does fix the flaws though, but tends to lose the crisp original vibe

[cyan stirrup]

hi, im a beginner in SD. currently using counterfeitv30(mostly)/control v11p openpose sd15 and waiNSFWillustriousSDXL_v140/openposeXL2 as check point/controlnet.

can anyone give me some advice to get a better result.
i am trying to create a character using this (skeleton) poses. i keep getting results like this.

either there are faces or there are some furniture adding. what am i doing wrong?

PS. when im using SD1.5(counterfeitv30): i can generate multiple poses from multiple skeleton in a single image it has the same problem.

[near elbow]

well...

this is simply "anime boy in a blue suit, standing at ease pose, arms stretched out to the side" with qwen, no magic required

"anime boy in a blue suit, standing at ease pose, flexing right arm, looking at the right arm, left arm stretched out"

for fun

[left wind]

You’re not doing anything wrong — it’s just OpenPose’s limits:
It only controls pose, not face, clothes, or background. So SD fills those in randomly unless you guide it.
Multiple skeletons look like a “scene”, so the model adds desks, chairs, or other extras.
SDXL is stricter than 1.5, meaning it needs clearer prompts and stronger control settings to behave. (I'm new too so I could be incorrect but I double checked with ChatGPT )

[cyan stirrup]

can you share me which checkpoints and controlnet you are using (if it has one)

[near elbow]

that's qwen running under comfyui, much more modern model, reacts to prompts like you would expect, no controlnet needed
just google for "qwen comfyui"

needs a good GPU though

[cyan stirrup]

also, i am just using chatgpt to check if my prompts are okay. haha im trying to find a general list for prompts but cant find one haha

[near elbow]

complex prompts are overrated. with a modern image generation model, just say what you want...
e.g. "anime style, a friendly robot neko sitting at a computer generating AI images, cluttered nerd cave background, speech bubble text "Just prompt for it""
with SD 1.x and SDXL you have no chance to do that, they don't do text well, and they don't understand prompt details and objects and relationships between them

[cyan stirrup]

i see. so its better to switch to comfyui already.

[near elbow]

comfyui is just the platform which runs the models (it can do SDXL, Flux, Hidream, Qwen, etc etc etc). personally i don't like it very much since the GUI can be very inconsistent and confusing, but if you go with the official example workflows, it just works

give it a try and have fun!

[cyan stirrup]

yes. thanks!

[quaint oar]

Lux

[quaint oar]

Marry Me lux

-melts-

[cold talon]

thanks

[near elbow]

happy sunday!

[cyan stirrup]

i tried using comfyui.
sadly my vram doesn't have the capacity to run qwen image.
so i tried using what i am using at sd. haha.
if i miss something so i can use qwen image let me know.
im usiong rtx 3060 6gb vram. 16gb cpu ram

[near elbow]

ouch, that's way underpowered. if you have the budget, get a more recent nvidia GPU with 16GB VRAM or more. if not, cloud computing might be an option, but is not for free usually

[cyan stirrup]

i see. its okay. i am just trying to learn anyways. i am planning to buy better system in the future.
well that means theres no way to use qwen for now on my current system right.

[keen hollow]

I recommend GPU rental. I stopped doing local generation, it's just too .. cheap? ... to do it online.

Plus, added benefit of less skethcy crap loaded to your computer. 😄

@calm oar if you're still here, I did see your DM request but I do not respond to DMs. To answer your question, that was just made with Vidu, I was just trying their free demo.

[tawny wasp]

Loona From (Helluva Boss) V1 Illustrious Apex_Ai
https://civitai.com/models/1894832

[keen hollow]

hmm

dang tried to upload a few images and they just.. disappeared

[past junco]

As someone who doesn't have a high end PC, I can run SDXL pretty well with SwarmUI even when ComfyUI and Automatic1111 failed

Worth looking into at least

And don't let anyone tell you that 1.5 isn't good (it may be old, but it's still gold)

I can't make anything close to this style with SDXL or other

I have some neat tricks for anyone who listens

Next step for me is to generate training images for an SDXL model using my master merge models... once I get a high end PC :3

[fickle oyster]

Cmon guys wake up

[wooden sail]

Take this!

[fickle oyster]

WAKE UP!

[bold ether]

Grab a brush and put a little make-up ?

[fickle oyster]

yes

[fickle oyster]

KEEP THOSE EYES OPEN

[near elbow]

👁️

more wholesome version

[tiny salmon]

i love this what prompts did u use

[wooden sail]

Turn around 😮‍💨

[indigo oyster]

strong oshi no ko vibes

[slate narwhal]

She will devour your eyes... Nano banana amost got the set of teeth for eyes right. Inpiration input was the Corinthian from Sandman, second image.

