File: da noos 5.jpg (628 KB, 1344x768)

628 KB JPG

>mfw Resource news

06/21/2024

>Stability AI appoints new CEO, Receives Ballout
https://www.reuters.com/technology/artificial-intelligence/stability-ai-appoints-new-ceo-information-reports-2024-06-21

>Programmable Motion Generation for Open-set Motion Control Tasks
https://hanchaoliu.github.io/Prog-MoGen/

>Advancing Fine-Grained Classification by Structure and Subject Preserving Augmentation
https://github.com/EyalMichaeli/SaSPA-Aug

>Consistency Models Made Easy
https://github.com/locuslab/ect

>Invertible Consistency Distillation for Text-Guided Image Editing in Around 7 Steps
https://yandex-research.github.io/invertible-cd/

>WATT: Weight Average Test-Time Adaption of CLIP
https://github.com/Mehrdad-Noori/WATT

>UHDDIP: Ultra-High-Definition Restoration: New Benchmarks and A Dual Interaction Prior-Driven Solution
https://github.com/wlydlut/UHDDIP

>Taggui v1.29.0 Adds Florence-2 Support
https://github.com/jhc13/taggui/releases/tag/v1.29.0

>CosmicMan: A Text-to-Image Foundation Model for Humans
https://github.com/cosmicman-cvpr2024/CosmicMan

06/20/2024

>Comfy Org launches comfy.org
https://www.comfy.org/

>ComfyUI-UniAnimate
https://github.com/AIFSH/ComfyUI-UniAnimate

>SD-Latent-Interposer Adds SD3 Support
https://github.com/city96/SD-Latent-Interposer

06/19/2024

>Florence2 in ComfyUI
https://github.com/kijai/ComfyUI-Florence2

>Ilya Sutskever, OpenAI’s former chief scientist, launches new AI company
https://techcrunch.com/2024/06/19/ilya-sutskever-openais-former-chief-scientist-launches-new-ai-company

>ComfyUI_frontend: ComfyUI modernized
https://github.com/huchenlei/ComfyUI_frontend

>RunwayML Gen-3 Alpha
https://runwayml.com/blog/introducing-gen-3-alpha/

>ComfyUI-LuminaWrapper
https://github.com/kijai/ComfyUI-LuminaWrapper

>TroL: Traversal of Layers for Large Language and Vision Models
https://github.com/ByungKwanLee/TroL

>diff-sampler: open-source toolbox for fast sampling of diffusion models
https://github.com/zju-pi/diff-sampler

>mfw Research news

06/21/2024

>A Survey of Multimodal-Guided Image Editing with Text-to-Image Diffusion Models
https://arxiv.org/abs/2406.14555

>SafeSora: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset
https://arxiv.org/abs/2406.14477

>Video Generation with Learned Action Prior
https://arxiv.org/abs/2406.14436

>CollaFuse: Collaborative Diffusion Models
https://arxiv.org/abs/2406.14429

>VLBiasBench: A Comprehensive Benchmark for Evaluating Bias in Large Vision-Language Model
https://arxiv.org/abs/2406.14194

>ExVideo: Extending Video Diffusion Models via Parameter-Efficient Post-Tuning
https://arxiv.org/abs/2406.14130

>From Descriptive Richness to Bias: Unveiling the Dark Side of Generative Image Caption Enrichment
https://arxiv.org/abs/2406.13912

>GenAI-Bench: Evaluating and Improving Compositional Text-to-Visual Generation
https://arxiv.org/abs/2406.13743

>StableSemantics: A Synthetic Language-Vision Dataset of Semantic Representations in Naturalistic Images
https://stablesemantics.github.io/StableSemantics/

>IntCoOp: Interpretability-Aware Vision-Language Prompt Tuning
https://arxiv.org/abs/2406.13683

>Trusted Video Inpainting Localization via Deep Attentive Noise Learning
https://arxiv.org/abs/2406.13576

>Is AI fun? HumorDB: a curated dataset and benchmark to investigate graphical humor
https://arxiv.org/abs/2406.13564

>4K4DGen: Panoramic 4D Generation at 4K Resolution
https://4k4dgen.github.io/

>ARDuP: Active Region Video Diffusion for Universal Policies
https://arxiv.org/abs/2406.13301

>AniFaceDiff: High-Fidelity Face Reenactment via Facial Parametric Conditioned Diffusion Models
https://arxiv.org/abs/2406.13272

