/animu/ - animu

animu and mango


New Reply[×]
Name
Email
Subject
Message
Files Max 5 files50MB total
Tegaki
Password
[New Reply]


00134-1500590896.png
[Hide] (4.8MB, 2688x1536)
> Local UIs
WebUI ReForge, an A1111 fork optimized for SDXL (faster genning speed on low-mid vram gpus): https://github.com/Panchovix/stable-diffusion-webui-reForge
ComfyUI: https://github.com/comfyanonymous/ComfyUI | https://github.com/shiimizu/ComfyUI_smZNodes
Stabilitymatrix: https://github.com/LykosAI/StabilityMatrix

>Online UIs
Novel AI: One of the best anime/cartoon/furry models at the moment. Has a much more recent reference database than most Danbooru based models. Can also be used for writing.
https://novelai.net/links

>Sites to check image metadata:
Local: www.sdimage.info/
NAI: novelai.net/inspect

> Resources and guides
Useful links: https://rentry.org/sdg-link | https://rentry.org/rentrysd | https://rentry.org/sd-mashup
A1111 Wiki: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/
Tags: https://danbooru.donmai.us/wiki_pages/tag_groups
Upscalers: https://openmodeldb.info

> LoRA downloads
https://rentry.org/ponyxl_loras_n_stuff | https://rentry.org/ponyxl_lora_previews
https://rentry.org/illustrious_loras_n_stuff
https://rentry.org/trashcollects
https://rentry.org/zp7g6
https://rentry.org/hdglorarepo

Previous threads: >>109750
>>101185
Replies: >>113376
00132-1500590894.png
[Hide] (4.7MB, 2688x1536)
00160-4029125319.png
[Hide] (5.2MB, 2688x1536)
00187-3920000312.png
[Hide] (5MB, 1536x2688)
>Artist Recommendations List For Noob/Illustrious Models
https://rentry.org/kroiing9
00133-1500590895.png
[Hide] (4.7MB, 2688x1536)
00184-3920000309.png
[Hide] (4.6MB, 1536x2688)
00073-1935833351.png
[Hide] (3.2MB, 2688x1536)
>>113368 (OP) 
>no highlights from the previous thread
Time to rebel.
Replies: >>113387 >>113388
>>113376
Time do do yourself what you want done.
Highlight_1.jpg
[Hide] (1.6MB, 5000x3000)
>>113376
Poorly made in Krita.
Replies: >>113400
>>113388
Good effort, aight highlights.
Though after generating myself the bottom left girl over a hundred times I'm kind of sick of looking at her, haha.
Also, download pureref, its the best local tool for making collages. I'm using it all the time.
https://www.pureref.com/
Replies: >>113408
Highlights1.jpg
[Hide] (1.8MB, 5000x3000)
Highlights2.jpg
[Hide] (1.8MB, 6000x3000)
Highlights3.jpg
[Hide] (2.3MB, 6000x4000)
>>113400
I found Photocollage for Linux and it works pretty well.
00142-2906522549.png
[Hide] (5.9MB, 2688x2048)
00114-1430091017.png
[Hide] (4.4MB, 2048x2688)
00093-1033668245.png
[Hide] (4.1MB, 2048x2688)
Brown and dangerous.
00077-2797432943.png
[Hide] (4.1MB, 1600x2688)
00054-238750281.png
[Hide] (5.1MB, 2048x2688)
00010-101474857.png
[Hide] (6.3MB, 2048x2688)
Dog in need of hug.
00000-2201736647.png
[Hide] (3.9MB, 2688x1600)
00001-1973712391.png
[Hide] (3.7MB, 1600x2688)
What's this elf's problem? Why isn't she in a party?
00042-2146813824.png
[Hide] (5.4MB, 2048x2688)
00023-2145333562.png
[Hide] (4.4MB, 2048x2688)
Big sexy chicken.
00053-445631755.png
[Hide] (6.4MB, 2048x2688)
00069-2730840136.png
[Hide] (7.3MB, 2048x2688)
Cute sassy togruta.
00032-2799205979.png
[Hide] (5.1MB, 2048x2688)
00004-523493284.png
[Hide] (5.4MB, 1792x2688)
Legs:Ashley
00072-999581074.png
[Hide] (4.3MB, 2048x2688)
00076-1911330734.png
[Hide] (3.8MB, 2048x2688)
00044-3444103596.png
[Hide] (4.1MB, 2048x2688)
Strongest magical girl thighs.
00182-3676351792.png
[Hide] (2.2MB, 2048x2688)
00183-3676351793.png
[Hide] (2.5MB, 2048x2688)
00184-3676351794.png
[Hide] (2.6MB, 2048x2688)
00180-4202688241.png
[Hide] (2.9MB, 2048x2688)
00168-3022623329.png
[Hide] (2.8MB, 1600x2688)
00088-4150664370.png
[Hide] (5.3MB, 1792x2688)
00087-3587269519.png
[Hide] (4.8MB, 1792x2688)
This style combination really worked out.
00269-3079102549.png
[Hide] (5.2MB, 2688x1536)
00273-2164433751.png
[Hide] (5.3MB, 2688x1536)
Comfy scenery gens.
00062-1218078854.png
[Hide] (5.6MB, 2048x2688)
00037-462180828.png
[Hide] (4.4MB, 1600x2688)
00008-1727077414.png
[Hide] (5MB, 2048x2688)
Had to explain to someone how to get started with local genning. So I might as well post it here too.

