File: anchy37.png (3.2 MB, 1152x1632)

3.2 MB PNG

>>106766008
ANCHOR:
>/adt/ pixiv:
BakerAnon: https://www.pixiv.net/en/users/110320313 (suspended sob)
Nekotwins: https://www.pixiv.net/en/users/93453238
*not found* https://www.pixiv.net/en/artworks/134385090
RSetsuu: https://www.pixiv.net/en/users/91156181
tamzaiy: https://www.pixiv.net/en/users/38224130
匂い: https://www.pixiv.net/en/users/76374114
Otto: https://www.pixiv.net/en/users/106407853
owphoenix: https://www.pixiv.net/en/users/1661492
https://rentry.org/foxlolianon
cornholioanon: https://www.pixiv.net/en/users/1035594

*Anchor your Pixiv and CivitAI here to add it in the next thread.

>/adt/ CivitAI
https://civitai.com/user/Nyan666:
https://civitai.com/models/1922023/obmirsak

>>106762385
REAL THREAD
>>106762385
REAL THREAD
>>106762385
REAL THREAD

>>106766116
no

File: 1732922709974761.jpg (921 KB, 1248x1824)

921 KB JPG

>>106763784
>>106763796
I never experimented with any sort of speed nodes extensively because all of them seem to nuke either quality or creativity from my experience and i'd rather just let my PC work a little longer.
These pictures of Palma from Choujin i posted in previous thread were all done with base noobai btw, not neta.

>no rocketnon in collage
GOOD.

>>106766520
>collage isnt animated
>he only posted animations

>>106764123
not according to the chomos here or the collagebakers apparently

>>106766565
>>106766565
Yeah, the collage ITT is missing a ton of people who made good stuff. You can tell it was rushed... but can't really complain since anon does it for free.

File: 4839393993.jpg (150 KB, 1080x1920)

150 KB JPG

WHERE ROCKETNON???

>>106766013
>(suspended sob)
The pixiv purge continues they know we can generate more ToTs than their servers can handle.

>>106766630
>does it for free
he does it for the (you)s, anon

>>106766565
forgetting or more like "purposely excluding people i don't like"?

File: Gens.jpg (15 KB, 344x97)

15 KB JPG

Why are you fags arguing about a collage when you could be posting gens

>>106766754
SEXO
>>106766741
Uhg,...

>>106766754
because half these threads is complaining about collage and the other half is people asking for metadata they will never use

>Not in the collage
It's over...

>>106766804
Catbox?
I WILL USE IT THIS TIME I SWEAR >>106766811

Inspecting old gens Found these kino tags

negative prompt: bad hands, extra fingers, too many fingers, mutated hands, ugly hands, deformed hands, malformed hands, worst quality hands, low quality hands, jpeg artifact hands

Remember when we were starting and blamed tags and embeddings for fucked up hands, feet, and limbs?

>>106767014
The SD 1.5 cope era
I remember it yes

>>106766937
https://files.catbox.moe/0e3d37.png

>>106766719
>>106766754
>>106766630
its clearly multiple repeated attempts to get posters who are actually good to leave

>>106766909
>>106766811
>>106763657
they cant even get jerked off/jerk themselves off normally

>>106766659
confirmed dead

One day shitmixes will be able to make a keyboard, an assault rifle, a spatula, or a sword.

>>106767167
We have various terminally online schizos who have been plaguing these threads for as long as they existed and this is exactly their goal.

If you want a sure spot in the collage, you must gen cunny

There's gold in my gen folders but I'm not sharing it with anyone.

>>106766804
Cute saber! <3

>>106767383
Pedotroon superiority complex many such cases...

>33 / 8 /
can someone explain why every thread devolves into the same arguments about cunny and collage instead of actual discussion
POST GENS NIGGA

Remember there's an Anistudio general >>106762385
So it's not necessary to include it in the OP anymore, just saying

>>106766741
BUILT FOR HEADPATS

>he doesn't have a folder called "definitely_will_use_later" with 10tb of nonies catbox

>>106767588
Why? Nobody even shares 'boxes.

>>106767797
>people asking for metadata they will never use
Aaaaiiiieeeee I feel personally attacked!

>>106767627
He doesn't know about the secret hands tag, oh no. Don't tell him :3

>>106767014
>>106767939
I slopped

>>106766008
>anistudio in the OP
Why is it that every time I ask someone to send me an .mp4 of AniStudio working, at least doing a simple txt2img function, I always end up insulted and called a schizo. Is it easier to ask for UFO footage than a video of AniStudio actually working?"

>>106767355
Yes but at the cost of forgetting what? This model has a capacity, to learn new things it has to forget other things

>>106766804
CUTE AND PLAPEABLE

>>106768062
Actually there is a secret Anicord where we're all genning kino 24/7

>>106768062
>open the anistudio link
>first thing you see is a screenshot of the UI working
are you retarded or something?

