previous >>8992532>Pick one and install Stable Diffusionhttps://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file#installinghttps://github.com/lllyasviel/stable-diffusion-webui-forge?tab=readme-ov-file#installing-forgeComfyUI image/video model guideshttps://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file#examples>Recommended XL modelsNoobAI: https://civitai.com/models/833294/noobai-xl-nai-xlDHXL: https://civitai.com/models/1291404/dhxlv10Illustrious: https://civitai.com/models/795765/illustrious-xlConfettiTuneXL: https://civitai.com/models/184813/animeconfettitunexl-DMD2 speed LoRA: https://civitai.com/models/1608870/dmd2-speed-lora-sdxl-pony-illustrious>Uncensored Flux modelChroma: https://civitai.com/models/1330309/chroma | https://civitai.com/models/1807633/chroma-unlocked-q4ks (low VRAM)>AI image upscaler model databasehttps://github.com/OpenModelDB/open-model-database>LoRA metadata viewer in HTML formhttps://github.com/Xypher7/lora-metadata-viewer>Optional extensionshttps://github.com/DominikDoom/a1111-sd-webui-tagcompletehttps://github.com/hako-mikan/sd-webui-regional-prompterFor ComfyUI:https://github.com/ltdrdata/ComfyUI-Manager>IS MY GPU GOOD ENOUGH?https://www.tomshardware.com/pc-components/gpus/stable-diffusion-benchmarks>3DPD, Furry, Futa>>>/aco/asdg>>>/trash/sdg>>>/b/AI>Related threads>>>/g/sdg>>>/h/hdg>>>/e/edg>>>/d/ddg
>>9005298How do you make these galleries?
>>9006027Not the anon that did those but you have to be logged into/create an account on there. Everything you upload while on said account can then be managed into albums like that.
Requesting topless jeans pinups of Molly Jo (Amphibia).
>>9006226catbox for this?
Requesting Juniper Lee humping Wuya's ass.
>>9006304Here you go buddy boy.https://files.catbox.moe/k6cf3l.jpg
>>9006500thank you kindly :D
Did LoRA_Easy ever get step by step instructions? I can gen across a1111 and comfy but can't figure that shit out
I'm in love with the arthemycomics checkpoint.
Requesting this (https://files.catbox.moe/9ljkr8.png) with Masha (The Owl House).
>>9006695
>>9006795
>>9006796
Can I get Pregnant Anubia?
>>9007218Absolutely LOVE this character!
>>9007336
>>9007301Same. Since I didn't make any mention of her before, it's Frankie from Knights of Guinevere for those unaware. muscular shortstack tomboy is quite the combo that ticks some real boxes for me.And suddenly, an Amethyst.
>>9006795>>9006796>>9006797I see you are back, how are those Asuka pics coming along?
Need poses ideas for the statue of liberty for some reaosn
>>9007862kneeling, squatting, double V,
>>9008143>Patreon
man, the realism thread died out
>>9008143>signing your slopCouldn't be me>implying 'pay me for it'Definitely couldn't be me
>>9008596That's interesting, seen how much more active it was than this one. I didn't go there that much, I wonder what happen.
>>9008815Looking into it more, somebody split off Flux, Hailuo, WAN video and other non-SD models into a separate general. As if we didn't have enough of them already.
>>9006961
>>9005626>>9006057nice Frankies, LoRAs?
>>9008925Main one I used for her is this one: https://civitai.com/models/1972583 , you'll find my gens on that page with all the other models/loras noted on their post.
>>9008815>Don Blewd
>>9008815>>9008830Too many spiting dilutes the posters until its all dead. I wish this thread had more traffic as I tend to like cartoon style the best.
>>9008830No SD models were ever banished from /asdg/.Playing gacha/training DALL-E and spamming kissing animations (Hailuo, that has long been dead) were.
>>9009541I'm just saying Flux used to make up a large portion of /asdg/ activity, and that has now moved elsewhere.
