Discussion of free and open source text-to-image modelsPrevious /ldg/ bread : >>101882089>Beginner UIEasyDiffusion: https://easydiffusion.github.ioFooocus: https://github.com/lllyasviel/fooocusMetastable: https://metastable.studio>Advanced UIAutomatic1111: https://github.com/automatic1111/stable-diffusion-webuiComfyUI: https://github.com/comfyanonymous/ComfyUIForge: https://github.com/lllyasviel/stable-diffusion-webui-forgeInvokeAI: https://github.com/invoke-ai/InvokeAISD.Next: https://github.com/vladmandic/automaticSwarmUI: https://github.com/mcmonkeyprojects/SwarmUI >Use a VAE if your images look washed outhttps://rentry.org/sdvae>Model Rankinghttps://imgsys.org/rankings>Models, LoRAs & traininghttps://civitai.comhttps://huggingface.cohttps://aitracker.arthttps://github.com/Nerogar/OneTrainerhttps://github.com/derrian-distro/LoRA_Easy_Training_Scripts>Fluxhttps://huggingface.co/spaces/black-forest-labs/FLUX.1-schnellhttps://comfyanonymous.github.io/ComfyUI_examples/flux>Pixart Sigma & Hunyuan DIThttps://huggingface.co/spaces/PixArt-alpha/PixArt-Sigmahttps://huggingface.co/spaces/Tencent-Hunyuan/HunyuanDiThttps://huggingface.co/comfyanonymous/hunyuan_dit_comfyuiNodes: https://github.com/city96/ComfyUI_ExtraModels>Index of guides and other toolshttps://rentry.org/sdg-linkhttps://rentry.org/rentrysd>GPU performancehttps://vladmandic.github.io/sd-extension-system-info/pages/benchmark.htmlhttps://docs.getgrist.com/3mjouqRSdkBY/sdperformance>Try online without registrationtxt2img: https://www.mage.spaceimg2img: https://huggingface.co/spaces/huggingface/diffuse-the-restsd3: https://huggingface.co/spaces/stabilityai/stable-diffusion-3-medium>Maintain thread qualityhttps://rentry.org/debo>Related boards>>>/h/hdg>>>/e/edg>>>/d/ddg>>>/b/degen>>>/vt/vtai>>>/aco/sdg>>>/trash/sdg
Reposting because new bread: qrd on forge and why should i use it over comfy?Also can Flux do controlnets yet?
>>101886105Flux is easier to use, more retard proof.The reason you should use it over comfy is personal preference.
>>101886105>Also can Flux do controlnets yet?https://huggingface.co/XLabs-AI/flux-controlnet-collections
Blessed thread of frenship
>>101886155Facts
>>101886139Alright, what am I doing wrong? Surely this doesn't only work on square images?
Teebs gens are like an untouchable dysgenic from Dalit was given a PC with flux installed to gen on, you cannot refute this
This is great for converting my old SD images into a Flux boomer prompthttps://huggingface.co/spaces/fancyfeast/joy-caption-pre-alpha
>>101886347What is teebs?
>>101886470genuinely curious, are you the dev promoting your shit?i don't mind either way since the tool is really good and should probably get put in the sticky, but i've seen you post a link in the past 5 threads now
Day 1 of learning AI image generation in order to catfish people online. Wish me luck Anons. Hopefully Flux will be good enough to pull it off.
>>101886139>sample images all blurrylol
>>101886562>are you the dev promoting your shit?No. I found it randomly while trying to figure out how to caption images for the LoRA I want to train. This is literally the first time I have posted this particular link.
>>101886629Poor brunette has a lisp
>>101886105forge is lovecomfy is passion
Flux with Grok is quite based
>>101886584You can't. Flux can't make consistent character
The person who gets vanned for posting certain images has it coming.
>>101886772What's the difference with the one we have in our home? it's using the pro version?