>>106768277
>screenshot

>>106768277
>works on my machine

>>106767355
And armor.

There's also this issue where if you ask for areola slip it often tries to gen two nipples.

>>106768062
Ani and the /sdg/ schizo are friends and they kept putting it in OP. We don't have enough active users in this thread to remove it like/ldg/ does

Tourist here
How many of you do this locally on your own GPUs versus using a paid online service?
Ideally I'd like to throw money at a service to generate NSFW, but I am always paranoid about privacy, plus I know some services have red lines you can't cross.

>>106769602
NovelAI it's you best bet for that also is SOTA in anime and NSFW

>>106769602
I use both
I guess NAI is up your alley then, I think it's the only weeb service allowing whatever NSFW you want, unless I'm wrong
I don't know how long it will last for them before processor payment or some new laws fuck them again though, they already got hit once, so it's not a bad idea to think about getting a good GPU somewhere along the lines

File: image.jpg (471 KB, 900x1496)

471 KB JPG

MiaoMiao Harem 1.75 Update Notes

 Core Update Highlights

    preserving realism. While maintaining a realistic texture, 1.75 has made subtle adjustments to the overall visual style to better match the early look of MiaoMiao Harem while

    Bug Fixes and Stability Enhancements : Fixed known bugs from v1.7 and improved overall stability and motion consistency .

    Continues to support LoRA training with a clean, style —ideal for customization and detail enhancement. consistent 

 Usage Tips

While 1.75 is a significant optimization update, it's not suitable for all creative scenarios. We recommend that users flexibly choose the version based on their needs; 1.0 Base, 1.6G, and 1.7 remain excellent choices.

While 1.75 offers significant improvements, it may not be the best fit for every creative workflow. We recommend choosing the version that best suits your needs—1.0 Base, 1.6G, and 1.7 remain excellent options

https://civitai.com/models/934764/miaomiao-harem

Model Output Parameters
Scheduler euler euler a
Steps 30
CFG Scale 3.8 ~ 6

Model Recommended Resolution 768X1280 896x1152

PVC Style(v1.2~v1.5)

miaoka\(pvc style\), miaoka\(figure\),3d,

Positive Prompt (default )

{Add Prompt here} ,masterpiece, best quality, absurdres, newest, very aesthetic, amazing quality,highres,sensitive, highres, ultra detailed, best anatomy, HDR, 8K, high detail RAW color art, high contrast, 

Negative Prompt

(bad hands,mutated hands and fingers:1.2),lowres,(bad),bad feet,text,error,fewer,extra,missing,worst quality,jpeg artifacts,low quality,watermark,unfinished,displeasing,oldest,early,signature,artistic error,username,scan,[abstract],english text,shiny hair,3d,realistic,

File: image.jpg (379 KB, 900x1158)

379 KB JPG

these are weak weeks for good quality checkpoints. I'm sharing new checkpoints that have relatively good scores.
Bismuth Illustrious Mix UPDATE

Recommended setitings:

Sampling method: Euler a, DPM++ 2M, DPM++ 2M SDE, DPM++ 2S a, DPM2 a, DEIS

Schedule type: Karras, SGM Uniform

Sampling steps: 20-50

CFG Scale: 3-7

Highres.fix + Adetailer for face are recommended

positive:

masterpiece, best quality, 

negative:

worst quality, low quality, lowres

>good
>mix
lol

>>106767167
This.
There's definitely at least one person in here who routinely attacks good genners, often times just straight up fabricating things to attack them with. Wouldn't be surprised if he bakes some threads.

>>106762436 (Cross-thread)
If you're still here, it was this post. Haven't tested it extensively enough for myself though.
>>106443690 (Dead)

Somehow, I was in the wrong thread. How which /adt/ in the archive has that Regional Prompter tutorial?

>>106769998
let me search

>>106769998
https://desuarchive.org/g/thread/106240788/
106261557

>>106769998
>>>106443690(Dead) (Dead)
Isn't that pretty much what you can do with Krita? You paint the structure, then clean up with details
Well, live painting would be the fastest way to do this but it changes quite a big chunk around, and bleed, sadly

>>106769959
Looking back at old threads, I saw great gens that were shared here. But generals naturally change as people come and go. It's just as important that we appreciate the current anons who are sharing their gens with us today.
Shout outs to >>106769251 and >>106768731 who have been here since the beginning.

>65 / 13
We're not very inspired today, are we?

File: NetaV4.png (700 KB, 740x736)

700 KB PNG

Neta Yume Lumina v4 preview ! It's happening!

>>106770301
Neat.

>>106770332
Neat Lumina

>>106770301
What's the threshold for a character and style appearing on NYL v3? Seems like it's only big characters/styles and cannot do a recent -100 pic on booru, can it?

>>106770375
>What's the threshold for a character and style appearing
what is the threshold?