You can get past the censorship for the eyes she is holding by using "spectral eyes" in the prompt

[terse ore]

What's the largest qwen edit gguf model one can use with a 12gb ram gpu?

[near elbow]

why change the eyes when a simple smile can be soooo charming 🙂

[tough matrix]

ahahh did you take my prompt

[wooden sail]

Your image. I took it
And then run it through gemini

[tough matrix]

its cute yeah

[near elbow]

satyr inspired

[hard garnet]

@fickle oyster how can i generate art style like that? is it like what they call a specific lora?

[tawny wasp]

Loona From (Helluva Boss) Noobai Apex_Ai
https://civitai.com/models/1906659?modelVersionId=2158074

[cyan stirrup]

Needing some help.
im trying to use image to image to generate different views of specific part of this image(2nd image attached). like wide front view of the 2nd image attached. is it impossible to do it?

[gleaming flame]

img2img can make the initial view different/better, but won't change the angle etc

[wooden sail]

Only large Transformer-based models can do stuff like that

[wooden sail]

Gemini in this case

[cyan stirrup]

what do you mean by that sir/maam? i'm still new in this field sorry for my ignorance.
but i see some transformers installed in my kohya was it the same.

[wooden sail]

Go to Google Gemini website, upload your image and ask it to draw this scene from the view of someone looking directly at the vending machine

Also search for 'nano banana' on Twitter or reddit for more info

[tawny wasp]

Momokusa Yomogi From (Tougen Anki) Illustrious Apex_Ai
https://civitai.com/models/1909676?modelVersionId=2161474

[tough matrix]

Super High Resolution attempt at wallpaper material

[hollow island]

looks like a typical Flash animation

[hollow island]

Terra has had it with your s*** and is coming to beat you to a pulp, lol

[hollow island]

Awkward laughing would be so Terra. I think FF6 kind of left out that she would be kind of socially awkward due to her upbringing.

[desert island]

okay this was fun... back to work now 😁

[fickle oyster]

HANDS ITS ALWAYS HANDS

[errant pike]

this looks good

[tawny wasp]

Ayase Momo From (DanDaDan) Illustrious Apex_Ai
https://civitai.com/models/1918627?modelVersionId=2171543

[tawny wasp]

Nice job 🙂

[near elbow]

spammers beware 🙂

[empty pelican]

Awesome! It's more cute than scary.

[near elbow]

robots at your service

[terse ore]

Theres a guy selling his 5090 PC for $2700 near me, what a deal

[near elbow]

... that being said, i'm quite happy that i bought a new 4090 rig when that came out in 2022, still good enough for anything (more VRAM would be nice of course). the 5090 is even crazier

[empty pelican]

I've never used the 50 series GPU, but I've heard it's often incompatible with Python libraries, leading to dependency errors. Is that correct?

[near elbow]

well the 50 series drop support for older versions of quite a lot of things (e.g. also PhysX), but most major image generation infrastructures support it now, and the number will grow in the future. we're not in the early adopter phase anymore, should be interesting. maybe wait until a non-flagship version with 24GB VRAM or more comes out, that should be similar to the 4090 but maybe require less power...

and of course, as always when talking about the future 😉

the longer you wait, the cooler the hardware gets, and the cheaper the current performance level will become - but on the other hand you're missing out on having fun with the hardware RIGHT NOW 😉

[empty pelican]

Awesome!

[keen hollow]

How's everyone doin'?

[keen hollow]

cat

[near elbow]

now that you mention it 🙂

[keen hollow]

trying to start a vast instance but they're being slow today 🙁

[near elbow]

you need to bring in the RACING cats then!

[keen hollow]

one brave cat without goggles

Tifa cat?

[fickle oyster]

copy paste?

front page of civ?

[fickle oyster]

looks like identical to this one

[left wind]

First time using Wan not the best but I tried

[bold ether]

flame is a bit weird but it looks pretty good to me otherwise

[tawny wasp]

Dream Core Style Illustrious Apex_Ai
https://civitai.com/models/1930749?modelVersionId=2185260

[near elbow]

how about some action? 🙂

[keen hollow]

yep same prompt, though different model

most of those in that line were just me combing various prompts

after the tifas 😛

[fickle oyster]

k.