>>101886805Skill issue.
waiting for flux on reforge
>>101886805>Flux can't make consistent characterwith InstantID it'll be easy enough
>>101886821Enjoy default dreamshaper face
>>101886470Man image recognition has improved like crazy over the last few years.It's absolutely nuts how well this rambling machine speak understands itself. It's damn good at replicating the results all things considered.Pity it's not exactly all that intuitive for a human to talk or think like that, not all that easy coming up with all of the descriptive fluff bullshit that's apparently required for best results.
>>101886843>InstantIDwhat is this?
>>101886843wrong
>>101886869With just one picture you can make any kind of variant, no training needed no anything, only works for SDXL so far thoughhttps://instantid.github.io/
>>101886866>Flux can't make consistent character>>101886866>Enjoy default dreamshaper facePick one, retard.
>>101886905Lol stupid
>>101886867It's easy if you are a fiction writer. I am not. But I'd imagine it would be.>>101886891Oh this, yeah. It's heavily Chinese biased though but could be the future of Character and celebrity LoRAs. Just a single image and transfer it to whatever gen.
>>101886954It's your fault for trolling so ineffectively.
you can get vanned when there is intent behind pictures, Amazon pictures are to sell swimsuits, we both know even by the camera position what your pictures are designed for, so yes, you can get vannedalso please, keep being a nuisance so they are forced to come up with a permanent solution for you
>>101887110okay anon have fun in prisononly a matter of time and you're so unstable you're going to be spamming nudes soon
It's completely legal to castrate yourself with a rusty pair of shearing scissors but it doesn't mean you should do it.
what is the difference between this thread and sdg? both are just stable diffusion threads no?
can't wait to have to login and verify to use 4chan
>>101887198/ldg/ has a resident pedo spammer/sdg/ has a resident thread squatter who has shit up the thread daily for the past 2 yearspick your poison
>>101887198Barely anyone here uses Stable Diffusion. /sdg/ is a discord for avatarfags mainly.The intent of this one is to be about all local models, not just Stable Diffusion.
>>101887198>stableno, we only run Flux now
>>101887245dare you to post them
I miss early 2000s rave girls high on e and grinding on anything with a pulse. Take me back
>>101887219>pedo spammersounds like the thread for me, thanks>>101887242>>101887244I've been out of it for so long I just figured SD was still the main model or main talking point
>>101887303It was the main model. In fact, it was the only local model for quite a while. We are now spoiled for choice, and everything SD has ever done has been completely mogged by Flux
>>101887303SAI shit the bed since SD 2.0. They have been handwringing about safety for so long they ran out of money, scared away all their engineers (who in turn, they made Flux), and SD 3 was a complete and utter disaster that can't even make females without making monstrosities.
Can someone please share flux inpainting workflow?
@101887332this coomer being saying the same shi for 2 years opinion discarded
>>101887407I'm concerned anon, did you have a stroke? SAI hasn't been relevant since SDXL and honestly without Pony, no one would be using SDXL either.
@101887443u flinging insults really makes you point coomer kek
>>101887443the NSFW POV lora for XL was pretty good and it came out way before Pony, you couldn't get things that consistent with 1.5
>>101887482sirs that is not how you bloody ping someone sir!!!
>>101887198There is no difference. You can post the same gens and talk about the same tech in both. A discord group split the general because they couldn't figure out how to share a thread with anons they disliked. Now they wage a culture war 24/7 over it because they have nothing else to live for.
>>101887535It really doesn't look that good, it looks like AI and everyone has the same body.
<3
>>101887585well that's just how it was then also the images people are making with it on civitai are not the best it can do
>>101887563Flux is what SD3 should have been desu
>>101887198one is gay (bara) the other is gay (bishonen)
>>101887607That wouldn't be very safe, Anon.
SD3 truly was the last stable diffusion model
>>101887563Did you see Lykon posting a woman on the grass as if it was some massive achievement
>>101887640I'm actually shocked they haven't released SD3 8B because Flux otherwise is uncontested.