>>9009089neat! i used another Frankie LoRa for more show accurate pieceshttps://civitai.com/models/1990099/frankie-knights-of-guinevere-il?modelVersionId=2252717
>>9009735and also, an Andi onehttps://civitai.com/models/1974047/andi-or-knights-of-guinevere-or-il?modelVersionId=2234419
Flexible girls are hot
someone remind me again how NAI hasn't been sued?
>>9010094People are generating too much Slop that they haven't had a chance to sue them.
>>9009735Hmm, you might be right, the one I was using didn't really account for her actual nose shape very well. Might have to use it more in the future. Seems like none of these loras account for her missing teeth, though, but that's simple enough for me to fix. (had to manually edit that into that one gen that had it)
>>9010162i think your pics look really good regardless of the missing details, anon, better than the ones i make with the shitty built in Civitai generator because i'm too lazy to install it manually
>>9010193Yeah I'm not sure what the civitai gen's deal is. might have to sink more buzz into it for more steps or something like that but it seems like it doesn't quite match local genning even with all the same loras present. draft mode isn't very useful since drafts and non-drafts are quite different in quality and composition even when on the same setup. I also use remacri upscaling but civit doesn't detect this nor does it allow me to manually add it to resources. that said, it wouldn't change the base gen makeup.And yeah, I think mine were accurate and good enough. Since regardless of lora accuracy you're still gonna get a lot of varying personal interpretation aligning more that genner's taste. I remember hearing official design notes from the show indicate she canonically has a powerlifter-esque build, but there hasn't been enough of her shown to really determine how much body fat ratio she actually has, but I'm certainly not alone in at least thinking she should have at least some pudge even if she is built, they're not mutually exclusive after all.
Succeed in doing a fem Fawful that DIDN'T look like shit. Why Fawful? Idk, was bored
>>9010094Because Nai is a big company, so Nintendo won't touch it unlike small harmless bunch of people's that fo roms and fangames
>>9010269Adding "muscular" and "plump" to the prompt does give her a more strongfat physique in Ilustruous, which is how i think she looks like
How do I go around to making a scene with 2 or more characters using Comfy?Or am I supposed to generate an image with some generic characters and then inpaint over them with what I really want?
>>9011187Don't use Comfy, but other than using regional prompt on the other ones, you just have to inpaint.
>>9011190What should I use then?
>>9011190I don't see why not, and yeah regional prompter is the way to go.
Almost lost ya guys
She doesn't like the uniform
it's actually Kool-Aid
Any tips on how to make characters do cosplay using LoRas of other characters?
>>9011343Well, the top effort option is this >>>/d/11386033 to which you'd have to add regional lora scheduling. Which is guaranteed to work, but maybe there are easier ways.
>>9008830 urge to balkanize into smaller threads over trivial, temporary issues is prehaps the weirdest aspect of AI art threads.
Crystal milk maids
>>9011343Sometimes a lora at low strength is enough to be able to use that character's outfit. Like start with a strength of 0.6 and keep reducing it until it no longer works.Negative prompt for the character's physical features that aren't the clothes. Completely dependent on the lora quality whether or not the character's clothes and features are hard baked together.Of course some popular characters have explicit costume/cosplay loras.
>>9011444That's a BIG one for her! ^^Can we see her fucking it nude in Reverse Cowgirl position? (Same thing for Amy too?)
>>9011200I use Forge/reForge. I don't usually go too crazy on the number of characters I gen at once, If I really want to go big, I just try to gen in pairs or individually (depending on the use of character Loras) and then put them together on photoshop or something similar.>>9011343Like other people have said, you can use the costume from another character by using their Lora and just start low strength on it until you archive the desired result without messing with the other character or the style you are aiming for.
Eva succ pp
>>9011904
>>9011905
>>9012118
>>9012123
>>9012127
>>9012128Thus ends the grotesque erotica of a MadMark/Namespace hybrid style.