>>106770375
Beats me, it doesn't know some big names like akihiko yoshida but can do several niche as fuck drawfags.

>>106770230
it's thursday, anon, we're busy

>>106770375
Not quite compared to Illustrious, base Neta is underdeveloped and Yume is doing all that work alone

need like, huge anime titties

>>106770391
catbox?

>>106770478
Oh shit, this is awesome please, collage maker, think of those kino anons too! Not only 1girl, youthful, cute

File: 1754141984453476.jpg (3.07 MB, 2432x1664)

3.07 MB JPG

>>106770558
The general needs more non-1girl gens

>>106770488
unfortunately won't be all that useful if you're after the style since I'm just spinning through real abstract dumb shit right now on top of doing a lot of post process.
<https://files.catbox.moe/c5ql48.png>
the lora is a low step count(512) test trained on a single image and the output was an 1:1 i2i to, effectively, do a light style transfer.
the original lowres it's based on is <https://files.catbox.moe/0x4cze.png>

>1boy, standing, looking at viewer

File: 03433-3672663510.png (3.63 MB, 1792x2304)

3.63 MB PNG

>>106770717
>lelouch vi britannia
kek

Stop being gay, thread.
Don't be gay.

>>106769673
>>106769705
Thanks, I'm genuinely impressed by the NAI results. Anyone know if the highest paid tier is worth it?

>>106770582
Oh I see, thanks though

>>106770829
cute boy :3

>>106770569
THIS

https://files.catbox.moe/n37cop.mp4
https://files.catbox.moe/q3f2tf.mp4

>>106771140
where caN I test sora for free?

>>106771140
I hate giving money to OpenAI, will there ever be a good local video generation model
(Also something that can generate something longer than 10 seconds)

>>106771443
the $200 option apparently does 1080p and 20 seconds, but the $20 is 100 gens a day for 720p and 10 secs which is actually a good deal imo

>>106771465
yes for sfw waifu video genning is good but I wont pay OpenSlop

>>106771504
the good thing is chinese companies will likely use sora gens to train their own models, but lets be honest, only us companies actually innovate / lead

>>106771516
yes, and let be more honest, I cant afford a 3k GPU

>>106771523
GET A FUCKING JOB THEN ANON
STOP BITCHING AND GET OFF YOUR ASS

>>106771533
ok I have a job now. thanks anon.

>>106771551
that's good:)
hope you can save up for the 3k GPU anon!

>98 / 18 /

>>106771568
when the thread that discusses anime generation discusses anime generation

File: catbox_lwpr08.png (2.48 MB, 1152x1632)

2.48 MB PNG

>>106771013
if you want to dick around with something similar, here's a 1dim lora(it's only 5.6mb!) using a different config that's not too far off https://files.catbox.moe/tktp11.safetensors
you'll need/want to use jz235 as a trigger/activation tag and I'd also recommend using faux traditional media, colorful, painting \(medium\) tags since they'll help the style express. I'd also recommend dropping weight to .6-.8~ for actual t2i so that it doesn't go too wild on trying to force out specific things of the image it was trained on. Though it was also trained on noob 0.5 eps so mileage may vary on whatever model you use.
image related is a raw upscale(at 1 weight) of a lowres generated with it at .7 weight with a super basic bitch prompt.
another example is https://files.catbox.moe/7223sf.png which was a 1:1 i2i from the same source image for >>106770582

>>106772227
lewd
I wonder what the text says?

>>106772227
Damn good image. Difficult to generate. I imagine you trained or downloaded a Lora. Either way good work.

File: ComfyUI_temp_snhsx_00013_.png (1.62 MB, 904x1328)

1.62 MB PNG

ah amazing!
a little sesame seed that was stuck between my teeth came out, feels great.

>>106772287
It's an old meme from 2channel from a news report I would say around 2016? If memory serves. Where a couple is seen under an umbrella and the reporter asked the guy what he thought of the weather when he said it gives him a romantic feeling when sharing an umbrella with his lover. His girlfriend/wife was so embarrassed she covered her face but the guy was completely bold in his declaration.

Japan is a somewhat repressed country especially in acts of public affection so it was extremely cute and gathered a 2channel following where people replaced the couple with their favorite characters. In this instance a yuri couple from Madoka Magica.

File: ComfyUI_temp_snhsx_00017_.png (2.09 MB, 904x1328)

2.09 MB PNG

>>106772250
nta but ty for sharing!
>>106772325
aah, very cute indeed

File: 00059-489496360.png (900 KB, 1024x1024)

900 KB PNG

>>106772287
the text itself is likely a complete nonsense but yeah like as other anon kindly explained its a meme reference.
>>106772306
thanks but its actually 0 effort txt2img pretty much, I'm just fooling around with meme tags instead of going to sleep

>>106772227
Can you gen this meme without the text? So you can, you know, place proper text on it with an image editor?