Here's an easy way to setup for making AI gens local.
>Download and install Stabilitymatrix. Works on both Windows and Linux. You'll need around 50GBs I think.
https://github.com/LykosAI/StabilityMatrix
>Download the Stable Diffusion WebUI reForge
>Download a model. I recommend these 3 for starters. They go in the folder "/stabilitymatrix/Data/Models/StableDiffusion"
https://civitai.com/models/827184/wai-illustrious-sdxl
https://civitai.com/models/1201815?modelVersionId=1925170
https://civitai.com/models/835655/illustrious-xl-personal-merge-noob-v-pred05-test-merge-updated

>Get a VAE. I recommend these 2. They go in the folder "/stabilitymatrix/Data/Models/VAE"
https://civitai.com/models/1684460/illustrious-xl-v20
https://civitai.com/models/285852?modelVersionId=2331459
>Get a few textual inversions. I recommend these. They go in the folder "/stabilitymatrix/Data/Models/Embeddings"
https://civitai.com/models/1302719?modelVersionId=2121199
https://civitai.com/models/1302719?modelVersionId=2268235
Replies: >>113447 >>113533
>>113446
When you first start up reForge UI on Stabilitymatrix it'll open up a new tab in your browser with a local host or loopback address. Your UI will show up here. From there you'll see some tabs at the top. Click on "Extensions". From there you go to the "Available" tab. Click on "Load From:" and it'll load a bunch of extensions that you can download. Here's what you'll want starting off:
a1111-sd-webui-tagcomplete
adetailer
sd-dynamic-prompts
sd-webui-lobe-theme
sd-webui-prompt-history

>Optional ones to get once you know what you're doing:
sd-webui-vectorscope-cc
sd-webui-resharpen

Read the descriptions or their pages to understand how to use them. Dynamic prompts is the only one that'll require additional input on your part as it requires you to make wildcards. We can worry about that later though.

With all that done reload your UI via Stabilitymatrix and all the new features will be loaded. Check the tabs on the screen saying "Checkpoint, lora, textual-inversion" to see if anything pops up. If so then the files are in the right folders and you can use them to start genning.

