[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
▼ Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: adt_153.webm (2.67 MB, 1016x1276)
2.67 MB
2.67 MB WEBM
Previous: >>110086767

>UIs
AniStudio: https://github.com/FizzleDorf/AniStudio
ComfyUI: https://github.com/comfyanonymous/ComfyUI
SwarmUI: https://github.com/mcmonkeyprojects/SwarmUI
KritaAI: https://github.com/Acly/krita-ai-diffusion
Forge Classic/Neo: https://github.com/Haoming02/sd-webui-forge-classic

>Image Models
IL: https://rentry.org/illustrious_checkpoints
Anima: https://huggingface.co/circlestone-labs/Anima
Krea: https://huggingface.co/krea/Krea-2-Turbo
MOSS: https://huggingface.co/Comfy-Org/MOSS-Image/tree/main/diffusion_models

SAAS:
NovelAI: https://novelai.net
Pelune: https://pelune.studio/

>Video Models
H3: https://huggingface.co/Comfy-Org/MiniMax-H3
Turbo lora: included above

>How to generate anime:
image : https://rentry.org/comfyui_guide_1girl
video: https://rentry.org/wan22ldgguide (No H3 guide yet)

>Style Explorers
Anima: https://anima.mooshieblob.com/ | https://animadex.net/
Illustrious: https://tagexplorer.github.io

>Checkpoints, LoRAs, Upscalers, & Workflows
https://civitai.com
https://civitaiarchive.com
https://tensor.art
https://openmodeldb.info
https://www.seaart.ai
https://www.liblib.art/

>Resources:
Share Metadata: https://litterbox.catbox.moe
Img2Prompt: https://huggingface.co/spaces/fancyfeast/joy-caption-beta-one
Online metadata viewer SD/NovelAI: https://spell.novelai.dev
Catbox/Metadata Userscript: https://gist.github.com/catboxanon/ca46eb79ce55e3216aecab49d5c7a3fb
Output cleanup: https://rentry.org/RemovingDiffusionGunk , https://www.mediafire.com/file/vipr23exc5htmnt (batch processing python script)
Inpainting Guide from an Anon: https://files.catbox.moe/fbzsxb.jpgcollage >>106520607

>/adt/ logo:
https://files.catbox.moe/pakfgt.pngcollage
>>
>>
The MOSS people said they are gonna train a new model exclusively for Anime, so you better go and spam them the datasets from booru and whatever bs else so they make something decent
>>
File: 198.jpg (306 KB, 1080x1092)
306 KB JPG
>>110108241
ngl I don't have a lot of hope
>>
I'm back o these threads y'all
>>
>>110108296
>NoobAI dataset
>furries and cosplay shit
please don't pollute this as well with that shit.
>>
>>110108296
They can't just "use the noob dataset" they would have to re-caption it
>>
>>110108616
Do you think they have a good Booru dump or something? I'm afraid they will train with a nothinburger or algo and will be a missed chance. The base model isn't too bad
>>
maybe tdrussel can lend a hand and save local again :D
>>
File: Monkey paw (Boob job).jpg (421 KB, 1467x2144)
421 KB JPG
>>
>>110109150
Skin parts are really a luxury in a devastated city.
>>
File: vtuber.jpg (275 KB, 768x1024)
275 KB JPG
>>
File: cover.png (3.78 MB, 1536x2244)
3.78 MB PNG
>>110108296
It will be a slop anime model like pic related, they call anime to whatever shit
>>
File: moss-image_00307_.png (1.39 MB, 1184x880)
1.39 MB PNG
>>
>>110108616
You have no idea what you're talking about, novelai has the best dataset we've ever seen. The fact that it trained on e621 only compounded to improve on its quality.

Go back to genning your 1 girls standing in the blandests generic artstyle possible while I gen 5 on 1 horse gangbangs of your favourite waifu.
>>
>>110109586
While you pay each month 25 bucks like a cuck and even then have limits on it, lmao.