[keen hollow]

wish the text on this one was better 😄

[tawny wasp]

ACGS Style Illustrious Apex_Ai
https://civitai.com/models/1931475/acgs-style-illustrious-apexai

[acoustic agate]

[near elbow]

change of style 🙂

[near elbow]

flux fusion to the rescue (a bit better with text)

[keen hollow]

I've been messing around with nano banana

[glacial coral]

I’ve returned to pass u one image after 12 years of non existing

[tough matrix]

wallpaper resolution

[keen hollow]

I like this one... I have a hard time with the 'cyberpunk' images because I think they all look the same. But this one seems more unique with some interesting details.

[glacial coral]

Cheers it can be a challenge some times tho the 748cm style Lora helps changing it up

[tough matrix]

What are youse strategies for hands?

[glacial coral]

Hand detailer or i inpaint

I wanted to use Wai V15 and I love V14 and V15 finally came but it seems to create 6 fingers a lot compared to V14 😞

[fickle oyster]

Wai us horiffically overrated imo.

[glacial coral]

Does what I need it to do other than that I use SIH 1.5 or Hassle

What other models u recommend?

[fickle oyster]

Honestly I mostly use my models and I think they come out cleaner, depending. Janku is pretty good though

[fickle oyster]

older versions, I think v10 and below were better for Wai as well

WAI being the #1 model, there are many comparable ones that have better colors and more "pop"

[glacial coral]

Ah I see cuz I’ve seen some models have issues with more than 3 Lora usages perfect illus seem to be the best when it comes to handling lora stacking

[fickle oyster]

prefect is pretty good, yeah.

[glacial coral]

I do enjoy those I do wanna try more NoobAI based models

I’ve only got SIH rn

Cuz I’m clueless on Noob models apart from that one I was given

[fickle oyster]

Understandable. There are... A lot.

but with A6000 you can afford to experiment

[glacial coral]

I’ve also been trying Koheh HRFix over normal Hi-Res

Idk if I spelt that name right

[glacial coral]

True qwen seems to take the piss tho to gen with

I wish Chroma was as easy to rock with prompt wise and stuff as Flux DEV models like fine tunes or Krea

[glacial coral]

I really like this

[fickle oyster]

Thank you, I can toss you a flow. You can def run it

[glacial coral]

Tho one thing I miss from moving from MJ to local is how simple MJ is to get results with but u have 100x more control with local and no subscription

[glacial coral]

Bet toss me it

[fickle oyster]

If you like the style, it's: https://civitai.com/models/1465491/ramthrusts-nsfw-pink-alchemy-mix

now for the vids

[glacial coral]

Solid I like it I’m more very dark and muted palettes sum u would see in horror stuff but still simple but I don’t mind this

[fickle oyster]

GET THESE WAN MODELS

and these loras:

[glacial coral]

Dw I got like almost most WAN 2.2 models downloaded XD

[glacial coral]

I have that also

[fickle oyster]

perfect

yeah the fp8 scaled stops shit from exploding in a big way

[glacial coral]

Bet some things I might be limited with since my card is Turing based basically a RTX 2080 TI with 24gb vram

Or sum

That’s what I was told

IIRC speed is like 4060 TI fuck knows

But hey I got it pretty cheap

[fickle oyster]

I got a 4090 so I'm spoiled

[glacial coral]

Maybe one day XD 20 year old me can’t get £1.1k+ on a GPU yet

[fickle oyster]

"How do you find out if someone has a 4090?"

"Wait 5 minutes"

[fickle oyster]

Yeah completely understood.

[glacial coral]

Still need to upgrade my cpu with a platform change and ram upgrade to DDR5

[fickle oyster]

I had to plea with my wife to spend my own money to get one

[glacial coral]

And then after I shall see what I can get

[fickle oyster]

DDR5 upgrade is... not as good as you'd expect. With DDR4 computers reboot instantly

DDR5... like 5 minutes even at 6000MHz

[glacial coral]

I have no choice XD going AM5 I game more so 9800x3D is my choice of CPU unless I get a good deal on 7800x3D

[fickle oyster]

OH FUCK

[glacial coral]

I’m on AM4 rn since Ryzen 3000 days so still rocking b450 board

[fickle oyster]

I'M SO SORRY ABOUT THE NSFW LOAD IN THE FLOW

Ah fuck....