>Convolutional Kolmogorov-Arnold Networks
https://arxiv.org/abs/2406.13155

>SwinStyleformer is a favorable choice for image inversion
https://arxiv.org/abs/2406.13153

>Sampling 3D Gaussian Scenes in Seconds with Latent Diffusion Models
https://arxiv.org/abs/2406.13099

>SAI purchased by Weta FX
>An investigation into Wētā's workplace culture by Kiwi public TV broadcaster 1 News...
>In their testimonies, workers identified the existence of a male-only pornographic mailing list called "Caveman", which originated in 2002 following a company-wide tradition known as Porn Friday
>Several reports also alleged that the company's IT systems required upgrades in order to accommodate the volume of pornographic content hosted on the company intranet
Boys, we're so, so fucking back and you don't even realise it.

Cursed thread

Comped thread?

High vibes thread, good mood and friendliness to all. Welcome

fixed groupnorm for ndhwc and fixed conv3d. neat

>>101092036
no work only gen post catgirl

despite how wacky sd3 medium is with humans, theres no way its as useless as people want us to think when it can follow your directions.

File: file.png (781 KB, 704x512)

781 KB PNG

>>101092060
is a humangirl ok

>>101092139
fiiiiiine

>>101091979
Weta only makes closed source software and has been holding back the VFX industry for years. They will probably change the business model of SD instead of improve the current free model one.

File: 1712331969502359.jpg (75 KB, 700x860)

75 KB JPG

So things like ComfyUI, Automatic1111, Forge, and SD.Next are just differences in GUI? I see at the very bottom of my Forge tab it is running on v1.9.4 of SD? So does that mean these GUI's use their preferred version of SD? But I am also running whatever models I want (that are compatible)..? If everything is baked into the models then and the UI stuff is just preference or no? Different models just require different settings/tweaks/prompt rules? I was getting OK results with T-ponynai3 but then switched to Pony Diffusion V6 XL and got whacky results with same settings.

Is Stable Diffusion just like the brand name for AI stuff? Like the Xanax of alprazolam?

File: file.png (670 KB, 512x704)

670 KB PNG

>>101092188
i dont gen often, next time ill give her cat ears

>this is bait

Reminder regarding the doxing discord
>>101085325
>Deanonymizing (to an extent) this general was necessary to get some folks to be more peaceable.

>this thread is just a handful of anons spamming and shit stirring

>this is /sdg/ as he wanted it

File: 400607-TFT_1240179324239.jpg (1.07 MB, 2560x2560)

1.07 MB JPG

fidget spinner autism bunny

>>101092507
you got the spinner, nice

>>101092507
he's cute

>>101092507
tits too small.

I wonder if Stability has enough talent left to start competing with Midjourney on API image gen.

File: tftWIP1.png (681 KB, 906x542)

681 KB PNG

wip, this is gonna need a lot of inpainting

>>101092768
is that a man?

>>101092768
do you go back to 1.5 to inpaint? i dont see any inpainting models for xl

>>101092831
you don't need an inpainting model you just inpaint with the normal model

>>101091912
>duckbros in /co/: Donald and Daisy, not with each other
>duckbros in /trash/: ducktales waifus
>duckbros in /g/: a horse

File: 00001-1851270299-1920x153(...).png (3.5 MB, 1920x1536)

3.5 MB PNG

howdy
everyone enjoying their summer with some refreshing worm icecream?

File: 00002-369705690-1920x1536(...).jpg (795 KB, 1920x1536)

795 KB JPG

>>101092985

File: 00003-3153211867-1920x153(...).jpg (785 KB, 1920x1536)

785 KB JPG

>>101093029

File: cudahelp.png (34 KB, 1316x743)

34 KB PNG

asking again if anyone can help me with this? impact nodes blew up and fails to load, saying I need to install a version of a package. when I do that, it says theres no CUDA_HOME, which has never stopped me before.

googling around, it seems like this variable gets set when you install the cuda toolkit, which I've never had. idk whats going on. and pls don't say "use venv lol" cuz thats not helpful

File: file.png (1.47 MB, 832x1216)

1.47 MB PNG

>>101093128
linux

>>101093170
looks like her face is pasted on from some completely different image

File: file.png (1.25 MB, 832x1216)

1.25 MB PNG

>>101093189
K.