>>9011444Is there a sequel to this pic maybe?
>>9011841Tried that, it gives me a costume that seemes more "inspired" by the character rather than the actual costume but again, this is the shitty built in Civitai gen we're talking about
>>9012049https://files.catbox.moe/4obo82.pnghttps://files.catbox.moe/stal9z.pnghttps://files.catbox.moe/idajdq.pnghttps://files.catbox.moe/qw7hkh.pnghttps://files.catbox.moe/jf4718.pnghttps://files.catbox.moe/fplnos.pnghttps://files.catbox.moe/9euwnx.png
>>9012717More of her please :D
first post ever hope you likey.
>>9013038
is there a good tutorial for setting this up? I have all the filesfeel free to call me a retard along the way
>>9013208i think you just install comfyui or something and then look up tutorial on youtube on how to use it.that's what im doing anyway.
>>9013226its completely inscrutable. i downloaded the models recommended on this thread. i can't get them to run. not even sure if i'm installing them correctly
>>9013234Which stable diffusion did you installed? Make sure you are putting the models in the correct folder and the loras on its own folders as well.
>>9013256i'm not seeing any explanation for what goes where. or how to create a workflow. i'm picking up bits and pieces but i don't see a start to finish explanation anywhere
>>9013258comfyui btw
>>9013258Here is a guide that might help you.https://rentry.org/sdgguide
Hey, while we've got comfyUI people on here, does anyone know how to do the equivalent of the "only masked" inpainting that webUI had in comfy for more detail and to fix AI weirdness like screwed up hands?
>>9013262sorry this doesnt say anything about how to set it up after you have it installed. I can't imagine every single person just figured this out on their own
Used a different LoRA for Evangelyne succ
>>9013323
>>9013324
>>9013400Specifically a MrPotatoParty LoRA and a Namespace one.
Massive thank you to whatever Anon told me what this kind of outfit is called.
>>9013505This LoRA combination gives her a thick, cock-milker of a tongue.
>>9013507
>>9013510
>>9013515
>>9013523
>>9013524
>>9013525nice set
>>9013208>>9013277For 95% of people comfy is unnecessarily complicated with no added benefit, just get Forge or reForge.Either way, to best way to learn is to steal a catbox from an AI thread and drag it into your UI. That gives you a full workflow of what they did to achieve it. Unless it's made with novelAI which is an online service.
>>9013208If you go for it just do like >>9013540 and taking from him you can also download pictures from civitAI from the stuff you like and replicate it doing the same>top left>ComfyUI button>Open>get your downloaded .PNG and open it>it will display the workflow right awayAlso it will show ONLY THE MISSING WORKFLOW NODES, with this I mean It WON'T download the models, or loras, you need to do that yourself.in order to make comfyUI download the nodes automatically just go to the top right to a buton with a puzzle piece called Manager---->Install Custom Missing Nodes, always download the latest/nightly versions and then you reboot ComfyUI and it should work.There is an exception with this in case you stumble on a Workflow with something called Ultralytics (long story short the AI feature you might know among many other things Facial Recognition) needs a bit of manual tweaking to make it work.There's many types of ultralytics the most common ones are,face,eyes,hands,body,NSFW(penuuuses,vaginas,bussy...)In order to make em work you gotta white list them.there are so many ways people do this so I'll tell you the easiest way to learn on your own.>run workflow until you reach an error from the ultralytics module.>copy the log ">_" (its an icon on the bottom left of comfyUI ">_", its in between a help icon "?" and keyboard shortcuts icon)>once you are in the log screen check you are in the propper tab cause there are 2 of em LOGS/TERMINAL>copy the whole log.>post it to chatGPT/coopilot any AI you use,>it will tell you your Ultralytics its not whitelisted and explain you step by step how to do it on your own.The only other thing which you have to do remember its downloading your models and copying in their right folder.My tip is make subfolders in each model folder so you know what category is it like Characters/poses/artists/backgrounds cause you might have a ton of loras soon.