>>101887648don't you get tired of arguing with yourself 24/7?
>>101887648thank you christianity for making our hecking wholesome moral system that hates and denies the sanctity of the human body
did pedo anon move from 6 year olds to jailbait, or do we have multiple schizos now?
>>101887640Cascade 1.6 is coming and will be runnable on HD5670 1024MB, FLUX is finished, finalised sir!
>>101887675it should be pretty obvious he's mentally ill
>>101887685SIR, SIR PLEASE CONFIRMDO THE NEEDFUL STABILITY
>>101887685SAI tried selling the idea that you wouldn't have to train one of the stages because "you wouldn't train the VAE" when that stage did a huge amount of work on the image lmao
>>101887728It is an interesting idea for architecture, a sandwich model. VAE -> Small Main Concept Model (expected to train) -> Universal Upscaler
>>101887706Its weird how you didnt answer the question
>>101887772judging by the timestamps it is the same person, also the style is a dead giveaway, he's also mentally ill so I expect him to samefag with multiple devices
tech or get tae fuck
>>101887675i just like cute girls anon, no need to over analyze like meta-schizo does
>>101887657I think stability was genuinely afraid of upping the requirements to run it, people did complain they couldn't run flux on their shit machines but that got drowned out by the model being good If you build it, they will come, ignore the seethers being gatekept
>>101887822>upping the requirements lowered the # of literal retards posting bad gens with your modelgenius actually
>>101887822It's just another example of them being handwringing retards. Also they should have the money to do the fp4 research, that's why it's all retarded. Every day it's clear they're just grifters afraid of losing the bag.
>>101887822chad mistral dropping 123B LLMs on us and saying just get a better machine
>>101887822their 8b model is a big boy though, and it was supposed to be given to us. At the end of the day, their way of training models is so cucked it could've been a 1000b model it wouldn't reach flux's level, they are just a bunch of retarded niggers there's nothing we can't do about it
>>101887822>people did complain they couldn't run flux on their shit machinesAnd now, 2 weeks later, even vramlets are running it. Trust the community.
>>101887917It's not like anyone did anything novel, SAI should have the researchers and talent to release 8B *with* an FP4 model and not just an FP4 model, but something finetunable.
newfren here, how the FUCK do i restrict controlnets to POSE ONLY (flux if it matters)i don't need it taking in other aspects like thickness of arms/body (say, if i wanted to add clothes to a character without them) because then it distorts the proportions and makes things look inaccurateall i need is for the hands (and maybe arms) to be positioned and posed how i want them to be
>>101887984wait for a OpenPose controlnet for Flux
>>101887888I used their 8b model via api. It wasn't bad, and it should have been the one they released.>>101887958Well they lost any and all talent they had and now we have BFL so it was for the better. Mismanaged dogshit company gets whats coming to them. Bankruptsy.
>>101887958they knew about fp8 for a long time, yet they cancelled the 4B model and shoved the 2B one down our throats. they knew exactly what they were doing.
>>101888022this
>go to reddit while the boards were frozenDamn, they really do post some bad gens there, huh?pic rel, the first thing i see on an album of "this is all created with flux" like this is not a disgusting failed gen
>>101887984Controlnets aren't working properly with Flux yet, it's only been out for 2 weeks so people are still figuring shit out. For now you can just try to prompt the poses. It's decent at most normal and basic poses.I did see one workflow of a guy who had comfy running in Blender and was posing a dummy and transferring the pose to gens, but I can't remember where I saw it.
>>101888042Reddit is basically soiboys, I've never really seen a good gen there, even those retarded billion pixel gens are fucking stupid and soi. 4chan image spam is at least somewhat tolerable but it's incredible what floats to the top on Reddit showing how doomed we are.