People that don't support open source are peak cattle
>>
Its always touchless virgins who care about genning gacha girls doing vanilla shit like showing their socks or whatever. Local is held back by the mediocre.
>>
>>110108296
They will likely produce a benchmark overfited poorly trained, gpt slopped anime model whose flawed dataset will be difficult for local finetuners to correct afterward.
This local after all, we ride on hype.
>>
>>110109601
Whats funny is that I also probably have a better GPU than you.
>>
>>110109627
And what are you so proud of? That you're a better slave to the system for me? Congrats I guess. But it doesn't matter. I fikght for the freedom from our oppressors. Open source frees us all from the shackles of exploitation, you're nothing more than a cattle for the corpos to exploit if you pay a single fucking dime to those fucking jews.
>>
CFG 3 is still ideal
>>
>>110109627
Are you legit retarded. Having a GPU or not doesn't relate with supporting close source trash, open source doesn't mean exclusively local
>>
>>110109604
If you wanna see some more decent sloppers you have to branch out to X and tiktok
>>
>>110109651
Go check if NAI recharged your stamina or w/e arbitrary limits they put on your meme subscription
>>
ani please save us from the saas paypigs and cumfart shills
>>
>>110109651
You dont fight for nothing, you consume whatever slop you're given for free, and on top of that give atrocious opinions online.
>>
>>110109651
Wake me up when you produce your own "open source" GPU.
Until then you will never be free.
In fact nvidia has been fucking you in the ass raw for years now, artificially limiting VRAM amount on goycattle cards so they can sell enterprise grade cards for $30k a piece.
>>
>SaaS inbred replying to a bait
the absolute state of these paypig retards, insane.
>>
>i-it was bait even tho most local lolcows unironically believe this
riiiiiight
>>
Animefags are going to have to swallow their pride and pay $25. It sucks that something like NAI doesn't exist locally, but there's no point in you and your rtx3060 alone trying to rescue a ship that's sinking no matter what you do.
>>
is it the same guy that trolled like this on /h/?
>>
>>110068160
>>110068268
>>110068281
>>110068294
>>110068331
These look amazing. If you're still around I'd like a catbox.
>>
>>110109729
Trani please stop talking about yourself in third person and hang yourself
>>
>>110109810
They could've had a good local model if they stopped listening to the majority, because the majority is always wrong. Artist styles are more important than the gacha girl of the month, e621 can help the model learn more complex poses, etc.
>>
>>110109814
qrd?
>>
>>110109601
>>110109651
Why the seethe? Did the trash not have a good lora for you today?
>>
why did the op stop using the script to make the collage? or is it a different guy making the threads?
>>
>>110109841
Basically same shitpost patterns of baiting people with "local is shit, saas is superior" and then it goes back and forth with anons (or him replying to himself) calling each others paypigs, freetards etc, the wording and behavior feel almost identical to shit that was going on in hdg around 1-2 years ago
>>
>>110109789
Being open source means it's for everyone, means everyone can develop it, acess the technology and if they don't have a GPU then several businesses can host it like OpenRouter does and compete against each other in price.

Meaning everyone benefits since technology moves faster and is cheaper for everyone, local or hosted on cloud.

You are literally the target audience of these closed SaaS because you are a NIGGER.
>>
>>110109867
Seriously? Was NAI competitive two years ago, pre v4.5? But locally we had Animagazine and Noob.
>>
>>110109877
NAI released anime SDXL finetune before anyone else (naiv3). For a couple of months it was unquestionably superior to local 1.5 models, and after pony release it was competitive.
>>
>>110109870
>means it's for everyone, means everyone can develop it,
Being open source means the barrier to entry is low, it says nothing about whether people actually organize effort efficiently.
In practice you get fragmentation of random no talent sloppers dropping half baked models with no direction and then nothing comes of it.
>>
>>110109910
And all that disorganization >>110108296
is exactly what keeps the SaaS companies in business.
>>
I hate that novelai is saas but they walk the talk, they arent even in the same competition as local, its only the "in the dark" retards who never used it who think otherwise, which means local has to get their shit together.
>>
File: mike.jpg (395 KB, 896x1152)
395 KB JPG
barely any pictures in this thread
curious
>>
>>110109963
Always that way when people are shilling NAI. Lots of talk how great it is, with barely any impressive examples.
>>
>>110109997
the guy who shills it never used it in his life
its trolling
>>
>>110109946
>local has to get their shit together.
It won't happen, since those who tried to save local anime are less than 10 people.