>>9013589ran out of characters, let me keep with the basics:that picture you're watching is the first iteration, in my case I make the first picture very smol and upscale it though what its called a KSampler and a Upscaler 1 or two times til it looks like the picture I just posted.But in order to learn properly once you have it installed just take 1 basic Checkpoint(NOOBAI(NAI),ILLUSTRIOUS(IL) or PONY) these are the big gooner models just pick whichever I like IL but everyone has different tastes, then grab one VAE.And build your first workflow so you get a basic grasp of how things work.Modules you gotta load:>Load Checkpoint (here you select the baseline model, in case you don't understand the .safetensor files with a heay weight like 6GB up to more space)>CLIP Set Last Layer (to explain it shortly its how much your words in the prompt will be understood almost., almost anyone uses them in their workflows so don't worry much about this one)>Empty Latent image (this will be where the picture will be saved, imagine a blank canvas)>CLIP Text Encode (here you will write the prompt)>KSampler (this is where the magic happens its where your computer will build the picture PAY ATTENTION TO, Steps,CFG,Sampler and Scheduler those are the meat of potatoes of picture generation)>VAE Decoder (it decodes the pictures from 0s and 1s to actual things we can see like a .PNG)>Save Image (self explainatory it saves your picture in /comfyUI/output (some workflows don't save them they keep em in temporary storage WATCH OUT cause once you close comfy it will delete everything on it so save what you want for future reference))
>>9013606 it takes a little time to make it work but its longer to understand every inner thing.And once you understand this you can start with LORAS,which expand on stuff you might not be able to generate on your own with a base model, for instance I want some especial competitive swimsuit that my IL doesn't generate, I'll download a Lora for it and just load it and set the parameters which will be almost always between 1 and 0.75.FOCUS on making the basic shit work worry about Loras other shit like wildcards (random word generation, people will get lists of tags from Danbooru sorted by character,artist,gender,camera angle,perspective,clothes,emotion...).This explains about 75% of everything we do with comfyUI in the goon threads in every single board, if you have more questions ask away.Don't remember what a node does?no worry click on the node and you will see a pop up with a help icon just click it and it show you the documentation about it.ALSO CHECK YOUR INSTALL READ BELLOWI'm assuming you have it installed properly, you can check by coping and pasting the logs after initializing comfy UI and asking AI about it (Pytorch version with CUDA,detecting your VRAM and RAM,displaying your CUDA device and the vram state should be NORMAL VRAM) you should do this to make sure you aren't running it on CPU mode, which will make picture generation impossible, and have in mind your own tech specs like the minum for it its 8GB of VRAM and 32GB of ram with a normal processor, with this you will be able to do any pictures and NO VIDEO OR ANIMATIONS (unless you want to spend an awful amount of time generating it)Also there is another shit you have to tweak if you use NVIDIA
>>9013622NVIDIA has a fucked up feature, when an app is using too much memory it will dump the rest on the RAM and when you are using most of your GPU asking your ram for its going to be shit cause it will FREEZE YOUR COMPUTER.when you do this fix only comfy UI will crash and your windows will plug and unplug your graphics card within seconds all it will be all fine (if you are watching a background video you might have to close it and open it again to get out of a green screen)>NVIDIA Configuration>NVIDIA Control Panel>3DConfiguration (there are two tabs global configuration and software configuration.>Software configuration>Select ComfyUI>Select High level processing NVIDIA>Specific Software Configuration tab>CUDA - System Fallback Memory>DISABLE IT>save the config>restart comfyUII think if you have AMD you don't need to bother with this shit.