>>106772525
The text is very nearly perfect. There's some issues, on the top the second kanji is barely readable, the third is nonsense. Down below only the 7th kanji is incorrect. The rest of them are directly from the meme and accurate.

https://gelbooru.com/index.php?page=post&s=list&tags=special_feeling_%28meme%29

File: 00062-4032918546.png (1.23 MB, 1024x1024)

1.23 MB PNG

>>106772525
Yeah I can inpaint it (or do another seed) but tomorrow
I'm off to sleep now

>>106772648
Sniffing Gabu who hasn't bathed in three days!

>>106773163
omg it migu
>no turquoise painted nails
Blasphemy.

File: AniStudio-00011.jpg (253 KB, 896x768)

253 KB JPG

>>106773251
Noble maid

File: ComfyUI_temp_snhsx_00093_.png (1.58 MB, 1024x1024)

1.58 MB PNG

>>106773317
And well prepared
If you're new to this hope you're having fun with it

>>106773382
That's cute!
Is there a dedicated thread for maids?

File: AniStudio-00026.jpg (379 KB, 1024x1536)

379 KB JPG

>>106773412

File: ComfyUI_temp_snhsx_00109_.png (1.48 MB, 1024x1024)

1.48 MB PNG

>>106773412
There is not but I see no problem posting maids in any /sdg/ thread personally

>>106773482
>>106773454
>>106773382
Very cute maids

File: ComfyUI_temp_snhsx_00112_.png (1.82 MB, 1024x1024)

1.82 MB PNG

>>106773574
uwu ty
I need to sleep now though, I encourage making them yourself too

>>106773601
Yes I will! I want to merge them maybe a little more mecha maids, would you mind to share a catbox? I'm not the AniStudio clarifying just in case because I think you were talking to him

File: forge photoshopea.png (520 KB, 1649x585)

520 KB PNG

News for Shoopers&Forge Classic users!

Found a fork of an extension that's one month old and currently maintained.

It's a Photopea extension similar to Krita but adapted for Forge Classic. Layers, inpaint, all that shoop jazz

https://github.com/Lord-Vibe/sd-webui-photopea-embed-ForgeClassic

This extencion is it not in the repository you have to install it via URL

>>106773636
>>106773622
Like this but with mayd clothes and a cute android girl with weapons!

>>106773622
I’d box but I’m in afk xd
I’m just using a single ksampler and Illustrious plantmilk-flax with a niji-style lora, if that helps

>>106773722
Oh I thought they'd be more hard. Nice maids tho! Nvm then, no need for catbox!

https://files.catbox.moe/hsr66y.mp4

File: 1759455814071.png (1.31 MB, 816x1280)

1.31 MB PNG

Returning to WAI 10 it's still a good model

File: 1759455814157.png (1.18 MB, 816x1280)

1.18 MB PNG

{1girl, solo,blonde hair, twintails, honey colored yellow eyes, black ornamental hair accessories, blouse, layered petticoat with ruffled edges, ornate golden chain belt with medallions and tassels, looking at viewer, emotionless, holding ceremonial scepter with jeweled details, one hand to hair }

File: 1759456182450.png (1.1 MB, 816x1024)

1.1 MB PNG

Same prompts WAI 15

File: 00000-1208264620.png (1.56 MB, 1024x1024)

1.56 MB PNG

>>106774299
your prompt is now mine

File: 1759456874192.png (1.26 MB, 832x1024)

1.26 MB PNG

>>106774361
Try this! Walnut it's kino, still very creative

File: catbox_hmhj75.png (2.33 MB, 1152x1632)

2.33 MB PNG

>>106772250
also here's a version of the lora trained on noob vpred 1.0, I guess, since I needed to test/adjust vpred settings for that config anyway(and still only 5.6MB) https://files.catbox.moe/cwhjsg.safetensors
pretty much same line of suggestion still applies, though I'd drop the "colorful" tag on vpred because it's completely shitfucked there(or at least, drop it if you're using it on straight t2i). If you want rainbow-ish texturing on hair I'd recommend doing something like <hair color>, multicolored hair.

File: 1759456764988.png (1.27 MB, 832x1024)

1.27 MB PNG

{from above, dutch angle, 1girl, elf, blonde hair, hair flowing over, big hair, braids, hime bangs, big hair bow, tiara, long pointy ears, jewelry, armlet, necklace, earrings, striped dress, sitting, knees together feet apart, holding book, light smile, head up, looking up, floating feathers, reaching, sunflowers, dappled lighting, butterfly, warm colored, Lots of white, mint and black colors, (narrative-rich composition:1.3), bold composition, intricate details, natural color palette, simple background, chromatic aberration, cinematic composition, high contrast, film grain, masterpiece, best quality, amazing quality, very aesthetic, absurdres, highres, newest, 2023, 2024,HDR, 8K, high detailed, subtle,4ur4_illu,beautiful light, best shadow,color blink}

File: 1759456874415.png (1.62 MB, 832x1024)

1.62 MB PNG

Same prompt with hemp2 but didnt work the same, different models without doubt

File: 00009-958287987.png (3.77 MB, 1600x2048)

3.77 MB PNG

Creating Princess Bride's 6 fingered man (girl) was not easy. Stupid prompts couldn't be forced to do it.