That's it. It's basically a 20min process. There are some optimization stuff you'll probably have to do on your end but there are guides on how to do that if you care.
00082-3334312942.png
[Hide] (2.5MB, 1600x2688)
00114-360495960.png
[Hide] (3.4MB, 2048x2688)
Thick Linne thighs.
>>113446
>Stabilitymatrix
Didn't know about that one, seems interesting, thanks for sharing.
>>113446
>recommending reForge for newbs
>not comfy
I might be biased because I started with it, but I think comfy's a much better choice, both in terms of ease of installation, as well as future potential for expanding what it can do.
Replies: >>113534 >>113571
20251122062753-novaAnimeXL_ilV130-Euler_Max-Align_your_steps_gits-steps27-cfg3.2-seed1350609170.png
[Hide] (998.4KB, 1152x1472)
20251122063909-novaAnimeXL_ilV130-Euler_Max-Align_your_steps_gits-steps27-cfg3.2-seed639452108.png
[Hide] (1.3MB, 1152x1472)
20251122064551-novaAnimeXL_ilV130-Euler_Max-Align_your_steps_gits-steps27-cfg3.2-seed3744298454.png
[Hide] (1.1MB, 1152x1472)
20251122071335-novaAnimeXL_ilV130-Euler_Max-Align_your_steps_gits-steps27-cfg3.2-seed1411411637.png
[Hide] (903.8KB, 1152x1472)
>>113533
unless you're already used to node UIs it looks like a pain in the ass to use, unlike the simple dropdowns, tabs and checkboxes in forge that everyone is familiar with
Replies: >>113535
>>113534
>unless you're already used to node UIs it looks like a pain in the ass to use
Lmao, no, the basic template you start with immediately gives you all the tools you need.
Also pissing is gross, so stop that.
Replies: >>113536
>>113535
>pissing is gross
what do you do when your wife or gf squirts in your mouth? say "EWWWWW THATS GROSS!!" and become gay instead?
Replies: >>113537
>>113536
Why and how would I even end up in a situation where's that possible?
No, I'm not going to lick there, fuck off. She can suck me off, but it works only one way.
00023-2479951128.png
[Hide] (4.8MB, 2048x2688)
00011-3553884740.png
[Hide] (4.4MB, 2688x2048)
00208-1246361800.png
[Hide] (4.2MB, 1600x2688)
>>113533
Comfy is better overall and doesn't require playing with the settings to get exactly what you want optimization wise. I started with reForge so I just find it easy. Alternatively there's SwarmUI.
00007-2613062901.png
[Hide] (4MB, 2048x2688)
00008-2613062902.png
[Hide] (3.8MB, 2048x2688)
00014-2325977759.png
[Hide] (3.7MB, 2048x2688)
Liking this artist mix so far. Here it is if anyone is interested. 
https://files.catbox.moe/czmaho.png
00053-1953417433.png
[Hide] (5.9MB, 2688x1600)
00003-2935873882.png
[Hide] (5.7MB, 2688x1600)
00077-1314610228.png
[Hide] (3.2MB, 1856x2688)
00120-3309863680.png
[Hide] (3.9MB, 1856x2688)
00145-4100929462.png
[Hide] (4.5MB, 1920x2688)
00045-2305780853.png
[Hide] (4.3MB, 2048x2688)
00049-75015130.png
[Hide] (4.8MB, 2048x2688)
00057-4192015633.png
[Hide] (4.3MB, 2048x2688)
Oppai Liner Illya Prisma
00107-3181907374.png
[Hide] (4.5MB, 2048x2688)
00106-3181907373.png
[Hide] (4.8MB, 2048x2688)
00111-1166775006.png
[Hide] (4.9MB, 2048x2688)
00068-1030507836.png
[Hide] (4.3MB, 2048x2688)
00042-731495313.png
[Hide] (6.4MB, 2688x2048)
00115-1642507482.png
[Hide] (4.5MB, 2048x2688)
Big titty committee.
found this
https://tagexplorer.github.io/
have fun
Replies: >>113825
00067-2810413207.png
[Hide] (3.8MB, 1536x2688)
00059-2743674314.png
[Hide] (3.3MB, 1536x2688)
00057-3620802761.png
[Hide] (3.3MB, 1536x2688)
>>113784
Nice find.
00002-2225415862.png
[Hide] (5.4MB, 2048x2688)
00007-4093524674.png
[Hide] (5.7MB, 2048x2688)
Myrrh with cow tits just fits right.
00176-2859066446.png
[Hide] (6.2MB, 2048x2688)
00180-3538707082.png
[Hide] (5.7MB, 2048x2688)
>Sexually suggestive
00131-2954165503.png
[Hide] (5.9MB, 2048x2688)