LAX
Xinsir
Comfy's grant to Anima
Chenkin
Kohaku
Kohia
Animagine
Illustrious
Astralite

Everyone else just makes garbage and is a clout chaser, for example all the current noobcoders are useless clout chasers.
>>
Yeah, name drop more diffusion e celebs, please.
>>
>>110109910
Except that's not the case because NoveAI was built on top of open source technology, as most SaaS do.
You understand that if we didn't have open source models most of this and the entire hobby would't exist right? It would be at most 3 American companies milking turbo retards and pricing it as they please

Sure it tends to be less organized, but it can accomplish massively. All the relevant tools in the space are open source. Look at Invoke, Comfy, or how Anima came to be; they used a business model where it's open source and they charge for hosting.
NovelAI could be doing that and still make bank, since most people don't have a good GPU, yet still are greedy and hate open sourceprogress.
The sole fact that Anima exists is another reason why NovelAI is not charging you 50 bucks instead.
You are in /g/ dont be so ignorant aboug software nigger
>>
File: Moss gen.jpg (227 KB, 896x1184)
227 KB JPG
>>110110113
Also a moss gen
>>
>>110110113
Real talk, I'd rather pay novelai whatever than tard wrangle anima, its like not even comparable. Especially for professional use.
>>
>>110110128
Chinese cartoons are professional work to you?
>>
>>110110113
>NovelAI could be doing that and still make bank
They could do that, but they sure wouldn't be making bank, don't kid yourself.
Not because your statement about most people not having a good GPU is false, but because they would be far from sole hosters of the model like they are now.
>>
File: 5.jpg (101 KB, 720x570)
101 KB JPG
>>110108296
>some ai generared data slipped throught trainig teheee~
They are actively benchamrking using GPT anime slop for the dataset, it's over.
Also, they want to make an anime model and don't know the lore behind it, not even NoobAI.
>>
>>110110128
you WILL pay whatever they want thats how it works. Glad you're happy about it
>>
>>110110166
This is not the win for local you think it is bro
>>
>>110110150
I make money with anime bestiality, most of my customers are japanese
>>
With reverse engineering AI this is the future of the last closed source businesses. Just staying in the cloud forever.
>>
>>110110182
Lol nice
>>
>>110110177
You're a complete inbred retard worthless to have a serious discution with, but at least someone else can read my prev posts
>>
File: images.jpg (20 KB, 513x597)
20 KB JPG
>>110110196
>>
keep paying cucky and remember to wait for your refill. Oh oops we increased your wait time, suck it up
>>
>>110110185
>>
>>110110124
Depressing, we could do that even with SDXL
>>
>>110110229
A novelai gen cannot be made locally, thats the thing you lolcows dont seem to understand.
>>
>>110110243
fucking kek
>>
>>110110243
Dayum, they pay CIA veterans so little these days, that they resort to leaks for bribes
>>
>>110109859
As you can see by the seething reply to my other post, it really is a retarded subhuman """dev"""
>>
https://huggingface.co/lodestones/Kromoss
why finetune an already nsfw model tho
>>
File: kay stomach ache 1.png (666 KB, 1344x1792)
666 KB PNG
Made this in NB2.1, use unlucid.ai's Smart AI tool to shorten the skirt, then img2img in Anima at low denoise, fix the details in Paint.NET, then inpaint in Anima again at low denoise and add speech bubbles made with https://tools.oyasumi-gamedev.com/fukidashi/ in Paint.NET
>>
>>110110942
>>
>>110109191
Lifelike texture is expensive!
>>
I'm going to try doing a lora for moss-image-sft with my danbooru 40k dataset using the following slopbot-recommended settings:
      train:
batch_size: 1
cache_text_embeddings: true
steps: 10000
gradient_accumulation: 1
train_unet: true
train_text_encoder: false
gradient_checkpointing: true
noise_scheduler: "flowmatch"
optimizer: "adamw8bit"
lr: 1e-4
lr_scheduler: constant
dtype: bf16
model:
name_or_path: "OpenMOSS-Team/MOSS-Image-SFT"
arch: "moss_image"
quantize: false
quantize_te: false
low_vram: false