>>101092980
>"You're going off on a long trip while I'm pregnant, Anon? Seriously?"

>101093205
>oppai loli
>>>/d/
>>>/b/

>>101093128
use venv

>101093251
she's 18

>>101093258
old

File: 00046-3215140083.png (1.81 MB, 1024x1536)

1.81 MB PNG

>>101093231
But, babe my paladin bros all are going!

>>101093258
Just a super prepubescent looking 18 year old, yes, of course.

>>101093294
shut up bitch

I'll give you that the oppai loli's outfit and staff are awesome, at least.
But oppai loli goes in /d/ and /b/.

File: 00129-[humu_v10]_[DPM++ 2(...).jpg (1.18 MB, 2048x3072)

1.18 MB JPG

[candle | 1girl]

File: 00004-3598192711-1920x153(...).jpg (909 KB, 1920x1536)

909 KB JPG

>>101093127

>>101093410
Anon, try this with Vofan as the main artist. I'm curious.

>>101093448
Did you?

File: SD3_1X-Quokka.jpg (214 KB, 970x1274)

214 KB JPG

SD3 Quokka isn't very good

>>101093464
>score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_anime, (Vofan:1.0), (solo, 1girl:1.2), (miku hatsune, hatsune miku:1.2),(oppai loli, small breasts):1.2 ,(aqua eyes, aqua hair, very long hair, twintails, hair ornament, hair between eyes, aqua nails:1.0), (:o:1.1), (leaning forward, close up, breast focus, covering breasts, holding breasts:1.0)

>>101093479
Can you try "[Vofan:3.0]"? I just need to see if it has an effect. If not, it may require a style lora.

>>101093477
how can it fuck up a quokka

File: 00005-1880224332-1920x153(...).jpg (818 KB, 1920x1536)

818 KB JPG

>>101093416

>>101093410
More oppai, less loli, more 2d, less 2.5d, (chibi:1.3) (very nice)

>>101093600
>More oppai, less loli
>Chibi
Late night timeslot Miku?

>>101093407
>"The stupid eventually stops burning you, trust me, Anon."

File: 00006-4275230254-1920x153(...).jpg (807 KB, 1920x1536)

807 KB JPG

>>101093583

>>101093533
Jump! Jump! Jump!

File: 00007-2647274730-1920x153(...).jpg (847 KB, 1920x1536)

847 KB JPG

>>101093739

File: pepe-help.png (27 KB, 128x128)

27 KB PNG

>>101091912
I updated my Automatic1111-Directml and now I can't run half of my models. I am getting the following error.
NotImplementedError: No operator found for `memory_efficient_attention_forward`
I am not using Xformers as I have an AMD gpu.

File: 000000_13700_.png (2.05 MB, 1584x894)

2.05 MB PNG

>>101093834
Nice.

>>101093128
appealing to anyone smart in the thread pls. is there anyone smart left?

File: 0.jpg (226 KB, 1024x1024)

226 KB JPG

>>101093911
thanks. nice cubism.

>>101093919
TY.
>>101093874
https://github.com/lshqqytiger/stable-diffusion-webui-amdgpu
From here=https://youtu.be/mKxt0kxD5C0

fix errors:
venv\Scripts\activate
pip install -r requirements.txt
pip install httpx==0.24.1
edit webui-user.bat file inside of automatic1111 folder and add command line arguments and save:
--use-directml --onnx
Inside of automatic1111 folder
find modules\sd_models.py file, edit it
comment out lines 632 - 635 by putting a # in front of the lines and save file
close out Automatic1111
Now you can run Automatic1111 by double-clicking on the webui-user.bat file from windows, or make a shortcut to it if you prefer.
Automatic1111 should now work the way it used to and should allow optimizing ONNX models.

File: 00011-866314824-1920x1536(...).jpg (872 KB, 1920x1536)

872 KB JPG

>>101093811

File: 00012-2627092015-1920x153(...).jpg (925 KB, 1920x1536)

925 KB JPG

>>101093950

File: 00013-2756705912-1920x153(...).jpg (727 KB, 1920x1536)

727 KB JPG

>>101093975

File: 00066-2610005806.png (1.92 MB, 1024x1536)

1.92 MB PNG

>>101094016
Nice one

File: 00014-2127724040-1920x144(...).png (3.21 MB, 1920x1440)

3.21 MB PNG

>>101093991

File: RA_2_00298_.jpg (1022 KB, 2064x2648)