00154-3588624805.png
[Hide] (5.6MB, 2048x2688)
Reimoo
ComfyUI_10232_.png
[Hide] (897.8KB, 840x1368)
ComfyUI_10242_.png
[Hide] (1.1MB, 840x1368)
ComfyUI_10298_.png
[Hide] (1.1MB, 840x1368)
ComfyUI_10299_.png
[Hide] (604.2KB, 840x1368)
ComfyUI_10307_.png
[Hide] (1.1MB, 840x1368)
I love Lotte so much its unreal
>y tho?
She has revealed herself to me in a dream.
Base Wai/Illustrous can generate her alright, but her glasses are off. Loras help but its still problematic, especially if she's supposed to have her eyes closed.
Replies: >>113838 >>113844
ComfyUI_10331_.png
[Hide] (2MB, 840x1368)
ComfyUI_10352_.png
[Hide] (1.6MB, 1272x840)
ComfyUI_10354_.png
[Hide] (1.9MB, 1272x840)
ComfyUI_10355_.png
[Hide] (1.5MB, 1272x840)
ComfyUI_10373_.png
[Hide] (1.5MB, 1136x952)
>>113836
Also Masikakuma, Orenjipiru and Gunsmith Cats loras are pretty neat. Though compatibility with the Lotte lora (specifically her eyes-through-glasses thing) is wonky. Also the Lotte loras often enforce their own style really hard. Like the first pic, it was supposed to use the Jan Matejko Lora, at a pretty high strength too, yet the lady herself is practically untouched.
Comboing the Chronozbergh tag with the Masikakuma lora sometimes produces a nice result (pic3).
Replies: >>113839 >>113840
>>113838
Sry, MasiXChrono is the 4th pic, not the 3rd. The 3rd is pure Chronozbergh. Not lora, just the tag.
ComfyUI_10389_.png
[Hide] (1.2MB, 840x1368)
ComfyUI_10383_.png
[Hide] (1.1MB, 840x1368)
ComfyUI_10376_.png
[Hide] (1.2MB, 1136x952)
>>113838
ENTER
WORST HEABY METAL GUITARIST EBER
LOOOOOOOTTEEEEEEEE
Yayahiyayayaka's style is so great too.
Artifact ridden eyes aside, it was funny how the model added an unprompted Akko holding Lotte's leg.
00036-2692472506.png
[Hide] (3.2MB, 1792x2304)
00191-258234929.png
[Hide] (5.8MB, 2048x2688)
>>113836
That 3D one is pretty cute. Also I still have trouble figuring out what's a good 3D model or what's good for 3D styles. So my 3D gens usually look awful.
Replies: >>113862
00077-2459845792.png
[Hide] (4.9MB, 2048x2688)
00067-2555389915.png
[Hide] (5.3MB, 2048x2688)
ComfyUI_10432_.png
[Hide] (2.6MB, 1680x1768)
ComfyUI_10432__polish_1.png
[Hide] (3.2MB, 1680x1768)
ComfyUI_10496_.png
[Hide] (845.2KB, 1376x912)
ComfyUI_10502_.png
[Hide] (1005KB, 1376x912)
ComfyUI_10508_.png
[Hide] (1.1MB, 1376x912)
>>113844
That 3d one used a Rockman Legends style lora with "low poly, 3d" tags even though the lora in theory doesn't need them. Probably a compatibility thing with the Lotte lora.
Also, generate more Lotte.
Also also, rariatto's style is so pretty.
ComfyUI_00057_.webm
[Hide] (1.4MB, 1280x704, 00:01)
ComfyUI_00055_.webm
[Hide] (571.5KB, 1280x704, 00:01)
ComfyUI_00053_.webm
[Hide] (1.2MB, 1280x704, 00:01)
ComfyUI_00051_.webm
[Hide] (1.3MB, 1280x704, 00:01)
ComfyUI_00049_.webm
[Hide] (1.2MB, 1280x704, 00:01)
THEY ALL LOOK LIKE FUCKING TRASH, HOW DO YOU EVEN USE WAN 2.2???/?
Its way faster than framepack, but the gacha is utter shit. And sometimes I get truly cursed results. I wonder if its a skill issue or just actually shitty tool.
Replies: >>114054 >>114056
>>114053
I've never seen a passable 2D AI animation anywhere. AI video models are built primarily for 3DPD and until there's a serious high budget 2D-focused one that gets everything just right, AI anime will continue to look like shit.
Replies: >>114068
1girl,aura_(sousou_no_frieren)_,geso_smith,xiangzi_box,mucha__(muchakai_),uni8,0_s-3760316829.png
[Hide] (1.3MB, 832x1216)
00124-3651660516.png
[Hide] (4.7MB, 2688x2048)
00072-2915817270.png
[Hide] (4.9MB, 2248x2248)
>>114053
Speaking of which, this just came out.
https://www.youtube.com/watch?v=I_b2QN-B1W0
https://github.com/Lightricks/LTX-2