[glacial coral]

Nah Dw I do certain stuff when I generate lmao

[fickle oyster]

I totally forgot

same obv

[glacial coral]

I saw it and was like “usual behaviour”

[fickle oyster]

Yeah that was a model test for freckledvixon

(The queen of all loras)

[glacial coral]

Fair I mean other than illus models and maybe flux models haven’t done SDXL NSFW in a hot minute lmao

[fickle oyster]

(all hail da queen)

[fickle oyster]

Yeah this is true

[glacial coral]

I just cba with hands on SDXL

Apparently a lot is fixed now

But idk

Not tried

For SFW realism I use Flux Krea FP8 scailed

With ViT L HE long clip and T5 v1.1

[fickle oyster]

Oh with that same model I posted earlier... Not exactly horror but:

[glacial coral]

That’s solid

I love my dark themes and muted colours

[fickle oyster]

I'm a huge contrast fan

[glacial coral]

If anything I like my cyberpunk 2077 game colours and aesthetics if not my dark atuff

*stuff

[glacial coral]

Dark aesthetic with heavy contrast like this goes hard

[fickle oyster]

BULLMA

[glacial coral]

Bro is tryna make me go inside go upstairs and use my pc 💀

[fickle oyster]

Haha sorry

[glacial coral]

Nah all good dw

I’m just enjoying the images

While I play music in my ears full blast

[fickle oyster]

noice

Yeah

Just having fun now

[glacial coral]

I can tell I love it

[empty pelican]

How did you generated these identical person's images but in different clothes?

Could you plese let me know that? @wooden sail
Thanks

[fickle oyster]

Using this one, no less.

#🍥｜anime message

[wooden sail]

Qwen-Image-Edit

[empty pelican]

I mean, how can I change the clothes only?

[empty pelican]

Perfect!

Thanks

[fickle oyster]

Yes, and I'm saying he didn't gen the original image

@wooden sail How much better than Kontext?

kek

[empty pelican]

Which platform did you use?
Automatic 1111 or Comfy UI?

[wooden sail]

Better at some tasks. Less consistancy but feels more flexible. It's ComfyUI only

[fickle oyster]

yurp comfy user here

DIS IS GUD DATA

[empty pelican]

Could you please share your workflow with me?

Thanks

[wooden sail]

Rotating characters is neat.

[wooden sail]

It's this one
https://docs.comfy.org/tutorials/image/qwen/qwen-image-edit

[empty pelican]

And the prompt was what?
Could you share your workflow image with me please?
Thx

[wooden sail]

Prompts were
Draw side view of this character
Draw this character wearing woven sweater
Draw this character wearing red military uniform

[near elbow]

anime meets pixel art 🙂

[empty pelican]

Sorry for the repeated questioin.

Can I generate images here with gtx 3060 whose vram is only 12GB?

[bold ether]

those are two different questions.
"Generate here", that would be using artisan's bot. cf #artisan-faq
"generate [locally using a] 3060 ", yes no problem.

[empty pelican]

But the model used there is 19GB

[wooden sail]

Watch this to the end:
https://youtu.be/WOcxMUwKWIk?si=_TeuLX_VvpVIQya7

[bold ether]

Use some quantized GGUF versions. eg : https://huggingface.co/QuantStack/Qwen-Image-Edit-GGUF/tree/main (haven t tested those yet)

[fading vapor]

https://x.com/5anjoker/status/1964554086138990954?t=iX5wxDokMvNfbaOfKXqcPw&s=19

5aNjOkEr_XL (@5anjoker)

⚔️ Smile before the storm… 🔥🌙

#5an_AI #AIart #digitalart #darkfantasy #オリキャラ #イラスト #fantasyart #AIイラスト︎

X · Sep 7, 2025, 4:59 AM

[near elbow]

solemn mood

[tawny wasp]

Makima (Chainsaw Man) Illustrious Apex_Ai
https://civitai.com/models/1953801/makima-chainsaw-man-illustrious-apexai

[gusty palm]

kinda looks like black rock shooter

[tawny wasp]

Yoruichi Shihouin (Bleach) V1 illustrious Apex_Ai
https://civitai.com/models/1956801/yoruichi-shihouin-bleach-v1-illustrious-apexai

[versed shadow]

Hi what's the best model u guys recommend for NSFW anime image ,i have a project and i need a clean,accurate and very high quality model that can make flawless or at least near that ,.TY. for ur help in advance

[pine sierra]

NovaAnimeEX, Wai-nsfw-illus v15

[desert island]

And rounding out the Demon Hunter cast 😁

[waxen idol]

i can't believe the result

[near elbow]

get the latest in cyber eyes

[near elbow]

head wings, now that's neat 🙂

[near elbow]

collect them all!