I'm not sure how many steps it's going to actually need, but I'll be watching test prompts to see how anatomy or pose-adherence changes.
>>
>Astralite
LOL
>>
File: battle scarred.jpg (420 KB, 896x1216)
420 KB JPG
>>
>>110110907
Nice try Julien, derailing an anime thread with local model talk. Take that shit to /ldg/
>>
I've been using kohya-ss for the longest time, but they're slow. What's the trainer to use for Moss?
>>
>>110111333
>>110111164
https://github.com/KohakuBlueleaf/KohakuMossTrainer
>>
>>110111333
I'm using ai-toolkit. First time doping moss-image-sft though, so fingers crossed.
>>
File: Meishodoto-1.webm (3.85 MB, 592x880)
3.85 MB
3.85 MB WEBM
I wanna fuck Meisho's dotos
>>
>>110111348
>>110108077
reminder to add this to OP
>>
File: hatsune teto.png (746 KB, 1762x864)
746 KB PNG
Is this inpaint workflow based or cringe?
Always wanted to manually and visually pick the context the AI works with instead of guessing padding pixels each iteration. With these unrelated and buggy nodes I select the custom context I want the AI to see, then draw the mask, and so on.
>>
>>110111596
this,
also remove the """UI""" no one uses to free space in the OP
>>
who needs an UI I generate anime girls in my terminal
>>
>remove the UI some anon's been maintaining on his free time,
>leave the dead WAN 2.2 rentry up
peak thread priorities
>>
>>110111628
is rtx even a good upscaler for this stuff?
>>
>>110111798
Alternatives appart from seed vr?
>>
>>110111960
just anime focused upscale models? is there a comparison between them all for anime?
>>
>>
File: cropandstitch_test.png (997 KB, 1573x847)
997 KB PNG
>>110111628
crop and stitch is your friend
>>
>>110112544
except here I forgot to wire inpaint nodes so it was very rough i2i that gets bad on high denoise and if you add inpaint controlnet/anime_lllite it gets amazing
>>
>>110111628
>>110112544
>inpainting in ComfyUI
Has you ever encountered inpainting halos problem?
>>
>>110112440
>that toe cleavage
>that ultraslopped cheap painterly nano banana look
those fluxesque kind of API models, unless they're cheap or nsfw nsfl uncensored, there's no real usecase for them, even at a professional level.
>>
>>110112566
I prompted watercolor. It's also missing a few diamond-shaped waist cutouts (not prompted), I wonder if it would copy better if I cropped the ref sheet.
>>
moss-image-sft lora test images, training start:
https://files.catbox.moe/hpedug.jpg
https://files.catbox.moe/rlznln.jpg
https://files.catbox.moe/zx6hmj.jpg
https://files.catbox.moe/oinin3.jpg
https://files.catbox.moe/nucf6v.jpg