>>9012964Sure :)https://files.catbox.moe/ve4c6f.pnghttps://files.catbox.moe/2nv6j3.pnghttps://files.catbox.moe/l6pv3w.pnghttps://files.catbox.moe/xoc08f.pnghttps://files.catbox.moe/rcplfq.pnghttps://files.catbox.moe/08cue9.pnghttps://files.catbox.moe/rk76zu.pnghttps://files.catbox.moe/deiv0j.pnghttps://files.catbox.moe/7ijx3z.pnghttps://files.catbox.moe/v45e3u.png
>>9013624They also made it very easy now, they have an .EXE installer with everything included, at most you will be asked to install GIT and it will be done automatically.If I missed something I'm sorry, when I formatted my PC and reinstalled I forgot about half the stuff I've written in these posts.remember its about making it work, not having encyclopedic knowledge about it.you might want to check this:>https://danbooru.donmai.us/wiki_pages/tag_groupsMost Models are trained with Danbooru, so you can look up what you want how to write it properly.
and even I'm not that good at it look at these filthy whores with the wrong bodysuits on and with gloves that aren't theirs.
>>9013277>https://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file#examples
>>9013277It also has default workflows for showing you how basic shit works.If you are afraid to read documentation your only chance is to copy and paste workflows until you find something that works for you, comfyUI isn't for people who put 0 mental effort, like you are asking the two of us instead of googling or queering to AI
>>9013540Agreed. I use Comfy for certain things like video workflows, but I use reForge for standard SDXL stuff. The lora management, out-of-the-box memory management, and inpainting are all just easier for me.https://github.com/Panchovix/stable-diffusion-webui-reForgeMinimum extensions to install:a1111-sd-webui-tagcompletecomfyui_stealth_pnginfo (despite the name it works for reForge as well)
>>9013702I spent 3 hours last night trying to figure it out, going through countless resources. I wasn't asking someone to do it for me. I was asking for some basic resources that explain the process of how to get started.
>>9013863That being said, I do appreciate all of the info you provided, thank you.
I tried to do the Halo meme with a goblin:>I need you to take a picture with your ass out and your back bent in with your pussy barely peeking through from a low angle with your face in it winking with your tongue out and your feet in it and your arch is exposed and your toes clenched and your torso at 3/4 rotation with your side boob and your nipples just out of sight with a choker on and your panties around your waist with the middle pulled to the side and your hands through the sides of the panties lifting them upward as your cheeks are outIt couldn't reconcile a way to have the panties pulled to the side with the hands in the sides of the panties lifting them upwards but all in all I think it did alright.Local ZLUDA install of automatic on an AMD using an illustrious model.
>>9013624Depending on exactly how much memory you're short on it can be a big help. Keeping it on lets you generate slower instead of crashing, and the negative effects on OS performance won't show as long as you have the spare RAM. I've been doing this since Pony release and never got a freeze like that.
>>9013627Catbox?
>>9014071https://files.catbox.moe/rrra9u.png
>>9013965Figured I'd give this a go since it's pretty much a bunch of danbooru tags strung together and it's kinda funny.By the way, aspect ratio greatly influences composition as you can see.
noob from earlier. finally got forge running. some models working ok. trying to get chroma to work. pasting images into png info and copying the prompts/settings/lora/ etc identically and it is just spitting out a garbled blurry mess and takes 10x longer than other models that are working fine. is it only good in comfy?
>>9014368i can't help you with Forge, but Chroma is a heavier model than SDXL. you can find FP8 and fast low step variants if you read the quantization section in the page here>https://civitai.com/models/1330309/chroma>https://civitai.com/models/1807633/chroma-unlocked-q4ks (low VRAM)This should probably be removed from the OP since it's incredibly slow, even if you have VRAM.
>>9013505Does this has metadata
>>9014401I have 12g vram and it isnt maxing it out when i look at the top usage data. I may just give up on it if the issue is power. it seems more like a settings issue though but wtf do i know haha
>>9014412>powerlol it's not that. but you need to figure out sampler/ideal step count/cfg for the model. forge might be busted IDK.you need to use image gen previews so you're not dumping wasted step iterations on images.