[tawny wasp]

Sylpha Langlis (tensei shitara dai nana ouji datta node kimama ni majutsu o kiwamemasu) illustrious Apex_Ai
https://civitai.com/models/1942875/ilxl-sylpha-langlis-or-tensei-shitara-dainana-ouji-datta-node-kimama-ni-majutsu-wo-kiwamemasu-i-was-reincarnated-as-the-7th-prince-so-i-can-take-my-time-perfecting-my-magical-ability

[fading vapor]

https://x.com/5anjoker/status/1968553879148347711?t=TDU859UzqZ_CoVMqhWl0vQ&s=19

5aNjOkEr_XL (@5anjoker)

🌅💠 seductive but untouchable…✨

#5an_AI | #AIart | #AIイラスト︎ |#AIgenerated | #digitalart

X · Sep 18, 2025, 5:53 AM

https://x.com/5anjoker/status/1968597532218384438?t=xi6P9lLiOuHons_rvcRqjg&s=19

5aNjOkEr_XL (@5anjoker)

✨ Soft glow, golden vibes & a touch of midnight magic 🌙💫
#DreamyAesthetics

#5an_AI #AIイラスト︎ #AIart #AIGenerated #DigitalArtist

X · Sep 18, 2025, 8:46 AM

https://x.com/5anjoker/status/1968638691615678957?t=fjSVW4D53ysF2YHIHxLH1A&s=19

5aNjOkEr_XL (@5anjoker)

💎 Neon queen with horns & heat 🔫✨ #CyberSiren

#5an_AI | #AIart | #AIイラスト︎ #AIGenerated | #digitalart

X · Sep 18, 2025, 11:30 AM

https://x.com/5anjoker/status/1968683758648770888?t=n2X47fZBID4vYDBEUP60OQ&s=19

5aNjOkEr_XL (@5anjoker)

✨👑 A goddess draped in light and shadows, ruling with grace and power ⚔️🔥 #DivineElegance

#5an_AI #AIart #AIイラスト︎ #digitalart

X · Sep 18, 2025, 2:29 PM

[fading vapor]

https://x.com/5anjoker/status/1968736998543245411?t=kin06FxEUMmzNWOJ391j1A&s=19

5aNjOkEr_XL (@5anjoker)

✨ Gothic allure meets divine light 🕯️💀 | Sensual vibes in the cathedral 🌙🖤

#5an_AI #AIart #AIイラスト #GothicGirl #digitalart

X · Sep 18, 2025, 6:00 PM

[fading vapor]

https://x.com/5anjoker/status/1969094022909010371?t=A1puYAgNm1TB7YCN4R97Mg&s=19

5aNjOkEr_XL (@5anjoker)

Catch the tide — and my hat if you can 🌊🎡😝

#5an_AI #AIイラスト #Anime #DigitalArt #AIArt #SDXL

X · Sep 19, 2025, 5:39 PM

[gusty palm]

fycuking love these

currently upscaling

sposed to be black rock shooter

i got like 1 of 10 with the correct flaming eye thing but its hard to prompt

[fickle oyster]

Thank you

[gusty palm]

oh wow

u made it look easy

[fickle oyster]

eye trail, light trail

oh, eye flameas well

nano banana also does this if I ask it nicely, but at a much lower res

[desert island]

I've seen this one before 😁

[fickle oyster]

How about this one?

[desert island]

oooh I love the colors on this one!

[fickle oyster]

RNG baby!

[keen hollow]

Fall and Halloweeny... nice. Moar

[gleaming flame]

Saturday vs Sunday

[near elbow]

just kidding

[gleaming flame]

soon...

[wooden sail]

Is this Qwen?

[gleaming flame]

No idea who is Qwen so I guess no.

[near elbow]

i recommend seedream, incredible prompt following... this is anime style a cool goth waifu shopping in a record store, lots of manga art on the wall, buying an album by miku rockstar showing miku on the album cover label "Miku Greatest Hits" subtitle text "Another beautiful day with Seedream Image Generation"

no chance to get even close with older image models...

[wooden sail]

This

[near elbow]

a sequence of manga panels showing the evolution of miku rockstar from a stick figure to a photorealistic version 🙂

[wooden sail]

Have you tried seedream edit? Does it produce short stumpy bodies just like Kontext and Qwen?

[gleaming flame]

not on my gpu... still using XL

[near elbow]

not yet, just image generation... flux kontext is running locally quite well and i've been testing it out, but my main use case is generation from scratch, not editing.

[near elbow]

seedream isn't available locally at all, i'm using replicate to run it, well worth the 3ct per image 🙂

quite censored though, but that's no real dealbreaker for most usecases

[desert island]

FINALLY! took too long for RNG to get the vibe and style I wanted.