>>101888042I can fit my whole body between her titty chasm
>>101888071Don't worry about Reddit, they represent a very minor portion of any community. The biggest concern is Google training off of Reddit. Gemini is going to be the soiest of soi models.
i'm starting to hate people now, why can't I just stay at home comfortably genning? sigh
>>101886470Anyone knows an equivalent I can run locally in comfy (also without any bullshit censorship if there is any) ?
>>101888176Why is this so sexual? Can you share your prompt?
I'm going to bed. Night, retards.
>>101888245goodnight fellow retard
>>101888210Exactly.I'm using Flux, and i described her as a harlot looking passionately or seducing a passerby, no magic here excluding the magic of wordsLook at this!
>>101888335kys pedo
Any way to bypass this early access shit?https://civitai.com/models/61099/perfect-full-round-breasts-and-slim-waist
>>101888097>just stay at home comfortably genningIt's really nice getting lost into this isn't it? Which is really starting to make me wonder about the future.The moment this tech gets good enough and we can pump realistic real time world generation directly into our heads, will we just enter the matrix and never come out?I genuinely think that is the end game for humanity.Everyone ends up doped up living in some pod experiencing millions of new artificial fantasies all day everyday, or maybe it'll be more like a global Habbo hotel where everyone has their own room that people can visit etc..Matrix really was a soft disclosure about our future.
>>101888392it's shit, just wait for something better
>new nf4 quant>reportedly slightly slower + takes up more RAMwhat is he doing??https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/1079#discussioncomment-10335654
16GB VRAM should be a hard minimum
>>101888426>ilyas gatekeeping vramlets just so he can double down on his claims that NF4 is better than fp8kek, respect
>>101888426it is the same speed, takes 0.5 more GB, it is closer to FP8
I might subscribe to X to get Flux. Has anyone here tried it?
oh fuck, i just realized elon using flux for x means we're gonna get even more retarded techbros in the space>>101888509
>>101888526retarded techbros are $$$
>>101888526It has already startedhttps://x.com/search?q=grokNormalfags don't even know what Flux is. They think Grok generates the images
>>101888509Yes we have tried it. It is good. Cloud services are off topic thoughbeit
>>101888554thats how mafia works
>>101888554Who knew AI was going to win the Meme War of 2024.
stop giving the pedo attention, ignore and pretend they don't exist
>>101888392Just wait anon.
>>101888554>They think Grok generates the imagesI mean it's pretty clever.
>adetailer penis model cant detect big juicy flaccid uncut cock
>>101886805I'm not really worried about this honestly, more that the pictures won't be realistic enough. Aside for maybe 1 or 2 pics that I've seen everything that's generated looks like some stock photo shit
>>101888751americans are truly the bane of porn
>>101888509You can, but what's the point? It's going to have image filters on top of the removed dataset, at that point, you get a worse dalle that you have to pay for. The advantage of flux is that it can become something more meaningful through community effort.For example, flux itself doesn't have quality tags, it has quality phrases, due to the dumb way the llm tagged the dataset. This can be fixed with finetunes, while those still using the main model will eventually realize how generic the model output is, despite its great adherence.
>>101888769because everyone overly trains on stock photos instead of using Flickr
>>101888812Gotta be fast and collect datasets
>>101888812>instead of using Flickrmy flickr set is up to 216m desu
whats the minim images i need for an oc character lora? is it possible to use two or three images to make a shitty lora that helps nudge it in the right direction and makes it easier to create a set of 20 images or so?
Am I retarded for boomer-prompting a bad image in the negative? Isn't that how it works?
>>101889025>It It It It>And And And Andlrn2prompt
>>101889044>doesn't know how flux T5 prompts worksit's okay anon i'll wait for someone else to answer.
>>101889068vlms do not describe images like this so most of your tokens are wasted
>>101889025dunno you tell me, does it work?do you get better results than without it?
>>101889090>magnetnahalso kys pedo
>>101889025you are moving away from all concepts related to said tokens, and since the negative prompt contains many generic concepts and tokens also used to describe good images, it will do more harm than good.
