[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]


Janitor acceptance emails will be sent out over the coming weeks. Make sure to check your spam folder!


[Advertise on 4chan]


File: 1781082051911368.png (2.73 MB, 1792x2304)
2.73 MB PNG
Previous: >>109017507
NEW: https://huggingface.co/circlestone-labs/Anima
Anima1.0 is out

>UIs
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

>How to generate anime:
image : https://rentry.org/comfyui_guide_1girl
video: https://rentry.org/wan22ldgguide

>Style Explorers
Anima: https://anima.mooshieblob.com/
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.jpg >>106520607

>/adt/ logo:
https://files.catbox.moe/pakfgt.png

>keywords(test):
yeah no we are not a TikTok video
>>
thanks for bakering
>>
File: Sky_00068_.png (1.12 MB, 1536x1024)
1.12 MB PNG
>>
File: cui_lawine_0002.jpg (1.1 MB, 1408x2048)
1.1 MB JPG
>>
File: lass_00015_.png (887 KB, 896x1152)
887 KB PNG
>>
>>109047307
>wifey made the OP
sweet
>>
Can i request some obese pokegirl gens?
>>
File: skyla_00062_.png (1.63 MB, 1536x1536)
1.63 MB PNG
>>
File: CUNFY_00523_.jpg (2.39 MB, 1664x2432)
2.39 MB JPG
>>
File: Nemona_00060_.png (791 KB, 896x1152)
791 KB PNG
>>
>109047508
>109047442
>109047441
>109047410
>109047388
You have to change the bait at some point if you want me to engage with your shitstorm of bad quality slop
>>
>>109047504
>>109047442
Hot (the weather)
>>109047401
I think the plants look nice
>>
File: may-onsen.png (1.16 MB, 1152x896)
1.16 MB PNG
>>
File: 1463367786739.png (17 KB, 882x758)
17 KB PNG
>had bbox_fill on for several days without realizing it in my mask detailing workflow
>mfw
>>
>>109047388
going to need the view rotated 90 degrees to the right for this one.
>>
>>109047603
>I think the plants look nice
But I'm _cui!, the plants weren't my focus, my artistic vision goes deeper than that, what about the multi layered composition? or the character placement, the houses in the background, the vines making sense, and the autumn tree?? y-yeah I forgot the character's shadows and shaders b-but that was an artistic choice!! not because I'm a lazy slopper at my core and i want to prove the contrary or anything!!
>>
What was the bigger step up for anime gen, Illustrious or Anima?
>>
File: Mutsuki candy.png (2.03 MB, 1536x1536)
2.03 MB PNG
>>
>>109047796
Pony
>>
>>109047796
NovelAI
>>
>>109047798
I hate children
>>109047826
t. paid shill
>>
>>109047622

Everytime I see May I’m reminded she’s one of the first girls I did the deed for. She’s home to me, and I need to replace her with an actual woman eventually.
>>
>>109047671
You seem to be in this thread most of the day.
>>
File: cui_may_0001.jpg (1.13 MB, 2048x1600)
1.13 MB JPG
>>109047441
Fat yes obese no
>>
>>109047950
>Bush
yes
But why is she in the lake though?
>>
>>109047967
>look mum, I put an anime character in front of an MTG background, I am so introspective, so unique so _cui!
>>
>>109047991
>>109047950
neat xd
I've been using a few pictures to img2img but this works too
>>
>>109047926
yes, i'm /adt/ schizo and the baker
>>
File: ComfyUI_10495_.png (3.53 MB, 1800x1536)
3.53 MB PNG
>>109047967
It's meant to be rustic and low-class