Meh on the writing, this was a challenge for myself and I'm not bothered enough to care to fix it.

File: 00012-497469355.png (1.64 MB, 1024x1024)

1.64 MB PNG

>>106774419

>>106774439
So easy to inpaint anon, come on! You will like fixing hands!

>>106774450
>plapeable filter

>>106774453
Inpainting wouldn't do it, it simply created a second thumb. I had to draw the extra finger myself and run it back through three times to get it to work.

Not terribly difficult, but silly I would say.

File: 1759439271083229.jpg (180 KB, 1024x1536)

180 KB JPG

>>106773412
This is the dedicated maid board, so you can post them in any thread.

>>106774510
Cuteeee *and plapeable*

File: oekaki, 2girls, beach s-3(...).png (939 KB, 1216x832)

939 KB PNG

you shake it more than once you're playing with it

File: 1759458149832.png (1.2 MB, 816x1280)

1.2 MB PNG

{1girl, solo, long hair, looking at viewer, short hair, black hair, long sleeves, red eyes, ribbon, closed mouth, jewelry, upper body, earrings, frills, parted lips, puffy sleeves, hand up, white gloves, parted bangs, eyelashes, makeup, red dress, red flower, juliet sleeves, brooch, lace trim, corset, eyeshadow, lace, lolita fashion, skull, red rose, hat flower, red lips, red theme, red gemstone, holding skull, dark, black background, chaos, crack, cracked skin, eldritch abomination, madness, chain, insanity,}

>>106774473
Based

File: 00029-4088303807.png (1.32 MB, 1024x1024)

1.32 MB PNG

>>106774547
I kinda feel like a shorter prompt would work better with the chaos, insanity etc tags at the end of the prompt. also dunno why you've got two different hair length tags unless you want hair on the skull?

File: 00321-2958318049-1267daeb(...).png (3.17 MB, 1344x1728)

3.17 MB PNG

>>106774547

>>106774600
>>106774619
Neat. I think I messed up and forgot I already tagged the heads. Slop happens.

That dead expression with dark green colors with red it's great strong roten vibes > > 106774619

The broken face detail is pure kino, really captures the idea > > 106774782
Thanks for using the prompt!


---

>>106774831
Lovecraftian!!!

https://files.catbox.moe/n13ae5.png

Any SD1.5 model updates?

File: 70961_1588257063_0019ee4b(...).png (723 KB, 1272x666)

723 KB PNG

>>106775080
OH MY GOODNESS! :c

>>106775181
Oh fuck! This really great, is it rain in the background? Nice!

>>106775092
On Civit I only see a Lora here and there, maybe a checkpoint once in a while. Like 90% of the time it’s for SDXL.

>>106766804
This is cute :3

File: 00002-3151807653-ae1490f9(...).png (2.96 MB, 1344x1728)

2.96 MB PNG

>>106775232
I used 'raining blood' and 'surreal'

>>106773636
Now that looks like something out of a cover from a SNES or a Genesis game. Good Stuff (TM)

Is NoobAI still the best local model? It's been a few months since I last genned something.

>>106775382
>Is NoobAI still the best local model?
sadly yes. safetyslop is everywhere now

>>106772661
god i wish

File: 00014-4064046238.png (1.25 MB, 1216x832)

1.25 MB PNG

>>106772525

File: 00029-197966243.png (2.44 MB, 1824x1248)

2.44 MB PNG

>>106776630
also got some random template from google, text is different from original but i don't know the moonspeak so whatever
did img2img passes to make it blend, letters seem to be intact

>>106775283 sovl&kino >>106775429
SDXL with the right prompts could generate cool surreal horror imagery

>>106774782
Delicate lineart. I know it’s novel AI, but did you use any specific artist tag or particular tag?

>>106766008
ok, positive for soul check, keep going

Gm

File: 1759486747187383.mp4 (3.49 MB, 704x1280)

3.49 MB MP4

Friendly reminder that you could animate your gens with SORA

File: 1759340297165663.mp4 (3.41 MB, 1280x704)

3.41 MB MP4

For the foot fags

>>106769602
I run local because privacy, volume, and I love messing with my workflows and come up with new graph aberrations.

>>106776905
Lovecraftian eldrich horror workflows are the best workflows!

Do Pony loras still work on Illustrious shitmixes?

>>106774532
Totally adorbs

>>106776678
Can I bother you for the checkpoint and possible lora combo, or the style prompt? Id like to gen some maids in the same style desu?