>>9014412I wouldn't waste my time with Forge for Chroma/Flux/SD3.5 and stick with comfy for that stuff. Forge is kind of a mess.
>>9010724Not even X company, but digital artists being openly ripped off.
>>9007405Oh please, can I see this animated?
>>9014556
>>9014558
>>9014564
>>9014565
>>9014569
>>9014570
>>9014571
>>9014574another great set
>>9014670More.
>>9014695
>>9014717
Shhhh! Don't wake her up!
>>9014556>>9014558>>9014564>>9014565>>9014569>>9014570>>9014571>>9014574Oh yeah give me that sweet bubblegum ass, princess!
A cute goomba explorer
>>9015003...Who give you a peak
Btw, a question to the thread, not sure if the Huntr/x girls i.e. Rumi, Zoey and Mira, are too new. Despite them having danbooru tags, the online generator I'm using doesn't gen them, as there is no auto complete for such tags. So the option is to locally gen them with Loras?
>>9013277just get the krita ai-pluginit's easy to set up and it downloads and installs all models for you
>>9015716It's a shame people have suffered brain poisoning from shitty web ui interfaces. Krita is so much better it's not even close.
>>9015718I've tried the Krita plugin and honestly I don't care for it. The workflow is a bit smoother when you want to make edits but there's some funkiness going on with inpainting and overall the prompting part is clunkier.
Bit the bullet and decided to install ComfyUI but every time i tried to generate a image i get this error>CUDA error: no kernel image is available for execution on the deviceCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.>For debugging consider passing CUDA_LAUNCH_BLOCKING=1Compile with TORCH_USE_CUDA_DSA to enable device-side assertions."i checked on the github forum but there doesn't seem to have a clear instruction on how to fix thisany ideas?
Big off model tits
Try to inpaint Nina Caliente from sims 4 and not bad!
>>9016051
https://civitai.com/articles/20211Civitai decided to kill its site already? lol
>>9016278>Making blue buzz even less usefulAnd the desired goal was?
>>9016306To appease the payment providers, of course.
>>9016278>Q: Wait, I have tons of Blue Buzz and only make NSFW content. Am I screwed?>A: Honestly? Yeahwelp
>>9016278Those niggas really just Tumblred themselves.
>>9016278Big deal. The only real draw was the broad pool of LoRAs and the cheap price of gens. Now they've basically neutered the blue buzz, the reason to bother with their shitass little website is gone.
>>9016278Download every models you like there because they will delete it soon
>>9015684Yup, Lora is the way to go with new stuff.
>>9016161more of this style please :D
>>9016562Shit. If people do take off their checkpoints/models, is there any other sites to look up and download them?
>>9005446Because Civ has shit the bed I want to locally generated. Only cartoons really. Can I use the models I downloaded for stable difusion?
>>9016948> https://www.tomshardware.com/pc-components/gpus/stable-diffusion-benchmarks>Pick one and install Stable Diffusionhttps://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file#installinghttps://github.com/lllyasviel/stable-diffusion-webui-forge?tab=readme-ov-file#installing-forge> https://comfyanonymous.github.io/ComfyUI_examples/
>>9015735> The workflow is a bit smoother That's an understatement. I can scroll my mouse wheel 1 notch up or down on the UI and select a model with preset loras and sampler settings.Takes less than a second.The plugin does some stuff with inpainting behind the scenes (e.g. like if the image context resolution is too small), but generally inpainting results depend on which model and sampler settings you use.
>>9017008shading looks nice. i can no longer tell what models people use just by looking anymore.
>>9017026Base Noob v-pred.I don't know any other model that insists on this kind of shading, for better or worse.
Why am I getting garbage t2i gens on tensor.art? I'm just using autismix with standard pony prompting but I keep getting magic eye results even with no loras.
Hey i was wondering, /aco/ ai here is for thing that LOOK like cartoon or ARE cartoon?