The idea is to watch them and see if anatomy or pose gets worse or better. The last one is just a control image (clothed anime girl). Of the initial bunch only one has retarded anatomy.
>>
>>110112564
no. from little inpainting I've done it masks pretty nicely with Anima, but again only if you add inpainting controlnet lllite thing otherwise on high denoise it just draws very different image on top
>>
>>110112604
>https://files.catbox.moe/rlznln.jpg
Whoops that one's retarded too, didn't notice the three feet.
>>
>>110112633
oinin3 got 3 legs
>>
>>
>>110112648
Haha yeah, that's moss for you. As for what the lora's going to do, my guess is just make it worse. Only krea2 held up to such a big dataset. But I'll find out for sure in a few days.
>>
>>110112564
I'm pretty slop/newbie tier blind myself when it comes to spot inpainting halos, but I'll also use somewhere between 32 and 64px of blur mask, like I showed in my screenshot.
>>
inpainting is a fundamentally flawed concept
just generate what you want directly
>>
>>110112604
okay for now only the ToT gen is good
>>
>>110112707
That was my mentality for two years
Realizing you can approach it from both directions makes you a better equipped slopper, althoughbeithapsplaps
>>
>>110112707
would help if models can do what you want at all times
>>
>>110112544
But can you manually select the context? Doesn't seem like it.
>>110112707
It's fun if you have a good gpu
>>
>>110112773
it reads mask and creates rectangle around it according to parameters, adding padding and all that. and you can add resolution selector to output_target_width/height to select on what resolution inpainting sampling happens
>>
>>110112707
>He's never edited images
>any images
filtered
>>
>>110111628
Use Invoke, it's made for that use case.
>>
File: ComfyUI_temp_dekyx_00054_.png (2.31 MB, 1024x1536)
2.31 MB PNG
>>110109822
https://civitai.red/models/2877827/magic-knight-rayearth-style-clamp-anima
https://civitai.red/models/2560956/anima-element-7-model-suite?modelVersionId=2966419
>>
File: Pasted image (399).png (1.96 MB, 1120x1440)
1.96 MB PNG
>>
>>110113234
Thank you. A catbox for the one with the grey hair and the one with the pink hair would be more helpful though. I want to replicate the backgrounds and some of the clothes too.
>>
whats the best workflow for inpainting anima? In the past when I tried it sucked
>>
>>110113284
cute feet
>>
>>110112686
wonderful art
>>
>>110113392
In the present it also sucks
>>
>>110113392
Well, there is anima edit: https://civitai.red/models/2650553/anima-edit
>>
>been generating pics of me an my ex hugging and cuddling on the bed
>somehow the ai keeps on making me fat no matter how i describe myself, despite not even being fat
sigh...
>>
>>110113666
"skinny" not working?
>>
>>110113726
how did i not think of that
thanks anon
>>
Kinda off tipic, but what do you guys thinks is worth more as a side hustle, AI nsfw genning or 3D nsfw? I'm trying to decide which one to focus the most on. I had planned that if I was doing ai gens I would do doujins or something to sell, but idk. With 3D I would actually be limited in characters to pose since I depend on others right? Like I would just have to pick whatevr are available. I do like both, but I struggle seeing which one I would like the most to focus on so any opinions are welcomed
>>
>>110113824
Pose 3d model to feed into ai
>>
File: very fast horse.jpg (573 KB, 832x1216)
573 KB JPG
>>
>>110113906
>very fast
Randomized
>>
https://github.com/Rkkss/ComfyUI-Spectrum-MOSS
>>
>>110113955
did you just call my wife randomized
>>
>>110114104
I called her cute and her speed questionable
>>
File: ComfyUI_25059_.png (3.02 MB, 1680x2160)
3.02 MB PNG
>blows smoke on your face
what's with all this moss talk?
>>
>>110114028
>https://github.com/Rkkss/ComfyUI-Spectrum-MOSS
interesting
>>110114373
nice
>>110113565
undercooked
>>
File: ComfyUI_25065_.png (2.34 MB, 1680x2160)
2.34 MB PNG
>>110114428
its not nice, its too animooo
>>
File: ComfyUI_25105_.png (3.2 MB, 1680x2160)
3.2 MB PNG
>MOSS: https://huggingface.co/Comfy-Org/MOSS-Image/tree/main/diffusion_models
lots of shitpost potential, too bad I won't be around to spam the thread with it
>>
File: YOU.jpg (359 KB, 896x1344)
359 KB JPG
>>
>>110114581
Where you going papa?
>>
taking smallest int8 brainrot convrot turbo version of anima and plugging in loras still produces good enough for me results
are there any other models that generate fine after getting smol? never seen int8 turbo of illustrious models are those even exist or some architecture limitation
>>
File: ComfyUI_25117_.png (3.03 MB, 2160x1760)
3.03 MB PNG
>>110114830
overseas deployment
>>
File: ComfyUI_25149_.png (2.56 MB, 2160x1296)
2.56 MB PNG
>>110115139
anima is tiny, what sorta device you trying to run it on??
is there a kissing lora?
>>
>>110115163
Kill yourself
>>
>>110115163
4GB vram, I really thought it was over for me few years ago when SDXL took off and I only ever touched cloud image generation since then until I stumbled into anima that knows so much artists and fits on my laptop
>>
File: ComfyUI_25153_.png (1.92 MB, 2160x867)
1.92 MB PNG
>>110115189
makes sense, base anima uses 8.6GB of vram or so, but I imagine they made sure anima could run properly on 8GB cards as those are quite popular
>>
File: ComfyUI_25067_.png (3.66 MB, 1680x2160)
3.66 MB PNG
>>110115189
tho anima is quite slow, compared to krea2 anyways, so I wonder how long it takes for you to gen stuff at a given resolution?
>>
File: animaINT8.png (416 KB, 576x768)
416 KB PNG
>>110115287
rough, 1024x1536 took 129 seconds, 512x512 <20sec, some 512x768 ~20sec, this one 33sec
>>
File: meat bag.jpg (209 KB, 896x1184)
209 KB JPG
>>
>>110115529
>dinner for the whole tribe.jpg
>>
File: moss-image_00785_.png (1.14 MB, 1024x1024)
1.14 MB PNG
>>
File: ComfyUI_temp_biuge_00005_.png (2.62 MB, 1024x1536)
2.62 MB PNG
>>
File: 1775654142041728.png (558 KB, 640x480)
558 KB PNG
>>110114028
>run without node
Prompt executed in 318.58 seconds
>run with node
Prompt executed in 322.46 seconds
>>
File: Vendettive crack addict.jpg (190 KB, 832x1216)
190 KB JPG
>Bros waiting up to 6m per slop
I'd loan you some of my zimzam if I could
>>
I miss rocketnon
>>
File: Miyabi-1.webm (3.43 MB, 592x880)
3.43 MB
3.43 MB WEBM
I'm running out of girls to slop, guys, throw some suggestions
>>
File: ComfyUI_25184_.png (2.15 MB, 1472x1728)
2.15 MB PNG
>>110116423
slop some guys? what you need anyways, genned pics that you can turn to video?
>>
>>110116447