>>106777011
I was told they do, but not sure if true

>>106777011
>>106777049
yes and also Illu Loras on Noob v pred tunes, or at least works in my machine

File: 00009-4263003695.png (2.06 MB, 1728x1344)

2.06 MB PNG

>>106777036
It's nai, so no loras were used... got local-like filename because I gen with api on reforge.
can tell the artist combo though - is (orenji \(orenjipiiru\):0.4), (snegovski:0.5), (wagashi \(dagashiya\):0.8), artist:higashiyama shou,

havent been using local for a bit but i'd personally recommend using base noobai with this anon's loras mediafire com/folder/c449c4yww5y7t/noobpoint5 if you want a flatter look. baker's posting here too so I assume it's fine to share

>>106777177
>got local-like filename because I gen with api on reforge
Neat, how do you do that? you put your prompts in the same prompt box of ReForge?

>>106777177
Thank you, Imma try when home

Any way to recreate the local shiny skin style with NAI?

>>106777237
Oh, sorry :3 shiny skin is a local asset, BUT another anon in a thread or 2 ago shared metadata from some NAI milfs with shiny skin and I investigated them and they have a tag for the skin,

File: file.png (534 KB, 1895x925)

534 KB PNG

>>106777190
>Neat, how do you do that?
With this
https://github.com/Metachs/sdwebui-nai-api
And yeah, just using regular prompt box
I only gen with api on this laptop so I could extensively debloat the ui, but if you also have all the extensions and the usual reforge toggles and sliders it can become a bit annoying to use
>>106777212
Sure

>>106777302
This is great anon, and when you upscale/hiresfix, how do the webui logic works?

>>106777324
Hires fix itself doesnt work, you can send the pic to img2img
regular upscale uses your chosen img2img GAN upscaler and does it locally, and then, if you have cost limiter disabled, does img2img denoising with api for the usual anlas cost. you can also do tiled sd upscale, it does the gan upscale locally and then does img2img processing in smaller "free" tiles, essentially a sequence of 832x1216 (or whatever else base 1mp res you're using) img2imgs. obviously its not as good as full attention upscale but most of the time it's absolutely good enough

>>106777417
This is great. We needed some interference from NAI to at least save the images locally. I know there are also some plugins to do the same thing in Silly Tavern and create a kind of visual novel.

Any cooder here?
I asked GPT to make me a .bat file to run Invoke in headless mode.
The issue is that when clicking the .bat file, it freezes at the stage 'Activating virtual environment and starting InvokeAI...' and doesn't seem to advance.

@echo off
echo Activating virtual environment and starting InvokeAI...

REM Activate the virtual environment
call .venv\Scripts\activate.bat

REM Run InvokeAI
invokeai-web --root "My:\Path\Invoke Community Edition"

REM Keep window open if there's an error
pause

>>106777543
>>106777520
As always, sweet gens anon ^^

>>106777302
thats some edgy theme

>>106777302
Theme style? I hate the blue gradio theme

>>106777302
>but you are probably ok if you don't do something stupid like leave generate forever on for days
And then you can make NoobAIv2 with millions of NAI images as dataset

>>106777737
:( poor corgy

>>106777465
Your bat file is probably not in the correct directory for the path to work.

@echo off
echo Starting InvokeAI in headless mode...
echo.

REM Changing to the directory where InvokeAI is installed
cd /d "C:\PATH\TO\INVOKEAI-FOLDER"

REM Activating the virtual environment
call .venv\Scripts\activate

REM Running InvokeAI in headless mode
invokeai-web --root "C:\PATH\TO\INVOKEAI" --headless

REM Keep the window open to see any output
pause

>>106776837
I usually just use oekaki but I dont think i used any style or artist tags for that gen

File: 1743454874014215.jpg (2.14 MB, 2016x1152)

2.14 MB JPG

>>106777746
He realized that hoes don't matter and now enjoys his life.

>>106777872
based 3d hoes doesnt matter ;)

>>106777810
Thankies!

File: image.jpg (309 KB, 900x950)

309 KB JPG

New ToT checkpoint!

Recommend
Sampling method: Euler a
Step: 20~30
Size: >1024*1024
CFG: 5~7


Prompt:

highres, {prompt}


Negative prompt:

lowres, backlighting, cropped, signature, logo, watermark, username, {prompt}

https://civitai.com/models/2011798/k0mix

File: image.jpg (388 KB, 900x1350)

388 KB JPG

Update of famous Hag Checkpoint!
>MatureMAX!

Mature women can add the following item: ((mature woman)), (mature, milf), (mature female),

Choose the shape of the breasts you like: big breasts, or gigantic breasts, or drooping breasts, or perky breasts.