1022 KB JPG

>>101094064
Thanks

File: 00015-125843625-1920x1440(...).png (3.57 MB, 1920x1440)

3.57 MB PNG

>>101094065

File: 00016-3919750432-1920x144(...).jpg (816 KB, 1920x1440)

816 KB JPG

>>101094160

>>101092234
>If everything is baked into the models then and the UI stuff is just preference or no?
Yeah that's pretty much it. The only thing (You) need to care about are what models the UI can run, but since all the popular UIs are forked from Automtic1111 they mostly share the same compatibility. Not everything is baked in though, some models require a specific VAE and such, they'll usually be mentioned where you get the model from.
>Different models just require different settings/tweaks/prompt rules?
Yes. The place you get the model from usually has the instructions, like "Must use sampler Euler A -a- SDMBDSMPlus++ Ultra with 69 steps" and such. Some UIs let you preview metadata for each model when you select them, you can use these to keep notes, or to make pre-set profiles that automatically switch to the model you like + its preferred settings.
>Is Stable Diffusion just like the brand name for AI stuff?
Not a brand, a type of model. SD is currently the most popular one that most UIs ONLY support. There are alternatives such as Stable Cascade but currently there's not that many good models for them. This will likely change in the future.

>>101093128
... so have you installed the cuda drivers? They aren't strictly required for ordinary gen but you need them if you want to do anything extra like training or GPU acceleration.
https://developer.nvidia.com/cuda-downloads

i miss schizo anon

>>101094299
I'm right here. AMA

https://desuarchive.org/g/search/text/i%20miss%20schizo%20anon/

>>101094307
how you feeling? and why do you do all of this?

>>101094326
>its real

Man Civitai is shit but it's become even shittier and now they're introducing payments for creators (ie. incoming censorship en masse). I liked uploading to it because it's got a nice search tool to find LoRAs you want for specific models and has (seemingly) unlimited hosting space for your models. It was also nice as you can upload your locally generated images and have them automatically link to the used LoRAs (though you get banned if you use a LoRA that has the "Don't let creators make NSFW with this model" enabled).

Is there a competitor that has the same feature-set or having a good search engine paired with being able to upload your own creation WITHOUT getting banned? Huggingface seems more scientific related (and has poor searching tools/tagging) while Tensor Art doesn't let you upload your own images.

File: CumfyUI_EAXL_00060_.png (1.87 MB, 1408x1088)

1.87 MB PNG

>>101094392
nice airship, though maybe a bit too literal

>>101094291
tried this and now I'm getting a totally different error message. seeing if I can progress from here. will report back
also, it installed cuda 12.5. I dunno if this is gonna cause version mismatching

File: 00018-1194440878-1920x144(...).jpg (966 KB, 1920x1440)

966 KB JPG

>>101094242

>>101091912
Ok but why doesn't the main post talk about "Stability Matrix". The thing installs everything for you and has an integrated interface so you can use it instead of the browser.

>>101094471
if you're on a1111 it 100% will have mismatch version issues because it uses 11.x

>>101094498
should I just uninstall/install a 11.x cuda?

>>101094521
I don't know what you haven't/have done already. But I would suggest to just go through the manual install again. People hate to hear it, but use a venv too.

>>101094538
I just want to save me comfy instal......

File: 00020-713566202-1920x1440(...).png (2.86 MB, 1920x1440)

2.86 MB PNG

>>101094477

>>101094471
Post your error

>>101094366
>Huggingface seems more scientific related
But they still allow you seemingly unlimited space for your models, a hosted inference api so people can try your models on the model's page directly, an option to apply all the loras you want on your models so people could try how any lora looks on any of your models, and it allows you to add images of your samples so people know how it looks like by just copying and pasting into it.
Civitai was garbage a year ago and that's why I've dedicated my life to uploading my models to huggingface and mirroring all the civitai ones that I like. And then on huggingface I get 50k downloads/month, not too shabby.

File: 00023-1043835413-1920x144(...).png (3.17 MB, 1920x1440)

3.17 MB PNG

>>101094522
that's a nice style
>>101094555

File: error2.png (109 KB, 1897x680)

109 KB PNG

>>101094583
its not helpful. I tried the install it recommended but nothing changed

File: 00024-1783403592332656955(...).png (3.37 MB, 1920x1440)

3.37 MB PNG

>>101094608

>>101094613
It says "you're missing a tool, get it from https://visualstudio.microsoft.com/visual-cpp-build-tools".