nah video is the boring part desu, more like girls to make gens for
>>
>>110115163
>>110115240
>>110116447
styles are pretty neat, catbox pls
>>
File: ComfyUI_17381_.png (1.1 MB, 1216x1536)
1.1 MB PNG
>>110116460
shermie from king of fighters, she's a cutie, ever did motoko kusanagi? how about violet evergarden?

>>110116479
can't share a catbox, too many chinese custom nodes
how's your workflow anyways send a screenshot, I can guide it to gen something similar
>>
>>110116502

I like the Motoko idea, thanks
>>
why is moss getting shilled? it's just another 3DPD model like krea that can't do good anime gens. just keep using anima
>>
Moss is uncensored and if you like certain anime styles it can do them. Plus they're gonna make an anime version
>>
>>110115240
why are they trembling
>>
Here's my danbooru 1024 explicit/questionable lora for Moss Image SFT:

https://huggingface.co/quarterturn/moss_danbooru40k_v1

Hard to tell if it really workeds out. Moss shits the bed in terms of anatomy all on its own, so I went by improvement in anatomy and pose vs number of fucked-up test images. Seemed like step 7000 was the best balance.
You can get a shitty Kuroki Tomoko merely by using "kuroki_tomoko" and her state of dress.
>>
>>110116423
smol girls wearing angel's (kof) obiviously oversized outfit
>>
File: ComfyUI_25063_.png (3.65 MB, 1680x2160)
3.65 MB PNG
>>110116902
laser printer is shocking them thru capacitive coupling, totally legal activities