>>101887822The high requirements has been the greatest shift to happen to image gen AI in a while. Not only does it allow for better models but it limits the amount of images poor 3rd worlders can poop out, and even then they’re stuck with lower quality quantizations. It’s a win across the board.
>>101889025you have to say please for it to work
>>101889157It'll also finally kill the need for a billion shitty Loras and no anime/realistic model split, 12B is enough brainspace to hold basically every concept you'd want. Flux could be the model that lasts five years until we can run 24B models.
>>101889087>>101889081>>101889108You see, that was my reasoning as well. But every time I've made comparisons, it's always led to better results.https://imgsli.com/Mjg3NTU4
>>101889259Is it better? if you were just aiming for a realistically ugly middle aged man they both got it spot onthe left is just blurry.
There is a rumor that something big is gonna happen at civitai soon. New kind of policy of what is allowed and what is not and it would nuke sizeable portion of the uploaded models etc from the site.Have you already downloaded and archived all you want and need? There is a good chance that soon what you want is not there anymore.
>>101889259outside of blur it's impossible to comparethe left image is actually better
>>101889275Its funny i spammed the shit out of their servers with failgens labeled with variations on YWNBAW and i wasn't even touched for it but i hear they'll randomly ban you for the pettiest reasonsi wouldn't be surprised if that rumor is correct, but everyone already should've been saving and backing up whatever they like this whole time, that platform's been a ticking timebomb for ages.
>>101889275>source: my ass
>>101889275>>101889293Shit thats not good news at all
>>101889275most models and loras are ass so the question is if models like Pony are affected or notif it's just a crackdown on celebrities or whatever, whatever
>>101889275What's the source of that rumour?
>>101889275>torrenting models and lora will become a thingwild times ahead
>>101889327his gaping asshole
>>101889275>civitai goes down literally one second after this postlmaoo
>>101889349holy shit
Damn, It's not just normies on twitter that don't know Grok is using flux, It seems to be MOST of twitter. Even people i follow that should know better are saying "grok made this!"Huh. Interesting. I wonder what the further implications of this will be.. Hopefully nothing incredibly disastrous!
>>101889335we all know it was a matter of time until they cracked down on overt porn posting and celebrity loras were always going to be a problem that doesn't seem worth the hassle
>>101889349well shit
>>101889349oh FUCK
>>101889363anon they literally made multiple categories you can participate in to classify images in different age restriction brackets.and by default they blur/hide any nsfw
>just purged all my models to save space ohnoohnoohnoohno
>>101889362so?
>>101889380>we all know it was a matter of time until they cracked down on overt porn posting and celebrity loras were always going to be a problem that doesn't seem worth the hassleI'm not changing my assertion. If you don't like it, you can make your own porn-obsessed Civit and find out.
>>101889406How would you have felt if you didn't eat breakfast this morning?
Is it owari?
>aitracker.art devoid of image generation models We did this to ourselves
>>101889455it hasn't even begun, we were born too early really
>>101889362Who cares, things will pan out eventually. If you get too hung up on this stuff some metaphysics-brained nerd will explain to you how saying "flux made this!" is equally wrong.
>>101889438that's Nona, talk to him
>>101889349What a curse to be a prophet who can only see a minute into the future.
>try a non-binary cope prompt for big tits from earlier today>it doesn't workDamn
>>101889268>>101889278I see what was happening now, my cfg was too high which was making any blank negative prompts return a blurry image.
>>101889543base models have such a severe lack of real saggy breasts
>>101889486I'm really not giving this any more thought than that post, at worst i was thinking "muskrat is using this to get flux very banned at federal level which domino's into all of open source A.I', i figured this point was a lot more obvious but my bad for assuming i guess.We'll see.
>>101889362Twitter is like a different dimension, everyone on there has severe brain rot. It used to just be leftist and commies, but now there's also the entire musk dick riders. It's a terrible place to be.