Positives:
masterwork,detailed,high detail,very aesthetic,amazing quality,absurdres,newest,detailed background,{prompt}

Negatives!:
anatomical nonsense,interlocked fingers,extra fingers,watermark,simple background,transparent,low quality,logo,text,signature,(worst quality, bad quality:1.2),jpeg artifacts,username,censored,extra digit,ugly,bad_hands,bad_feet,bad_anatomy,deformed anatomy,bad proportions,lowres,bad_quality,

https://civitai.com/models/1368727/one-obsession-branchmaturenoobaivpred

>>106778215
nice

File: file.png (1.96 MB, 1344x1728)

1.96 MB PNG

>>106777627
Thank you very much!

>>106778253
Why the long face?

What artist do you recommend for Neta Yume 3?

>"hm today I won't gen any 1girl, I'll more something more artistic"
>start doing some medieval type scenery
>an hour later
>already doing 1girl nun in a church with huge nipples
sigh it's just too easy, the temptation

>>106778415
Remember: you are 1girling forever

Idea: Maid Mondays

File: 20250823_0055_Sweltering (...).png (3.05 MB, 1536x1024)

3.05 MB PNG

I have always liked the default Sora anime style tool. But I'm assuming you're locked to SFW.

>>106778702
Yes, and locked also to the yellow filter, but maybe you can maek a 2 pass workflow, generate via Sora, inpaint via SDXL

>>106776905
What's your hardware setup?

File: 20250902_2330_Akari's Sno(...).png (3.21 MB, 1536x1024)

3.21 MB PNG

>>106778716
Doesn't have to be yellow.

File: 00000-1268956003.png (2.04 MB, 1344x1728)

2.04 MB PNG

>>106773622
>>106773601

>>106778702
>assuming you're locked to SFW.
yeah, no pantyshot ever, it's kind of impressive how good they are at censoring this stuff

>>106778728
Wow, how did you removed the yellow filter!? Are you using Gpt image gen via proxy?

File: 00002-1922508347.png (2.68 MB, 1344x1728)

2.68 MB PNG

>>106773636
>>106773664
This is really cool!

>>106778733
I have to leave for work right now, but when I was playing around with Whisk which also blocks NSFW content, I was able to get the prompter to give me underwear if I phrased things creatively. Sports bra isn't usually blocked. And "undergarments" as opposed to "panties" usually worked. I know we're comparing google to ChatGPT, but similar wording could work. I was able to get bikinis though about half of those prompts got blocked as well.
>>106778743
That was from the web ChatGPT Sora back in August. So things may have changed.

>>106776678
Text changes between samples based on what news organization is reporting. So you'll have examples like Alice Margatroid and Marisa Kirisame under the umbrella reported by Aya Shameimaru, Rem and Subaru reported by Lugunica news, Kirari Moroboshi and Anzu Futaba being reported by 765 Productions and so on. So the text is expected to change based on the characters.

In your case it's being reported by ... Oracle. As in ... Oracle
https://www.oracle.com/

Not sure if that was your intention or if it's related to Madoka Magica somehow. Specifically it's being reported as occurring in front of the Oracle building itself. That's gotta be intentional, right? I can't imagine that was random.

>>106778732
Amazing

how do i plap a robot maid?

File: android girls.jpg (123 KB, 1200x857)

123 KB JPG

>>106779085

>>106778358
long face=motherly sex

>>106779201
longs legs are long

File: i3076.jpg (354 KB, 1280x896)

354 KB JPG

>>106778733
challenge accepted

File: 00005-8164234107.jpg (384 KB, 1146x1706)

384 KB JPG

>>106778424
Maid Mondays, I support this. But only if we also have ToT Tuesdays

>>106779238
how many to get that?

>>106779278
i just read the thread
aka >>106778799

>>106769602
Want privacy, unlimited slop, unlimited inpainting? Go local.
Want SOTA with zero skill and don't mind your degenerate prompts being logged by a company in a non extradition country? NAI is your friend.

>>106779238
Will I ever understand the problem with corpses and eroticism?

>>106779361
*corpos

>>106775582
tragic, but atleast it's still better than pony

>>106779361
>>106779376
lmao
28 days to go

File: 20251003-1218-waiNSFWIllu(...).jpg (859 KB, 1664x2432)

859 KB JPG

>>106769602
local
fuck corposhit with their jewery and other bullshit, take back control and do it yourself

>>106779451
BASED. Local it's our personal controled slop.

Anyone tried this >>106778253 ? The sample images look kinda samey. Is it just another overbaked Illustrious mix?

>>106779584
It's an overbaked shitmix of an alredy overbaked shitmix that is OneObsession

>>106779604
Please understand I need the overbaked shitmix because I am not good at the prompting

File: 00000-364252567.png (1.15 MB, 720x1200)

1.15 MB PNG

>>106769602
local, specifically Forge

File: 00035-154384754.png (1.99 MB, 1824x1248)

1.99 MB PNG

>>106777697
Taithrah/Minimal with some half-assed css adjustments.
>>106778805
Thanks gpt.
I also got "Oracle" translation but thought it's some sort of translation mistake

>>106779897
>want to stick my dick in that mouth

>>106776904
Thanks, anon.
It's not very good.