I was going to try out LTX-2 this weekend. It works on low VRAM while supposedly also being more accurate and making twice as long animations (15 seconds). It already has a bunch of loras for it as well as artstyle lora tutorials for it.
Replies: >>114062
media_G-Q7yHpW4AEf-Qp.png
[Hide] (409.1KB, 634x511)
media_G-Q75GjWIAAd-O-.png
[Hide] (231.1KB, 553x680)
media_G-Rsy1XWUAACnSK.jpg
[Hide] (145.7KB, 1070x1330)
Could you guys make some AI slop using these images as a base?
>>114056
So you're telling me there's still hope
>20-40gb model files
I really hope you don't need all of them to run it, just 1 max.

Man, it'd be neat if there was something passable locally...
Well, there is, all those posts seem to have been made locally
>>111572
>>111630
>>112876
>>112728
so its most likely a skill issue on my part.
>20-40GB models
has the price of 32GB of RAM gone back down to $50 yet like it was a year ago?
Replies: >>114065 >>114066
>>114064
ram prices are tru the roof and now ssds prices are going up
good luck
>>114064
Its not even RAM, you need VRAM. 24 and 32gb are the highest end consumer grade graphics cards on the market.
>>114054
>until there's a serious high budget 2D-focused one that gets everything just right, AI anime will continue to look like shit.
Cygames opened an AI Studio last month. Goo chances there will do something animation related, but it's more likely it'll be a closed model.
Is this a good place to get help regrading image generation? I'm having issues such as characters with defective hands or carrying deformed/unnatural objects. I have a bunch of questions about things that are confusing to me. It would be nice if someone experienced here could help me out with these problems.
Replies: >>114079 >>114084
>>114078
>oh no, a shy retarded autist
Don't ask if you CAN ask a question, just shoot you baka.
Inpainting_Guide.jpg
[Hide] (1.2MB, 2500x2900)
3girls,friends,aura_(sousou_no_frieren),akiyama_yukari,nice_nature_(umamusume),_s-2444535691.png
[Hide] (1.4MB, 1216x832)
>>114078
Deformed or unnatural objects most likely means you're using the wrong resolutions. You're resolution for images needs to be divisible by 32 as that's what everything is trained on. Put these numbers down on a sticky note or something you can quickly read for reference.

Image resolutions:
>512, 544, 576, 608, 640, 672, 704, 736, 768, 800, 832, 864, 896, 928, 960, 992, 1024, 1056, 1088, 1120, 1152, 1184, 1216, 1248, 1280, 1312, 1344, 1376, 1408, 1440, 1472, 1504, 1536,
Modern models use 768 or 1024 by default instead of 512. Don't go above 1344 unless you have a model that specifically says it can go up to 1536. If you want larger images, use either Hi-res fix, img2img, or Extra (if on A1111 UI/forks) to upscale. All have different results and uses. Easiest and most straight forward is Hi-res fix. 

Hands? Try using adetailer as that'll make it easier. Adetailer is an extension that inpaints certain areas right after you gen. Making its own mask and changes based on the prompt you give it. Basically auto inpaint. Read the github page if you want to learn how that works. As for inpainting, the image is a good guide on that.
Replies: >>114137
>Put these numbers down on a sticky note or something you can quickly read for reference
or in (re)forge you can change "txt2img/Width/step" and "txt2img/Height/step" to 32 in ui-config.json
00049-4279651381.png
[Hide] (8MB, 2248x2248)
00173-1445884837.png
[Hide] (5.5MB, 2048x2688)
https://rentry.org/kroiing9
Updated this artist list for AI usage. I've been slowly adding images for the artists.
Replies: >>114205
00034-1813582274.png
[Hide] (6.9MB, 2048x2688)
00025-1561313156.png
[Hide] (6.5MB, 2688x2048)
This is the new face of the Extremism.
>>114084
Thanks for the advice. Did you make that inpainting guide yourself or did you get it from someone else? How long have you've been making image generations? I use a cloud service (vast.ai) for generating images. I'm not sure if it's safe to generate stuff like loli, though.