>>101889543It never worked for realistic images, try an artwork or something
>>101889543works on my machine.
is it too big or is it okay?
>>101889613The genie is out of the bottle, they will have to ban the code and ban GPUs. I have everything I need with my computer right now to make a 2B model from scratch, it would just take forever.
can flux make penis
>>101889692die faggot
>>101889718you need to ignore that guy
>>101889703it has never seen even the hint of a penis
>>101889672The professional photograph features a high-quality, original depiction of a non-binary person with tanned skin. They have long, pure white hair. They have huge breasts, which are so large and heavy they sag - and a tiny waist and very wide hips. They are adorned in a skimpy lingerie, which has a very large amount of cleavage. They are holding up a sign that says, "non binary btw".With a high cfg
>>101889656Even if GPUs are banned eventually everything will be done with CPUs and RAM. The new unified memory Mac architecture will be standard soon enough.
Is there a publicly available dataset of live-action porn with natural language captions?>>101889362It's no different from the generals on /tg/ and /trash/ referring to DALL-E as Bing.
>>101889742whats the point then
>>101889790none. bye.
>>101889755the distinction is important, the gpt+ version and the azure endpoint versions are different enough to justify it. before microsoft unified it, there was also a distinction between bing, designer and copilot (higher token limit)
>>101889745>it worksthey really had to make their model safe and effective isn't it?
>>101889703There's a LoRA for it:https://civitai.com/models/628763/androflux
>>101889987Lights should be dimmer, but I'd still have a drink there
>>101889999Understandable
>>101889208that might be true but for us to have a single model with all the celebs, characters, styles, fetishes, etc that are separate loras right now you'd need a group to collect all those data sets and train the model on them
>>101889648could be bigger desu
You can't handle the truth.
I am the truth.
No I am the truth.
Is there an fp16 version that works on Forge? I've only seen the NF4 quants. I have a 3090 so I could load just fine the fp16 dev model.Why does the full model don't work? Will it eventually or is there any architecture incompatibility?
>>101890289wait so anons here have been shilling Forge and it doesn't even support the best model?AHAHAHAHHAAAAHA
Truth is for the little people.
>>101890305Bestest, not firstest.
>>101886818Using pro and just like Dalle using some sort of prompt injection to "enhance" your prompt. Leads to less control over what it gens. Nothing you can't replicate locally.
>>101890306short kings
what image dimensions are the go-to for training flux loras? I remember someone saying 512*512, is that true? should I be downsizing my sdxl dataset for optimal results?
run girl,run for your life!
>>101890410what did she do
>>101890373what, no, vramlets are doing it at that res so they can train the lora at allFlux can adapt because it was trained on a range of pixel counts but preferably you'd target 1 megapixel
>>101890456>double belly buttonPFFFFFTHHAAHAHHAHA
>>101890524die in a fire
>>101890540Cunny poster bro don't make her wear "blacked" bikini make her wear a bikini that says "blacked posters are jews with AIDS"
how many IPs does this guy have?
>>101890631to many
>>101890631probably a turd worlder who can just reset his internet and get a new one. if hiromoot cared they'd rangeban all of india/SEA and we'd all be better for it
>>101890631over nine thousand
>>101890661no he is german.. he admited to that previously
I'm reading this guide and I don't understand why if i want to run flux schnell I am not downloading the files directly from black forest labs huggingface repo.https://comfyanonymous.github.io/ComfyUI_examples/flux/
>>101890682why are germans like this?t. knows a german who did things with a girl at the minimum age of consent and tried to justify it, absolutely sickening
>>101890682aka turdworld immigrant
>>101890682turd worlders often larp as first worlders.
>>101890681you are sick>>101890717I am also German but I am not sick
>>101890717>minimum age of consentits 14, and legal, and ethical. you are supporting feminist propaganda.
>>101890741the pics you post are not 14
>>101890741>feminist propagandaCall it what it actually is. It's Jewish anti-natalist propaganda. Ridiculous AoC laws and women's rights are disastrous for White population growth..