[gleaming flame]

not the best clothing choice

[wooden sail]

First season
Second season

[mossy hollow]

I prefer the first one 😍😍 beautiful light and shadows. The second one I appreciate the dark atmosphere but is not my favorite style. Gg for the work bro 😎

[desert island]

💯 Yup love the style of the 1st season too

[near elbow]

seedream is fun (i might have mentioned that once or twice already 🙂 )

[near elbow]

"a sequence of anime images showing the evolution of AI image generation" with seedream 🙂

yeah, seedream is incredible

[desert island]

was it all in one image generation? without masking or anything?

[near elbow]

literally that prompt, one single image, first try, no postprocessing
(yeah, we've come a loooong way since the stable diffusion beta here 🙂 )

[near elbow]

more fun with seedream

[gusty palm]

wow

whast seedream

[near elbow]

https://seedream-4.com/

warning, paywall etc, but you see the use cases

[gusty palm]

playing with this openoutpaint thingy

[gusty palm]

this thing is so cool

https://i.imgur.com/kE4QTqV.mp4

[thorn jolt]

hola

[desert island]

one more 😄

[keen hollow]

Then the sequel...

[desert island]

🤣 OMG This was too funny! 👍 nice work!

[keen hollow]

no work involved lol

[desert island]

🤷 well you know what I mean.

[keen hollow]

haha

just messin' with the banana

[desert island]

https://tenor.com/view/winking-banana-gif-20160391

[keen hollow]

yep, him

[terse ore]

Does illustrious have a controlnet like xinsir union pro for sdxl? I think illustrious controlnets arent great but I've not checked for a while

[lyric zealot]

at this point, just automod and ban if message attachment contains 1.jpg

[near elbow]

but the comply thing is for the ED-209, not the cyborg robocop (i love that movie)

hmmm, impressive but seedream doesn't seem to know ed-209
robocop ed-209 in a three panel comic strip, sarcastic, poking fun at AI

bonus points for coming up with 1987 without that being in the prompt, though

[lyric zealot]

ain't no way they change the attachment to image.png

[pale mural]

Security breachhhh 😂

[desert arch]

Hello chat

[near elbow]

she's embarrassed because the screen just went black 😉

[bold ether]

or because she just got scammed buying that keyboard

There's not even proper arrows keys

Ooor she's a leet dev with macro keys on the right. Your choice.

[near elbow]

chatGPT?

[keen hollow]

The banana.

[near elbow]

neat

[fickle oyster]

@autumn roost Hey boss, we have a SCAMMER on the INTERNET!

D:

[desert arch]

Gm

[tawny wasp]

Amo Empool (Gachiakuta) illustrious Apex_Aihttps://civitai.com/models/2000504?modelVersionId=2264343

[near elbow]

cleeeeeaningwoman? (with apologies to dead men don't wear plaid)

[quasi shadow]

Tried this style for the first time
Turns out I really like it!

[fickle oyster]

it's so much fun

so many things you can do with it

Limited only by your

https://tenor.com/view/rainbow-spongebob-imagination-gif-11877335675029242228

[quasi shadow]

How do you get this poses tho?

I have a quite bad pc so I use gpt to try and get the prompt perfect and maybe get the image first try

But poses and "camera" positioning, that i can almost never get right

[fickle oyster]

booru tags

[desert arch]

Afternoon chat 🥕

[near elbow]

well, with modern models you can just prompt for it, no additional magic required
vintage anime style, an enticing attractive female, left arm up showing the devil sign, right arm on her hip, tilted hip, balancing on left leg with the right leg up in the air, subtitle text "It's all about the model nowadays"

devil's horns with red fingers, nice detail

(seedream 4.0, first try)

[desert arch]

Like RAM said, booru tags works as well but it also depends on how well the model follows certain tags

I used these for the image below, with the art style and character lora excluded (i assume it's still safe to post this)
medium close-up, collarbone, white tank top, cleavage, hand pulling shirt, grin, wink, leaning forward, from front, breast gap, indoors, loose shirt, sweaty, blushing,

[keen hollow]

is seedream local? Or online service only?

[near elbow]

seedream on replicate, 3ct per image at 4k, convenient and reliable. there's free services out there but there's usually a catch with that. AFAIK there is no seedream model publicly released and it's probably quite large anyway, and i "only" got 24GB VRAM with the 4090 🙂

[acoustic agate]

She's balancing on her right leg tho. Failed to follow the prompt. AI still has a long way to go

[near elbow]

well it's a start

[tight linden]

🎌 Anime Showdown Challenge! 🔥
Hi everyone! I generated these 3 anime battle scenes with Flux Schnell for a challenge. Which one do you prefer?