>>110117113
what's this, the KoF fan meet-up thread?
>>
File: ComfyUI_25272_.png (1.7 MB, 1344x1728)
1.7 MB PNG
>>110117141
adding an eye color for the midget was a bad idea
>>
File: nb outpaint.jpg (725 KB, 1024x1024)
725 KB JPG
>>110117141
sure
>>
I have good results with 50 steps on mosd and using linear_quadratic
>>
File: ComfyUI_25275_.png (3.38 MB, 1680x2160)
3.38 MB PNG
>>110117180
is that a brother brand laser printer? what's AI trying to imply here??
>>
>>110117180
WTF, got gymnastics going on while shota gets cucked
https://litter.catbox.moe/ei0l60j5cffx0v7x.png
>>
File: ComfyUI_25281_.png (3.04 MB, 1680x2160)
3.04 MB PNG
>>110117227
anon c'mon don't be a noob, all you had to add was "mother and son" and "vaginal" and maybe "onee-shota" and it would work the way you wanted
https://files.catbox.moe/okhf9y.png
bonus https://files.catbox.moe/xxsgqk.png
>>
>>110117270
>Image cannot be displayed because it contains error.
won't load either in browser or downloaded
>>
>>110117289
https://imgpile.com/m/GbEyDiM
https://imgpile.com/m/mFsxHUA
I hope their servers aren't located in commie europe
>>
>>110117319
ok that kinda makes now, bro got mad lean
>>
I've been getting into NovelAI recently
How do I figure out art styles? Everything I gen looks hideous no matter what I do
>>
File: 1778323070594180.jpg (1.69 MB, 1792x2400)
1.69 MB JPG
>>
>medium complexity
>0.4::artist:xxx, artist:yyy, artist:zzz ::
manually adjust weighty artists down and low image count artists up, so like
>0.4::artist:xxx ::, 0.6::artist:yyy ::, artist:zzz
Works backwards from usual so last artist listed gets the lion's share
Complexity goes from low (mspaint)/medium (flex)/high (fights you)/ultra (LOUD NOISES)
>>
>>110117720
>>110117617
derp
>>
File: 1786732696742624.png (1.98 MB, 1792x2304)
1.98 MB PNG
anima enjoys turning my girls into gyarus for some reason
>>
>>110117797
rub some sunscreen on that poor thing
>>
>>110117797
>dark skin, dark-skinned female
in negs?
>>
File: 1762930008836816.png (1.05 MB, 896x1216)
1.05 MB PNG
what models are you guys using
is it still anima, or is there a new mix? I'm always behind on this shit
>>
>>110115240
>makes sense, base anima uses 8.6GB of vram or so
huh? it uses 4.9GB for me
>>
>>110118554
Check again during generation when everything's loaded in.
>>
File: 1472666433520.jpg (33 KB, 232x173)
33 KB JPG
>>110110165
Why the hell do people upload sloppa to boorus anyway? It's clear it's going to poison the dataset down the line. Self-sabotage much?
>>
File: egyptian.jpg (475 KB, 832x1216)
475 KB JPG
>>
>>110116923
This general never appreciates anything, so take this as a rare thank you.
>>
File: Moss-Image_00004_.png (1.16 MB, 1024x1024)
1.16 MB PNG
Okay shit moss image looks promising. A reasonable size for danbooru finetuning too.
>>
>>110119199
I'll take a quiet general over a schizofest any day.
>>
>>110119381
>schizofest
At the risk of sounding like a humongous faggot, this is exactly why I don't tend to post in here. Well, that and my autistic dedication to one series that's better discussed elsewhere anyway.
>>
File: file.jpg (572 KB, 1536x1536)
572 KB JPG
>>
>>110119397
but...Touhou..?you can do that here too?
>>
File: moss-image_00104_.png (2.48 MB, 1920x1088)
2.48 MB PNG
>>
>>110116923
>Moss shits the bed in terms of anatomy
Same as Flux2Klein when training for anime. It's like the model barely learned anything, SD1.5 tier quality.
Spoiler: Flux2VAE might be poisoned and could be the anti local villain that has been ruining quietly all those base models.
>>
>>110119397
/adt/>>>>>>>>>>>>>>>>/2huai/>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>/naptroons/, just saying.
>>
>>110120004
Let's not start retarded shit like this.
>>
File: 00005-1368371549.png (1.74 MB, 832x1216)
1.74 MB PNG
>>
>>110120004
Yes please, thread wars!
>>
>>110119330
Kino
>>
>>110119009
Luddites flood boorus with AI slop on purpose, as a kind of sabotage.
>>
>>110118181
Anima 3.8b
>>
>>110117617
Share metadata. Also, why did you subscribe to NAI without first testing yourself fryed local models via Civit or SeaArt(both in the OP tho)?



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.