People recommended Linux earlier because most distros already have the tools like this already installed by the distro. You'll probably find that after installing the C++ compiler then you'll need to install even more prerequisite software.

>>101094666
I already tried that install but it gives the same error

>>101094326
grim

>>101094671
So the question would be, did those tools install successfully? If you go into Add/Remove programs, does it appear in the list of installed software?

The problem could also be that even if the compiler is installed correctly, Python can't see it, which might happen if you've installed Python to an unusual location or have some weird environmental variables set up. Based on your screenshot, it looks like you're just using the defaults. Did anything strange happen when installing Python like dismissable error messages?

File: cpp.png (42 KB, 612x857)

42 KB PNG

>>101094696
afaik it installed successfully. nothing indicated otherwise. when I look at the app list, I don't see it by that name but sometimes things install w/o being on the app list
> Did anything strange happen when installing Python like dismissable error messages?
I've not ever had any problems with python and have used comfy and all sorts of custom nodes for a year or so with no issue

File: 00026-618593905187605763-(...).jpg (658 KB, 1920x1440)

658 KB JPG

>>101094658

File: 00027-571991789299344836-(...).jpg (702 KB, 1920x1440)

702 KB JPG

>>101094746

>>101094732
That's just the redistributables, not the compiler. So I would recommend trying to reinstall the compiler again, reboot, and give it another shot. If it still errors out, you might be in a bind as to discovering what the actual issue is. And in which case, you need to decide if it's worth doing that deep-dive investigation, or starting from scratch (and using a venv this time so your system doesn't mess up the installation in the future).

Whatever you decide, good luck. Remain persistent, the pay-off will be worth it.

>>101094802
I think I'm just gonna kill myself instead

>>101094822
clean style

File: 00029-497803081933095957-(...).png (3.47 MB, 1920x1440)

3.47 MB PNG

>but anon he looks just like you! he's your son!
>no

>>101094787

>>101094595
I see, makes sense to do that. My other plan was to just make a MEGA to upload my loras to and link it here. And for image posting, I still post them here but specifically I wanted a permanent place to host them, I'll probably start using a booru or something.

>>101094366
>(though you get banned if you use a LoRA that has the "Don't let creators make NSFW with this model" enabled

wat? that's not a thing

>>101094940
yeah not exactly you're just not allowed to post it, but hyperbole is the essence of engagement (and who's to say they won't start banning you for making "infringements" like that? the fact it exists already shows the people running civitai are legitimately insane)

File: 00032-961187105111195300-(...).png (3.66 MB, 1920x1440)

3.66 MB PNG

mfw i've been using a pony lora on regular sdxl models
it felt a bit off night
>>101094870

is the AI porn scene still a thing or is it oversaturated already? I wanna make money

>>101094973
Get a real job.

>>101094999
I'm trying to. The tech job market is ass in the US especially for junior devs. I'm about to give up and start applying to burger flipping jobs but I'd still like to make money doing porn on the side (I'm too ugly to do IRL porn)

>>101095020
lie lie lie lie and lie
also lie
you're welcome

File: 00037-34415476694701-1920(...).png (3.6 MB, 1920x1440)

3.6 MB PNG

>>101094972

>>101095046
can I go as far as lying about my years of experience? I thought that shit turned up in background checks

File: 00039-24277764849380-1920(...).png (3.51 MB, 1920x1440)

3.51 MB PNG

>>101095050

>>101095062
Depends on how far you want to go.
Easy mode? Just make something up and use your friend as a reference for whatever contractor/freelancer job you put down.

Alternatives would be making an LLC (cheap & easy). You can also use the route of putting down prior history of recently defuct businesses. They can't check on the history if it doesn't exist anymore.

As far as the background checks go none of that is extensive unless it's a contractor job for the government, a government position itself (state/fed) or fortune 100 companies.

You'll be fine. Oh and don't use AI slop to make your cover letter/resume.

File: CumfyUI_EAXL_00059_.png (2 MB, 1408x1088)

2 MB PNG

>>101094973
are you better at it than everyone else?

>>101095088
Not a requirement. As long as you can do the job that's all that matters the rest is doing whatever you can to get your foot in the door.