>>109047991
Gandalf...
>>
File: Nyo.jpg (1.27 MB, 1420x1915)
1.27 MB JPG
>>109048013
That's not true, and you wouldn't pick that picture
>>109048169
I thought she was hunting psyducks
>>
>>109048238
ah, fml my fking screen again
>>
File: eliisa plains.png (1.2 MB, 1200x900)
1.2 MB PNG
>>109047991
Use cases for being unique?
>>
>>109048299
Don't know if this is you skizo kun but this is lovely innit'?
I'm not seeing the problem here
>>
File: ComfyUI_169297_.jpg (513 KB, 1080x1920)
513 KB JPG
>>
File: image.jpg (959 KB, 2120x1500)
959 KB JPG
>>109048299
>>109047991
Thread theme: Magic The Gathering lands + anime girls
>>
File: ComfyUI_03732_.jpg (317 KB, 1344x1344)
317 KB JPG
is it ethical to use anime diffusion models?
>>
>>109048702
what do you mean
>>
File: swsh_00037_.png (1.11 MB, 1536x1024)
1.11 MB PNG
>>
>>109048708
they were trained on the hard work of thousand artists without paying a single cent
>>
File: Nom.jpg (1.42 MB, 1464x1920)
1.42 MB JPG
>>109048554
Why ty
>>
yeah nah im out
>>
>>109048741
why? stay here
>>
>>109048630
that's a pretty sharp card and image
>>
>>109048782
>>109048735
>>
>>109048630
midget legs
>>
>>109047504
that's not AI genned
>>
File: 0.png (1.59 MB, 1344x936)
1.59 MB PNG
>>109048630
well that was fun
>>
Is it worth paying for video animation? How are their filters currently compared to others? Nothing explicit.
>>
>>109048901
you can run video models locally btw
e.g. WanVideo
>>
>>109048932
Why not LTX 2.3?
>>
>>109048882
this upscaled it's kino
>>
File: sw-lass1_00008_.png (1.37 MB, 1536x1024)
1.37 MB PNG
>>
File: 0.png (3.3 MB, 1824x1824)
3.3 MB PNG
>>109048970
I'm assuming that's a request with no punctuation. I actually re-downscaled it after i2i upscaling to match the original image resolution.
>>
File: WolfMTG.jpg (900 KB, 1500x1032)
900 KB JPG
>>109048882
>>109048630
I'm pretty shit at this
>>
>>109048990
neat
>>
not enough migu
>>
File: Fops.jpg (2.01 MB, 1600x2200)
2.01 MB JPG
>>
>>109028954
Can I reroll for the third kaleido?
>>
File: lorelei.png (2.89 MB, 1856x1536)
2.89 MB PNG
lorelei was the hottest one when i was a kid
>>
>>
>>109049396
Pokemom
>>
>>109047401
With regard to the number of elements in an image, sometimes less is more.
>>
File: takagaki mariya.png (2.64 MB, 1600x1600)
2.64 MB PNG
Happy Birthday, Kaede
>>
File: katsu2.jpg (854 KB, 1632x1264)
854 KB JPG
>>
>>109049546
You'll have to show me more to help me remember
>>109049574
She's going to eat him isn't she?
>>
File: CUNFY_00524_.jpg (2.36 MB, 1664x2432)
2.36 MB JPG
>>
File: cui__17427_.png (2.07 MB, 1024x1536)
2.07 MB PNG
>>
File: unamused.jpg (408 KB, 1344x1344)
408 KB JPG
>>109049287
if you're lucky
>>
File: GSC_AceTrainer_00009_.png (1.18 MB, 896x1152)
1.18 MB PNG
>>
Thoughts on my Skyla video gen?