File: sataniapointingupandtothe(...).jpg (172 KB, 571x553)

172 KB JPG

>>106779977

Anons, I want to show my inpainting workflow. Looking for a tool to compile 10 images in sequence (1 frame per second) showing the gen process. Need to export as webm (no audio) for 4chan. What cloud service free should I use?

>>106780047
Use Virtualdub to manually make a 10 frame, 1FPS avi, then use Handbrake to manually encode it to VP9 webm.
Fucking zoomers have to automate everything, goddamn.

>virtudaldub
uhh what year is it?

>>106780103
It still works. I guess maybe I should have said VirtualDub2.
It still works. It's still lightweight. It still has a simple, intuitive interface.
I still use it with Wine long after switching from Windows to Linux because it just works. It's simple, lightweight and just works.

>>106780069
Funny, Ironically, Im a techlet millennial that come here for the diffusion threads. I'll look into this suggestion. Also remembered I have PowerDirector on my pc that I've never touched, will check out both options. Thanks!

>>106780067
nononononono

>>106780160
>thread blessed and saved

>>106780152
With Virtualdub it's so simple - you just name the 10 images 01.png through 10.png, put them all in the same folder, then open the first image in the sequence in Virtualdub and it automatically loads then 10 images as the 10 frames in order of filename. Then you just change the FPS to 1, export as avi, and use Handbrake to convert from avi to webm.
The latter is super quick once you set up a webm preset in Handbrake.
4chan webm filesize limit is 4mb. To calculate bitrate for Handbrake, it's 4096 x 8, then divide that number by the number of seconds in the video, then set bitrate to just under that.

File: _c951e7d8-ff38-4c67-8261-(...).jpg (410 KB, 1792x1024)

410 KB JPG

>>106780009
OM NOM NOM

>>106779308
>>106779451
>>106779729
What's the barrier to entry in terms of hardware?

>>106780334
Ok thanks!

>>106780386
Maybe 6-8GB minimum. Oekaki anon has 4GB I think, always talking about VRAM.

But no worries if you decide to go with SaaS. /g/ is very focused on Open Source, so sometimes there are radical opinions about SaaS. But here I think there are 2 or 3 anons using NovelAI - it has good image quality and "just works." You don't get Loras, but you have vibe transfer, and the tag system is different from SDXL so you might need to adjust things there. Plus, all this local stuff requires a big upfront investment and not everyone has that money to start, and the hobby moves too fast. So if you want to go with NovelAI, don't feel guilty about it.

>>106780386
as the anon with 4gb, I'd say 8gb vram if you're doing images. I still think the best images are going to result from using a larger model like flux or qwen to create a base image and then using an anime model for img2img.

File: 1757021949700038.jpg (1002 KB, 1248x1824)

1002 KB JPG

>>106780386
I've been prompting for a year on 970 before taking 3060 12gb and i can use everything up to qwen, although videos are obviously too slow to bother with.

>>106780386
computer with 8GB VRAM Nvidia card
maybe it will work on 6GB VRAM idk

I feel like the hardest thing to do with pure prompting is having two (or more) characters that aren't interacting with each other but are instead each doing their own thing.

>>106780570
Man this is beautiull

>>106780570
Inspiring

>>106780570
on 970 were you prompting sdxl or 1.5?

>>106779977
Um, not GPT. I speak and read Japanese. Thanks though.

File: 1729042080113495.png (3.36 MB, 1536x1224)

3.36 MB PNG

>>106780703
Mostly 1.5. Here's my typical nearly two years old sloppa.

>>106780763
yeah alright then, that's still painful but conceivable unlike sdxl on 970

File: 1741120435658376.png (1.33 MB, 768x1024)

1.33 MB PNG

>>106780780
I did try some SDXL prompting, but back then SDXL was in its infancy and the results were underwhelming compared to the good anime 1.5 models i had on hand.

>>106780763
>>106780570
Anon, how did you make this in SD1.5??? It's genuinely beautiful. Hats off to you.

File: 1749193272624750.png (3.05 MB, 1536x1224)

3.05 MB PNG

>>106780860
This >>106780570 pic is a recent Chroma gen.
This >>106780763 was with an Akihiko Yoshida lora, and it's indeed a 1.5 gen. I think 1.5 models had the best Yoshida and Amano style loras.

>>106780860
anon 1.5 anime genning was essentially all about the backgrounds because that's pretty much the only thing that you could make interesting without having to stack 15 loras and controlnet

is there a way to generate the backside of something that was already generated? feed in one image and let AI figure out the back of it?
Otherwise all i know is to use WAN to rotate it and just get the frame.

>>106780926
>Chroma gen.
I want to test it! do you have a worfklow or use it throught neoforge!?

>>106781040
No SDXL model can do that, only edit models like Kontext, Qwen edit or Nano Banana