File: CumfyUI_EAXL_00058_.png (1.7 MB, 1408x1088)

1.7 MB PNG

>>101095046
hes right. the market has been total shit since the start of 2023. AI is making it worse because people think they can just plug in more AI instead of hire more people. but also the market is still fucky from all the personnel shuffling that happened during/after covid

>>101095109
maybe it wasn't before but its a requirement now. only way you can build any following is to out-compete the market or find one of the few unexplored niches no one has capitalized on yet. there's not much room in either direction to find success. only way to reliably make money in genAI these days is to write code

>>101095137
>maybe it wasn't before but its a requirement now. only way you can build any following is to out-compete the market or find one of the few unexplored niches no one has capitalized on yet. there's not much room in either direction to find success. only way to reliably make money in genAI these days is to write code
Yeah I was speaking in IT in general, mainly any junior software dev position. If you can demonstrate your ability (have a repo in your resume) that's clean code and tell some white lies you'll be alright.
And I'm not discounting the current playing field right now, it is competitive more so than it has been before.

>>101092514
is this base sdxl?

>>101095216
>base
who is genning on any base model except for the new chinaman ones?

File: pepe-cry.gif (14 KB, 220x218)

14 KB GIF

>>101093949
I did all that but I am still getting the same error. Also, it says that --onnx is an unrecognized argument.

>>101095020
Get in shape. Get a neat haircut. Get some nice conservative work clothes. Look good as good as you can. Keep applying for jobs.

File: CumfyUI_EAXL_00056_.png (1.92 MB, 1408x1088)

1.92 MB PNG

>>101095161
pretty hard to get budget to hire a junior these days when you can prompt and supervise an AI with about the same level of effort and wasted time. also, its pretty rare for any juniors to have boastable code. they might have an html resume or a maybe cloned repo but thats it. the hireable juniors are 1 in 10,000 and there's not that many jobs

>>101095234
how old are you?

>>101095238
>how old are you?
97

>>101095238
>its pretty rare for any juniors to have boastable code. they might have an html resume or a maybe cloned repo but thats it
Doesn't need to be boastable just something that demonstrates your proficiency. And this is a good example of the slop resumes people try to pass off. If you do better than this you are already ahead of the game.

File: 00043-65132621908380-1920(...).png (3.22 MB, 1920x1440)

3.22 MB PNG

>>101095073

>>101092569
for a second I thought armhairchads had won but it was just the cape

>>101095227
who knows, but the filename there is SDXL_

>>101093426
I can read AI text. Anon typed "sex with duck" here.

File: 00045-65132621908380-1920(...).png (3.21 MB, 1920x1440)

3.21 MB PNG

>>101095271

File: 00047-16840916022998-1920(...).jpg (658 KB, 1920x1440)

658 KB JPG

>>101095285

I shitted and I farted

File: CumfyUI_EAXL_00055_.png (1.94 MB, 1408x1088)

1.94 MB PNG

>>101095259
I dont think you appreciate how tight the market is for juniors. the swe market is tight all around and standing out as a junior, especially just getting your resume through the first round, is real hard. even top cs programs are having a harder time placing graduates into roles.

best bet is to continue on with a masters in something. but thats two more years of education, which means two more years of paying for school, and two more years of the markets shrinking even more

>>101091979
anon, that means that we'll never get SD3-8b, this company only makes closed source model, it's over, SAI is dead

File: 00049-12316851622508-1920(...).png (3.27 MB, 1920x1440)

3.27 MB PNG

>>101095321

>>101095341
>I dont think you appreciate how tight the market is for juniors.
yep, as a data scientist junior I got only 1 interview in 6 month, and the guy who interviewed me literally said that I should lie on my resume to get a job, like saying that I have some experience in a company or some shit, that's fucking brutal

>>101095375
This is why I said initially to lie lie lie, and lie.

>lie lie lie, and lie.
The /sdg/ motto

>>101095384
I just can't do that, my morals tell me that lying to people is bad, I don't want to be that guy just to get a fucking job, I know everyone lie on their resume but I wouldn't be able to look myself in the mirror if I do something like that, guess that I'm too much of a nice person for this fucked up world

>>101095397
It's a dog eat dog world anon, sorry to break it to you. But good on you to sticking to your principles.