https://litter.catbox.moe/m56xly.webm
>>
>>109050426
looks great! have you posted your gens in /ldg/? i think they'd love them!
>>
>>109050426
is this bw 1 or bw 2?
>>
>>109050502
Is there a difference?
>>
>>109050514
one is a minor the other isn't
>>
File: 12.png (101 KB, 936x912)
101 KB PNG
https://x.com/sesuji07/media
>>
can someone explain why Anima breaks above 1536x?? the noise gets blurry every img2img iteration.
>>
>>109051659
You have 95k followers, can you please shill /adt/ to your community, something like "very interesting new anime general on 4chan, cool anons."
>>
>>109052119
You're not too bright, huh?
>>
>>109050306
Good rolls. Thanks.
>>
>>109051659
nice. share the lora?
>>
File: laplus.png (24 KB, 208x152)
24 KB PNG
>>109052678
@sesujistyle, pixel art
Resize to 12.5% nearest neighbor for the true pixel size.
https://files.catbox.moe/y6w96i.png
https://files.catbox.moe/zhr1f8.png
https://files.catbox.moe/5l3283.safetensors
>>
>>109048702
>>109048735
Using the model for funsies I think is not unethical.
However, the model makers did do something unethical by using the training data knowing full well a large portion of the artists would not agree to it.
>>
File: giwtm.png (3.44 MB, 1856x1536)
3.44 MB PNG
>>109049396
>>
File: demon.png (34 KB, 336x432)
34 KB PNG
>>109052724
cool
>>
>>
File: ComfyUI_01061_.jpg (112 KB, 832x1216)
112 KB JPG
>>109052724
nice
>>
Does NegPiP work with Anima? How would the correct syntax look if I want to move a negative prompt into the positive prompt using negative weights?
>>
>>109054121
it should
syntax is (the tag:-1)
>>
>>109052724
ty for sharing
>>
>>109054227
and if it's more than one tag? (worst quality:-1.3), (old:-1.3)?
>>
>>109054326
the advice seems to differ depending on model
https://github.com/pamparamm/ComfyUI-ppm#clipnegpip
>>
Suggest me some NoobAI shitmix
>>
>>109054326
(tag1, tag2: -1)
>>
File: image.jpg (153 KB, 1072x1072)
153 KB JPG
are birds considered anime?
>>
Anima appears to be overfitted. When I try to gen a character wearing similar clothing and use tags that overlap with the character's outfit, the model tends to associate those clothing features with the original character and as a result, elements of the original character's design bleed into the new character
>>
>>109054610
post the prompt and we'll see
>>
>>109054610
you don't know what overfitted means
>>
>>109054590
Birds aren't real
>>
Can you guys explain to me what NoobAI is and how it differs from Illustrious? The description summary doesn't make it clear, and the rentry guide only says it's something "different"
>>
File: content.jpg (322 KB, 1672x940)
322 KB JPG
>>109054978
Illustrious is the actual base model developed by OnomaAI. It uses standard eps prediction, meaning it works perfectly with standard SDXL tools, ControlNets, and has a massive ecosystem of community merges on Civitai. The catch is that it is a clean base, so you will need to load external LoRAs if you want very specific niche characters or distinct artist styles.NoobAI is a massive fine-tuned fork built on top of an early Illustrious base by Laxhar. It uses v-prediction instead, which fixes the washed-out look of typical SDXL models and gives you true deep blacks and pure white backgrounds. The main selling point is the dataset. They injected 20 million more images into it, including a ton of e621 tags, so it has native support for furry/anthro stuff and a huge built-in library of artist styles without needing to clutter your prompt with LoRAs.As far as compatibility goes, you can use Illustrious LoRAs on NoobAI and they usually work fine, but NoobAI LoRAs will completely break if you try to run them on an Illustrious base.Basically, stick to Illustrious if you want a reliable base for training or want to use standard community mixes and switch to NoobAI if you want better native style variety, furry content, or deep contrast control out of the box.
>>
>>109054717
and you don't know what love means
>>
File: Fan.jpg (1.96 MB, 2000x3485)
1.96 MB JPG
>>
File: GSC_AceTrainer00020_.png (1.13 MB, 896x1152)
1.13 MB PNG
>>
Can i request some pokemon gens?
>>
>>109054961
Neither is your waifu
>>
...and then i gooned
>>
File: migu.jpg (533 KB, 1152x1536)
533 KB JPG
>>109047322
>>
Ogipote status?
>>
File: GSC_AceTrainer_00035_.png (1.13 MB, 896x1152)
1.13 MB PNG
>>109056079
>>
>>109056755
Just wait for Ogipote shizo
>>
It makes me want to vomit how Anima base alone can give you shiny slop gens very easily.
>>
File: 00246-3069604973.png (3.14 MB, 1792x2304)
3.14 MB PNG
>>
File: 00023-681777161.png (2.48 MB, 1408x1984)
2.48 MB PNG
>>
>>109057463
So you've also fallen to the crop top school uniform over swimsuit illness. This is sick. Nobody would wear something so perverted in school!!
>>
File: 00248-1235786348.png (2.29 MB, 2240x1152)
2.29 MB PNG
>>109057609
that's just a remake of an old gen that I couldn't quite achieve on SDXL
>>
>>109057643
What was hard to achieve? Classroom being Euclidean and making sense or something else?
Funny how the foot is out focus, your POV guy is probably more hungry for the popsicle. "Hey, I want that too, not your damn foot". The fool.
>>
File: 00030-3705622493.png (2.09 MB, 1600x1600)
2.09 MB PNG
>>109057676