>>9016355NSFW is still allowed
https://files.catbox.moe/1aci8s.pnghttps://files.catbox.moe/tv9xr3.pnghttps://files.catbox.moe/bol34s.png
>>9017791For now. Currently they purging every Loli stuff in their site (even flat chested ones). Ill give it a year before they removing every porn.
https://files.catbox.moe/le6uyw.pnghttps://files.catbox.moe/ojludm.pnghttps://files.catbox.moe/hwz4cd.pnghttps://files.catbox.moe/x35osi.png
Bought a 5070tiHow good is it for locally running SD?
>>9018937Should be perfectly fine if I can manage with a 6800XT.
>civit killed itself>refuse to pay for porn out of principle>potato card I have no desire to upgrade from just from the occasional genIt's been a pleasure posting with you, anons. Maybe it's for the best. I noticed I stopped doodling completely when I could pretty much get any lewd idea that came to my head delivered to me in seconds. I should get back to drawing.
>>9018979Dude, there are more sites than just Civit.
>>9019012With free pretty much unlimited generations?
>>9016278lmao
>>9019016>unlimited generationsYou don't need that
https://civitai.com/articles/20844>we heard your feedback! NSFW on site generations stay!...as long as you have an active civitai membershiplmao, fucking clownsanyway here's my first local gen, not bad for a GTX 1050
>>9019199That's on a 1050??? I have the same card and I've been avoiding genning locally cause I thought it couldn't handle it.
>>9019408What you're able to do will be extremely limited, and you'll probably be relegated to SD1.5, though you may be able to run SDXL (illustrious, pony) if you have enough system ram to offload to. It won't be quick, but it'll probably work. You could also just do CPU genning which wouldn't use VRAM at all.
>>9019471addendum: you can get a used but functional RTX 2060 on ebay for like $115, which would triple your VRAM (6GB) and be a massive speed boost. For $200 you could get an RTX 3060, which would sextuple your VRAM (12GB) and be even faster.
>>9019486different anon but you can use this lora to really speed it uphttps://civitai.com/models/1608870/dmd2-speed-lora-sdxl-pony-illustrious
>>9019496Provided you can actually load it into memory in addition to the model
>>9019520lol good point. I didn't realize the 1050 is 2gb. absolute destitution
>>9019535Holy shit, just stick with SD1.5 at that point, the system baseline probably uses half of that VRAM unless it's a headless server.Seriously, SD1.5 is pretty much the only practical option for CPU only inference in terms of the time vs quality tradeoff.
>>9019535>>9019726I think a few step SDXL with some LCM distill (don't know which one is the meta atm probably the one linked above) can be doable on CPU inference (with partial GPU acceleration or not).It's a 2.6 model at the end of the day, even easier if using something like Q8.Provided it is not some ancient DDR3 shit.What are your current set up and token per sec anon?You may also want to keep an eye on nunchaku. I don't think it has easy inference for their SDXL quant rn but in the future it might let you fit SDXL into your GPU.
Wanted to post this at /g/ but I am not sure if janny would confuse her panties for pussy so here it is.
>try tensor.>Limited but fine.>Can't do NSFW prompts.:/
>>9019760Only old accounts can still do NSFW prompts unrestricted. You have to use tricks to get around the filter by misspelling text or grouping words together.
>>9019760yodayo.com supports NSFW, and if you're missing a model from CivitAI it's free and easy to upload it (LORAs at least, don't know about checkpoints), but you get only 50 free points per day with the average image costing 4
>>9020422also run by a chinese company based out of kowloon
>>9020839if I pretend I do not see, my privacy won't be threatened in any way.
>>9020839>also run by a chinese companySo? it's not like american companies don't do way worse shit
>>9019408>>9019471Yeah, it takes around 11 minutes to make one image in Ilustrious with DPM++ SDE, but i really don't mind.
Amethyst
>>9021073Can you do Marcy topless in a pair of jeans please?