Vote with reactions:
🔟 → Scene 1: Epic Duel (Celestial Blade Symphony, 16:9)
💥 → Scene 2: Cyberpunk (Neon Phantom Strike, 3:4)
🐉 → Scene 3: Mythic Dragon (Dragon’s Wrath Awakens, 9:16)

(Full prompts & settings in reply!)

[near elbow]

i would change the language to english, since it's the server's official language and you will reach many more people if you use it 😇
and i understand enough french to get what you're saying, so here's my brief feedback:

• interesting mix of images and styles. i wouldn't consider #2 nor #3 as anime though
• personally, the cyberpunk kick #2 is most impactful from colors and dynamics

[tight linden]

Thank you so much for your feedback, @near elbow! 😊 Really appreciate the input—especially about Scene #2 (cyberpunk) being the most impactful!

To everyone else: What do you think? Should I focus on improving Scene #2 further, or adjust Scene #3 to fit the anime style better?

(Also, noted on using English—thanks! Will do for future posts. 🙏)

[tight linden]

🎌 Anime Showdown Challenge! 🔥
Hi everyone! I generated these 3 anime battle scenes with Flux Schnell for a challenge. Which one do you prefer?

Vote with reactions:
Scene 4: Neon Vendetta (CyberSamurai MechaBattle, 16:9)
Scene 5: Elemental Rivals (MagicBattle Epic, 4:5)
Scene 6: Cosmic Duality (Duality CosmicAnime, 1:1)

(Full prompts & settings in reply!)

[near elbow]

hard to say, they all suffer from technical issues (limbs, angles, ...). can you maybe use flux fusion instead of flux schnell (roughly flux dev quality with flux schnell speed)?
from the general mood i like the first best

just for fun, "anime style Neon Vendetta Cyber Samurai Mecha Battle" with seedream 🙂

[desert island]

okay, that was fun 😄. Thanks for the mecha theme idea, it got my noodle going 😁

[keen hollow]

ah ok, got it. wasn't sure. 🙂 I think I played with it on a site somewhere that gave 20 free images or something at some point.

[keen hollow]

These remind me of Simon Stålenhag.

[tawny wasp]

Nazuna Nanakusa (Call of the Night) illustrious Apex_Ai
https://civitai.com/models/2007409?modelVersionId=2272068

[desert island]

I just looked it up... and oh yeah... Electric State! I saw some images but wasn't really paying attention, my wife was looking through it but I only casually saw it on the coffee table. I wonder if that was where the idea popped up from 🤔

[keen hollow]

well now we need everyone to do 1000 more

[desert island]

💯 percent!

[desert island]

these are cool!

[desert arch]

Are these made with Gemini? (assuming because of the watermark)

[keen hollow]

Nano banana via AI studio

[desert arch]

Ah okay 👍

[desert island]

Okay last one I swear 😁

[tight linden]

Thank you very much for your advice, which is so valuable to me 🙏. I will follow it. 😊

[tight linden]

I hadn't thought of that. Thank you.

[stone cedar]

masterpiece!!

[jagged axle]

Hi very very beautiful picture anime. I like it 💖💖💖💖💖💖💖💖💖💖

[nova remnant]

Do y'all know a site that turns photos into ai

[fickle oyster]

yup

[desert island]

It's kind of weird to say given the context of the image. I don't know why but I love that street side tree on the left side of the image in the background. I actually just love the whole background in general.

[desert island]

this is a really cool style 👍

[near elbow]

mix it

[near elbow]

hip wings, always fashionable 🤪

[wooden sail]

Something something... Center of mass

[near elbow]

well hermes/mercury had wings on his boots, that's even worse

[desert island]

Yay! I love Albedo!

[gleaming flame]

huggy...

[past junco]

I love this style

[gleaming flame]

actually quite a mix of different styles

[near elbow]

someone at your door

[gleaming flame]

start with a decent illustrious model like diving anime or for more sketchy look animij, then add a solid dose of 0hu-ton and JollyJack styles (both have loras on civitai)... you can try other more "sketchy" and "stretched" styles if you know any...

experiment... always experiment... thats what makes AI fun, at least for me

that, and making sexy girls...

she be eating again...

[near elbow]

"moebius heavy metal cel shaded style, sketch look, cute female with chaotic hair biting into a sandwich in a futuristic mecha cockpit, smirking at the camera, speech bubble text "First try"" with seedream... not quite there yet 🙂

3rd arm on the house

[gleaming flame]

always handy

[near elbow]

"moebius clean minimalist cel shaded style, line drawing, cute female with long white hair parted in the middle, round nerdy glasses, torn jeans, skimpy puffy white blouse, biting into a sandwich in a futuristic spaceship cockpit, smirking at the camera, speech bubble text "Next try", view from below"
well, considering that this is with a totally different model (seedream) and all... good enough, although your version is still cuter

[left wind]

What's a good model to use I haven't been on here in a good while ?