the leg angle was the main problem, tried many combinations of tags but none of them worked like I wanted
>>
>>109057742
That's what you get for trying to t2i instead of CN-stealing compositions from artists like usual ehehheheh
>>
>>109057818
>CN-stealing compositions from artists like usual ehehheheh
opinionated opinion
>>
>>109057869
Every day I find another one, you know...
https://litter.catbox.moe/htjhql.png
https://danbooru.donmai.us/posts/7134355
I'm sure eventually they will ALL end up being CNed!!
>>
File: 00106-2508403136.png (2.47 MB, 2112x1056)
2.47 MB PNG
>>109057818
i don't have any meme reaction made with anima yet, can you believe it?
>>
File: 00111-4273471861_result.png (755 KB, 1664x2304)
755 KB PNG
>>109057912
holy shit, is this how are you using your time now? is everything at home ok?
>>
>>109057929
I am going to make a website called noodlegate where I will expose your crimes and then all your fans will unsubscribe from you, just you watch •̀﹏•̀
>>
Friendly reminder that Anima's problems largely come down to Russell captioning the dataset with a local (eww) Gemma (eww) instead of supperior WD tagger, using a model that has no clue about anime to caption anime training data is a great way to get garbage out.
>>
File: 00003-646748459.png (2.39 MB, 1984x1408)
2.39 MB PNG
>>109057973
geeeeeeeg alright, better than shitposting and harassing manchildren all day I guess
>>
>>109057913
this looks very gay
>>
>>109058180
I don't think it's inherently gay, but something about wanting to replicate that danbooru artist aesthetic specifically rubs me the wrong way, like bro you are an AI genfag, own it completely instead of trying to cosplay as a human artist.
>>
File: 00249-3448652707.png (3.33 MB, 1792x2304)
3.33 MB PNG
>>
whats the negative meta for anima?!
>>
>>109058483
>>109058498 here in /adt/ every day is illustrious day fuck off
>>
>>109057463
https://animefeet.blogspot.com
Do you even foot fetish? Every toe there besides the big one is just a normal toe with shorter phalanges...
>>
>>109058562
wtf
t
f
>>
>>109057463
Also the arch and the heel pad are out of proportion.
>>
>>109058562
some site you got there buddy
>>
File: CUNFY_00525_.jpg (2.33 MB, 1664x2432)
2.33 MB JPG
>>
>>109057913
Imagine these long ass legs wrapping all around you like snakes. They aren't exactly canon, but I'll take them, sure.
>>
File: evelyn casual.png (2.24 MB, 1112x1920)
2.24 MB PNG
>>
>>109058015
>implying first place can ever make merch that good
Cute ARIA fumos tho. [spoiler]IA's pantsu is black tho.[/spoiler]
>>
>>109059810
I tried to make it black but at the last step it went back to white for some reason
You are lurking here again huh...
>>
>>109059821
>You are lurking here again huh...
Always lurking in them heres. Also
>>109057913
Those wings look a lot better than the original. Good job and thanks for more Zako.
>>
I also lurk all the time. I only talk when drunk lol
>>
Anyone here tried to use AI to create their own unique OC and then train Lora for it? How did it go? Can AI even make interesting character designs? How difficult was it to give the OC interesting details and having Lora learn those details?
>>
File: good morning.jpg (474 KB, 1152x1536)
474 KB JPG
>>
>>
>>
>>109060196
>>109060198
Are you using Illustrious or Anima? And why do you put franchise and meta tags first and end with 1girl? It would make more sense to start with 1girl, then name, franchise, meta, and appearance, establishing what the subject is first before describing it.
>>
>>109059864
You first have to create consistency between your dataset. Consistency of clothes between gens I could only do it with SaaS edit models
>>
>>109060337
I'm using this https://civitai.com/models/835655/illustrious-xl-personal-merge-noob-v-pred05-test-merge-updated
Also I'll try that
>>
>>109060342
>Consistency of clothes between gens I could only do it with SaaS edit models
Thats what I'm thinking. Which one model specifically?
>>
>>109047307
Does anyone have any good method of forcing the model to output characters default outfit, without describing it with generic tags? Like non-default outfits tagged as <character>_<skin>_<series> work fine, but at times <character>_<series> won't properly return the original outfit. Is there any tag like official_outfit or something that'd help with this?
>>
>>109060529
It does if I use LoRAs to do so without tags when using IL.
>>
>>109060554
I'm not sure I understand, you mean you use character LoRAs?
>>
>>109060563
I like taking the clothes off character LoRAs so I can give them to other character LoRAs.
>>
>>109060655
Yeah but my point is that models like Anima at times won't return the character dressed in their default outfit but some random clothes, and I would prefer to avoid having to describe said outfit using tags like black_shirt or using LoRAs just for clothes.
>>
>>109060529
Even if such tag existed. The issue is that if character doesn't have enough samples in the dataset, the model simply won't understand that the outfit is part of the character. To it, any clothing is more associated with generic pieces in other art than the character. Giving it the hints forces it to include the pieces, and because those pieces are associated with the character, they look like it's official outfit.