>>9021290Buttjorb
>>9021292Which way, western man?
>>9021293Front
>>9021294...or back?
>>9020422tried, pretty dissapointing.
>>9021291Sure
>>9021472Pretty good! Thanks!
>>9021482you're welcome
>>9021573God I love her big eyes!
https://files.catbox.moe/50vyi4.pnghttps://files.catbox.moe/51x5m7.pnghttps://files.catbox.moe/k30wal.pnghttps://files.catbox.moe/mnc4pn.pnghttps://files.catbox.moe/xbgmvp.png
>>9022053
>>9022054
>>9022055
>>9022060
>>9022062
>>9022063
>>9022065
>>9022066
>>9021573NTA but can you make one where Marcy loses her swimsuit at a public pool and looks all embarrassed?
>>9011307Moar Mavis please
>>9022077out of free credits, will try tomorrow
>Devil's Candy finally got a loranice. shame the smutty gens go off-model with making Pandora buff, though.
Hey what do NTA mean, as i always thinks it mean Not The AssholeAlso tried with different styles
>>9022378'Not that anon' but same thing.
>>9005446Think anyone can make a gen of Wondy & Harley dressed up as sluttier human women versions of Optimus Prime and Bumblebee from the Michael Bay Transformers movies?Its just them wearing revealing armor that keeps their lady parts covered
>>9022381Aaah ok, thank bro! Want a free request?
>>9022417Could you try that Pandora lora and keep it on-model? Casual nudity and uninterested sex would be great.https://civitai.com/models/2048876?modelVersionId=2318813https://devilscandycomic.tumblr.com/tagged/pandora
>>9022516Here ya go! Also have no clue who she is but she cute
>>9022551She's from an old webcomic / western manga, by the same artist from The River City Girls game. There's a few cute girls in it, but probably not much material to train loras on.
https://files.catbox.moe/wm05co.pnghttps://files.catbox.moe/e9bm5d.pnghttps://files.catbox.moe/b34p5r.pnghttps://files.catbox.moe/h40056.pnghttps://files.catbox.moe/ynyqf9.png
>>9022568Oooh neato! Also I didnt LORA but NAI cause that the one im more used to
https://files.catbox.moe/219dr3.pnghttps://files.catbox.moe/g7zl7v.pnghttps://files.catbox.moe/i4j3nk.pnghttps://files.catbox.moe/80l6zs.pnghttps://files.catbox.moe/riu4m6.pnghttps://files.catbox.moe/ds6kf5.pnghttps://files.catbox.moe/6x1b0s.pnghttps://files.catbox.moe/ylul5r.pnghttps://files.catbox.moe/nbose7.pnghttps://files.catbox.moe/j7r4sy.pnghttps://files.catbox.moe/pwrj4t.pnghttps://files.catbox.moe/nyxek1.pnghttps://files.catbox.moe/uzke39.pnghttps://files.catbox.moe/rjrkg4.png
>>9019197
Mikasa found a way to speed up eren's titan regeneration.(I'm trying my hand at the sketch inpaint feature in forge, pretty cool, is there anyway to improve? maybe hook it into photoshop or something, that would be absolutely perfect)
>>9024269>maybe hook it into photoshop or somethingyou know about krita?
>>9024169
canonically this is what happened between henry and katherine after their escape
>>9024473I'm a retard
Panda.
Multi-Panda
captain undergoing a twisted endurance test meant to test focus under duress
lux working overtime to support her carry
Lara got busted unfortunatelynow she's trying her best to get away by bribing the guard who saw her
>Lara was always nice to her mates in university and pretty quick to feel and loosen tensions when most needed.Trying to find a style that I like with illustrious, thoughts?(I added shit with img2img so just ignore that)
I was kinda curious to see how 'active' the /aco/ /sdg/ was but it looks like its nowhere near as active as any of the other threads from other boards are. Is it always like this?
>>9026681yes