>>101095406
>It's a dog eat dog world anon
I know, that's why I'm not making kids, this world sucks too much to bring more people into it, can't see myself telling my kids they should lie and be a fucking snake just to survive

Catboxing just in case the jannies get pissy
https://files.catbox.moe/l7au83.png

File: CumfyUI_EAXL_00054_.png (1.95 MB, 1408x1088)

1.95 MB PNG

>>101095397
its not just morals, its a waste of time. you'll lie on your resume to squeak through the ai filters and the hr sanity check but as soon as you're in a technical when anyone, they'll realize you have no idea about anything. maybe you can get lucky and they'll think "whatever, this guy is cool and we can work with him" but its more likely that they have hundreds of more candidates to choose from

File: 00051-34585263948838-1920(...).png (3.63 MB, 1920x1440)

3.63 MB PNG

fug i'm high

>>101095369

>>101095471
>its not just morals, its a waste of time. you'll lie on your resume to squeak through the ai filters and the hr sanity check but as soon as you're in a technical when anyone, they'll realize you have no idea about anything.
I mean, you can say you worked on a company and you did some stuff you're familiar with, but yeah, the fuck is wrong with this system, juniors aren't good at their job, well no shit that's why they're juniors, what the fuck do they expect?

>>101095471
It's not a waste of time. Of fucking course you can't fake it until you make it in these types of environments. Though to think just because you're fresh out of college and have CS degree you'll be scooped up you're sorely mistaken (obviously now times have changed people are wanting "data scientists" among other shit, or a fleet of full stack devs wearing all the hats).

>>101095450
nooo you can't just heccin generate some arbitrary pixels in certain configurations next to each other ohh my virgin eyes help help i need the board police

>>101095517
They really hate even the barest suggestion of nipples. I bet jannies got bottle fed

>>101095528
>I bet jannies got bottle fed
interesting theory there anon

>>101095510
>Though to think just because you're fresh out of college and have CS degree you'll be scooped up you're sorely mistaken
this, the only way to get a job as a junior is to know someone from the company you'd want to be hired on or to lie on your resume, it's just how it is, companies are allergic to people with no experience even though to have experience, you need to be hired, this is fine :)

>>101095510
>obviously now times have changed people are wanting "data scientists" among other shit, or a fleet of full stack devs wearing all the hats
true, they don't want "data scientist" anymore, they want someone who can do a bit of everything (data engineer -> data scientist -> data architect)

File: CumfyUI_EAXL_00052_.png (1.95 MB, 1408x1088)

1.95 MB PNG

>>101095493
>the fuck is wrong with this system
money. juniors are an investment in the longevity of a company but longevity doesn't report on a quarterly. hiring managers get budget for performance, not investment, so they spend that money on devs that need minimum ramp. plus, like I said, there's this magical tale being told that all these productivity gains can be had from just throwing my AI into the machine instead. so instead of hiring on 5 new devs, they'll outsource some AI company and license a model instead.

>>101095510
>It's not a waste of time
>you can't fake it until you make it
pick one
> just because you're fresh out of college and have CS degree you'll be scooped up
thats the promise a lot of CS students were sold and it was true, if not for a fleeting moment

File: XL_gen_tmp_09.jpg (499 KB, 1536x1200)

499 KB JPG

Gn fellas

>>101095547
>thats the promise a lot of CS students were sold and it was true
not true anymore, ChatGPT killed that dream, now a company can hire one guy to do the work of previously 20 engineers because he can simply use GPT4 to get shit done

>>101095553
night

>>101095547
>>It's not a waste of time
>>you can't fake it until you make it
>pick one
Nothing I've said is contradictory. Have you not read what I've been saying? You have to be able to demonstrate your proficiency. As long as YOU KNOW you can do the job you should do whatever it takes to get your foot in the door.

what are you guys even arguing about, just post 1girls

>>101091979
anon... lay off that söylent, it's clearly too much you.

>>101095564
It's impossible for him to get along with anyone

File: 00053-93987372444511-1920(...).png (3.28 MB, 1920x1440)

3.28 MB PNG

>>101095553
gn
>>101095488

File: ComfyUI_temp_vsyre_00011_.jpg (988 KB, 1504x2192)

988 KB JPG

>>101095564
You're right I'll get back to genning but I don't know about 1girls.