>>109060554
Thats why Loras work, they over-fit the character's description to include it's outfit.
>>
>>109060690
>Even if such tag existed. The issue is that if character doesn't have enough samples in the dataset, the model simply won't understand that the outfit is part of the character.
That's not exactly true, especially given it does the alternate outfit consistently well, despite it having way less images than the "default". 800 images on danbooru vs 1700. I'm pretty sure this is more caused by people drawing the character in non-official outfits, but it all going under the singular base tag
>>
File: Anima_00249_.png (2.06 MB, 1280x1280)
2.06 MB PNG
>>109060663
It depends on if they are known enough in the model; just like using WAI. Makoto works well and you can give her whatever clothes you want.
>>
File: Anima_00251_.png (2.64 MB, 1280x1280)
2.64 MB PNG
>>109060802
I gave it her normal outfit and it gave a pretty cool result.
>>
I'm drunk so I'm bad at logic but if you want characters that a lot of models can recognize:
https://huggingface.co/spaces/flagrantia/character_select_saa
>>
File: ComfyUI_09727_.png (3.49 MB, 2560x2560)
3.49 MB PNG
>>109060854
>masterpeice, best quality, solo, standing, aisaka taiga, angry, holding weapon, wooden sword
>>
>>109060663
>Anima at times won't return the character dressed in their default outfit
Anima is a shit sucessor tho
>>
>>109061091
It still does the good shit though.
(sometimes)
>>
>>
>>
File: GSC_AceTrainer_00044_.png (1.15 MB, 896x1152)
1.15 MB PNG
>>109056778
>>
>>109059654
nice, would you share a catbox?
>>
File: brand miku.jpg (428 KB, 1056x1728)
428 KB JPG
>>
>>109063172
it's an edit of a pic found on twitter
>>
>>
File: VampireHunter.jpg (1.15 MB, 3528x2352)
1.15 MB JPG
>>
File: lana_00021_.png (1.69 MB, 1536x1536)
1.69 MB PNG
>>
File: lyra_00001_.png (841 KB, 896x1152)
841 KB PNG
>>
>>109063339
>>109064887
>>109064897
can you make them fatties please?
>>
File: CUNFY_00527_.jpg (2.41 MB, 1664x2432)
2.41 MB JPG
>>
Is there anything besides tipo that can maybe fill out a prompt with related tags? Something like yandere+smile="shaded face, empty eyes, knife" since an llm seems like overkill when https://danbooru.donmai.us/related_tag exists.



[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.