[stone cedar]

Chinese style

[near elbow]

... and also what your goal is - anime only, specific styles and characters, not only SFW, etc?
more modern models are just so much fun, e.g. this is seedream simply with happy anime style cel shaded cute female hacker neko, datacenter, speech bubble text "Depends on your setup - cloud? local? which GPU?" - no way you can do this with anything older than flux and even that will struggle with the text

[left wind]

I just bought a 5090 and all local

[near elbow]

that should be quite ok for the next few years - 4090 here, got it when it came out 🙂
with that firepower, for local generations i recommend flux in all tastes (e.g. flux fusion with 6 steps should run in less than 4s on your card for one 1024x1024 image, i prefer forge instead of automatic1111), more special stuff might benefit from chroma as a model (flux based), and for more complex things qwen image generation (works OOTB with comfy, no forge support unfortunately), hidream and hunyuan are great. take 40-150s for a 1024x1024 image on my setup, so it's less fun than just playing around and getting the result basically at once

currently i like seedream so much that i use cloud services, no local version available...

[left wind]

I'll give it a look I use comfy UI thanks for helping me out 👍

[near elbow]

all for a good cause 😉 - have fun!

[velvet nexus]

This style is so awesome. Can you share me what model and prompts are used?

[gleaming flame]

it's not that simple, I used multiple illustrious models for these, that's why every image is in slightly different style. Basically add onono imoko, rariatto, (and jollyjack lora) to your prompt as a starting point, or you can start with murata range style and evolve it with img2img and mix with mentioned artist styles

oh, models I used for these: diving illustrious semi-real 10, animij and illustrijQuill

[desert island]

Yeah you turned me onto diving and the whole illustrious line so thanks for that.

[desert arch]

Gm

[gleaming flame]

they are no-brainer if you want cool anime styles, often without using loras

this time with brand new lora

https://civitai.com/models/2007986/jackie-droujko-style if anyone''s interested

[gleaming flame]

not sure if still counts as anime...

[desert island]

i'll allow it 🧑‍⚖️

[gleaming flame]

phew

[gleaming flame]

aaand that's all for now 😛

[near elbow]

not really NSFW i guess

[gleaming flame]

one more take of moth girl

[gleaming flame]

ok, too much of sweetness will kill you... eventually... That's the last one.

[tawny wasp]

Toru Hagakure (Invisible Girl) - My Hero Academia - illustrious Apex_Ai
https://civitai.com/models/2024057?modelVersionId=2290759

[near elbow]

classical style

something different

[gloomy lodge]

Anyone know how to readd old samplers to sd i want all the old sampling methods back to run tests just dont know how to get them back

[near elbow]

automatic1111? forge? comfy? swarm? the answer totally depends on your setup 😉
and it belongs more into #🤝｜tech-support not into anime...

[gloomy lodge]

oh perfect didnt even see the chanel

[lyric zealot]

not the place for this - rule 5

[keen hollow]

The sequel - KPop Clown Rockers

[gleaming flame]

nope, they are australian psyche-indie electro pop-rock trio

[keen hollow]

ah, naturally

[pulsar geyser]

🎉 LIVE DEMO: 8-Step > 20-Step Diffusion Breakthrough

It's an implementation of the paper: Hyperparameters are all you need. I've just launched my HuggingFace Space where you can test this yourself!
the link of the paper is below: https://arxiv.org/abs/2510.02390
🔗 Try it:
counterfeit v3.0 version: https://huggingface.co/spaces/coralLight/Hyperparameters_Are_All_You_Need
xl version: https://huggingface.co/spaces/coralLight/Hyperparameters-are-all-you-need-xl-version
original sd version: https://huggingface.co/spaces/coralLight/hyperparameters-are-all-you-need-sd-version

some examples is below

✨ What you'll see:

• Side-by-side with DPM++2m
• 2.5x faster, BETTER quality
• Works with any model

📊 The impossible made possible:

• 8 steps generate images with FID performance comparable to the 20 steps
• No training/distillation needed
• 60% compute reduction

Would love your feedback! What models should I test next?

[dusky linden]

Spamming it in all channel makes me less inclined to try it

[devout sequoia]

Hi

[left wind]

Knight

[near elbow]

morning... cool down, relax