Is it a good idea to follow this inpainting tutorial as well? https://www.youtube.com/watch?v=DgphyDiD4TE

Should you start off with realistic models if you're a beginner and save the anime models for later? Or can you just go ahead and start using anime models? I'm not sure if the process of generating images with anime models and realistic models are different from each other and if one is easier/more beginner-friendly than the other.
Replies: >>114146 >>114147
00084-1039107195.png
[Hide] (6.3MB, 2048x2688)
00042-357446272.png
[Hide] (5.1MB, 1728x2688)
ComfyUI_00012_.png
[Hide] (1.5MB, 864x1208)
>>114137
That isn't my guide. It's one another anon made. I've been genning for 2 years now and it's still fun. Never heard of Vast.AI. Use NovelAI instead. You can gen whatever you want; furry, guro, loli, handholding, etc. It can even add text to gens if you want. Not only that but it has the most up-to-date artist database.

That tutorial in the vid is fine. The image I provide is just easier to reference but that video explains things in more depth. 

Just go straight to anime models. Illustrious and Noob AI models introduced in late 2024 have made genning anime girls much more consistent and easier and a large majority of modern AI image models used locally are based off those 2. That aside, NovelAI is its own model. So you won't have to worry about what model to select if you use it.

This is all assuming you can't gen locally though. As long as you have close to 8GBs of VRAM (from your GPU, not literally RAM), you can gen just fine at decent speeds. Best way to gen remotely from your computer is using your phone to access your computer and genning from there.
Replies: >>114149
>>114137
>Should you start off with realistic models if you're a beginner and save the anime models for later? Or can you just go ahead and start using anime models? I'm not sure if the process of generating images with anime models and realistic models are different from each other and if one is easier/more beginner-friendly than the other.
Realistic models are generally bigger, slower and can understand grammar and context more.
Anime models (all of which are practically illustrous derived nowadays) work based on danbooru tags, with very limited understanding of things that aren't tags there. So you should think how to describe a picture using tags you'd see on booru. Some tag autocomplete is a must, it makes the process way easier and faster.
While they can't really understand grammar or natural text, you can do few things outside of tags. There's no "red cat" or "giant cat" tags, but illustrous does understand colors and sizes and it generally should be able to generate a red cat or a giant cat.

tldr for starting with local: make sure you have at least 8gb Vram GPU, download the comfyui installer, install that, add autocomplete tag plugin, download some animu model (e.g. this thing https://civitai.com/models/827184?modelVersionId=2514310) and you're basically good to go. Everything else is a matter of skill and understanding of the tool.
>>114146
>pic 3
Call me when they release White Dandelion: Purification Palace.
r8i649z7sc42g5u4.mp4
[Hide] (13.3MB, 1920x1080, 01:07)
Do you want to help save the British Isles, Anon-kun?
Posting it here since the video is AI.
Replies: >>114196
00138-3412538236.png
[Hide] (3MB, 2688x1536)
00192-2019617660.png
[Hide] (4.5MB, 2048x2048)
00011-2157364711.png
[Hide] (3.6MB, 2048x2688)
Replies: >>114211
00067-1773630501.png
[Hide] (4.6MB, 2048x2688)
00135-2006399209.png
[Hide] (3.3MB, 2688x1600)
00116-3912410374.png
[Hide] (3.6MB, 2048x2688)
Clara's tails is so inconsistent.

>>114186
A well done trailer. Much better propaganda than the source material.
Replies: >>114211
>>114093
can you upload the first pic so i can copy it to comfyUI?
Replies: >>114217
>>114195
>>114196
Please share catboxes while the site is still up.
Better yet, come over to /monster/'s AI thread
https://smuglo.li/monster/res/212147.html
They'd like your gens.
00116-4177982483.png
[Hide] (5.1MB, 2048x2688)
>>114205
https://files.catbox.moe/accs2l.png
Bat wife.

Sure. I'll share some gens there.
Replies: >>114226
>>114217
arigato
[New Reply]
71 replies | 125 files
Connecting...
Show Post Actions

Actions:

Captcha:

- news - rules - faq -
jschan 1.7.3