I will post pics I've already posted
>>101890790what a shizo post
Has anyone trained a lora on 12GB yet?
>>101890859I say it again: Die in fire
>>101890848there's no way that's possible
>>101890428Tried to steal documents from the local castle . Exiting from the front door wasn't the right choice.
>>101890888everything is possible, it's possible to train these models like onions, it would just be extremely slow (which is the catch of all this)
>>101890881nice
when there's only one exit , you just have to run
>>101890971you're being purposely annoying at bestall you show is you have no friends
stop. engaging. the. pedo.
what if we close /g/ for a day, he'll get bored and leave
>>101890848I'm trying to set it up right now, bmaltais repo has it in the GUI now if you're a retard like me who doesn't like typing out all the training commands for kohya in command promptkohya says>--optimizer_type adafactor --optimizer_args "relative_step=False" "scale_parameter=False" "warmup_init=False" --split_mode --network_args "train_blocks=single" are what to use for 12GB vram
>>101891019>uhmmm it's totally normal for a guy like me in his twenties to be hanging out with a 14 year old child, not creepy at all okay?????? IT'S HECKIN' LEGAL OKAYY???KYS
>>101891017I wanna see how many bans until his ISP gets range banned
>>101891017its some obese pedophile nolife retard from india or some other shithole, he has no other place to go but the internet to escape his sad existence.but see it positive, at least while he is wasting his life here he isnt out there molesting real life children.
>>101891025suppose I thought that this was impossible 10 minutes ago, what's the pipeline?how do you tag a cache of images, I assume WD moat is no good
>>101891052don't get b8ed m8
Just ignore and report anon, it's not that difficult.
>>101891052As long as they're exclusive and get married it's fine.
>>101891103hehe kitty
>>101891190>kitty
>>101890937thanks
>>101891068not having tried it yet as I'm still downloading the dev checkpoint for training, same as normal, just change your bmaltais or kohya repo to the one that says 'flux' in the name. as for tagging, some people have been using >>101886470I'll probably just use chatgpt on non-nsfw, running a regular booru tag'd existing dataset first for initial test just to see that it runs desu https://github.com/kohya-ss/sd-scripts/tree/sd3https://github.com/bmaltais/kohya_ss/tree/sd3-flux.1
>>101891117Maybe in first world countries but this is a 3rd world board so gtfo
Is the Grok flux different? cuz it can do celebs
>>101889275i have all the loras
>>101891354could be based on an uncensored version of Pro
>>101891354Might be the pro model, Black Forest Labs posted about their partnership with Grok so they might have given them the pro model or another version of it.
>>101891354which celebs can it do that FluxDev can't? Haven't seen any so far.
>Grok is currently not available in your region or countryyou made me click a million checkboxes and agree to have all my data stolen for that, Elon?
>>101890971I say it again: die in a fire
>>101891441If you want to try it out for free, use LMSYS. The Grok 2 model on there is called sus-column-r but Flux is not included, so it's only the LLM.
>>101891441What 3rd world country are you from?
do. not. engage.
>>101886470>thow old anus1111 gens into this >allahu akbar!>>101890737based german landschaftengenerierer
>>101891557She looks like she has balls.
>>101886057which of the retard-proof uis is the best guys?
>>101891602for flux, unironically grok
She safely reached the town outskirts
lets be honest for a momentthis is still better than sharing a thread with the thread schizo
>>101891675what was in those documents?
>>101891687color me suspicious as to why only this thread has a ban-evading pedo
>>101891687kinda painful to see samegen noskill gens tho ngl
Let's move on to some fresh bread...>>101891791>>101891791>>101891791
I sure hope finetuners don't touch T5I sure hope every Flux finetune on civitai isn't a fucking 20GB file
>>101891718Oh shi-
>>1018878761t/s on cheap ddr4 128gb with speculative decoding is fast enough and easy and cheap for any consumer pc today