File: CumfyUI_EAXL_00051_.png (1.83 MB, 1408x1088)

1.83 MB PNG

>>101095559
thats not true (yet) but thats the [profitable] dream thats being sold. thats why no one is picking up juniors en masse rn, combined with the gutt of more talented devs on the market

>>101095563
>Have you not read what I've been saying?
I've been trying but you don't make much sense. I think you want to believe something to be true more than you've reasoned out the truthiness of it

>>101095611
>thats not true
>no one is picking up juniors en masse rn, combined with the gutt of more talented devs on the market
So it is true kek

>>101092412
thanks for the PSA

File: 00059-117982499327978-192(...).png (3 MB, 1920x1440)

3 MB PNG

>>101095589
1girls make it fun tho
>>101095576

File: analogist glitch first attempt.gif (3.95 MB, 512x512)

3.95 MB GIF

i tried to make a gif a new way but it didn't work

File: 00060-48048078514089-1920(...).jpg (669 KB, 1920x1440)

669 KB JPG

>>101095701

File: 00061-60025462875592-1920(...).png (3.7 MB, 1920x1440)

3.7 MB PNG

>>101095770

File: 00063-32489149937795-1920(...).jpg (735 KB, 1920x1440)

735 KB JPG

>>101095820

File: 00064-10759901849276-1920(...).png (3.27 MB, 1920x1440)

3.27 MB PNG

>>101095848

i love sdg

>>101094550
why? you're not going to lose anything. just don't delete your inputs/outputs/models folder, copy those to the new install

>>101094901
This is an example of something that you can do if you don't care about people using your stuff on the website itself:
>https://huggingface.co/Bruhn/7choma/tree/main
Bruhn just created an account, named the archive as 7choma, and now everyone can download any of the models on that repo.
And this is an example of what you can do if you want to let people try out your Loras without having to download them:
>https://huggingface.co/alperiox/cctv-lora
People type something and use the Lora on the page directly and get images in 10 seconds.

>>101094973
Generations still suffer from bad hands and wrong number of legs and arms, so you could help with that in general.
The models without those problems generate the same pose every time.

File: 00069-180082512401217-192(...).png (3.64 MB, 1920x1440)

3.64 MB PNG

>>101095897

>>101095020
Ask for help in /x/, some anons have used magic to manifest wealth without having to work, like me.

>>101096187
demon eyes are cool, she needs some tiny horns or somethin tho, listen you really gotta rub it in around here, you know?

File: 00071-180082512401217-192(...).png (2.54 MB, 1920x1440)

2.54 MB PNG

>>101096175

File: ia_00011.jpg (819 KB, 1664x3112)

819 KB JPG

OoO

File: 00072-109021104382853-192(...).png (3.08 MB, 1920x1440)

3.08 MB PNG

>>101096219

File: 00074-25924819887356-1920(...).png (3.13 MB, 1920x1440)

3.13 MB PNG

>>101096263

File: ia_00013.jpg (259 KB, 1256x2712)

259 KB JPG

devolving.. no, un-evolving. irrevolution. something like that.

File: 00077-10398536475557-1920(...).png (3.02 MB, 1920x1440)

3.02 MB PNG

>>101096283

File: 00024-1583202376.png (974 KB, 1064x1192)

974 KB PNG

Okay, I gived her horns

File: file.png (219 KB, 612x542)

219 KB PNG

>>101096313

File: ia_00014.jpg (300 KB, 1256x2712)

300 KB JPG

>>101096313
what're your tokens for the black eyes? this is very important to me for reasons i can't entirely comprehend

File: 00080-51687692030384-1920(...).png (2.78 MB, 1920x1440)

2.78 MB PNG

and with that, gn all
>>101096308

>>101096373
night

File: ia_00015.jpg (383 KB, 1256x2712)

383 KB JPG

>>101096359
"black eyes, black sclera"
this is NOT the way!

File: ia_00016.jpg (399 KB, 1256x2712)

399 KB JPG

i'm twisting in the wind here >>101096313

>>101096390
>black eyes, black sclera
It didn't work? That was all I used. I found the model makes a difference though and I am using an artist style lora, so there might be some images in the dataset that is helping it

>>101096428
i can do better than that

File: ia_00016.jpg (383 KB, 1256x2712)

383 KB JPG

MULLIGAN

>>101096438
damn. yeah, must be a model thing

>>101096575
GREAT GEN BRO! THIS IS EPIC! YOU BELONG IN THE DISCORD!

>>101096587
i'm deadly serious

>>101096587
chill anon, it was in reference to >>101096560

File: file.png (18 KB, 725x190)

18 KB PNG

>>101096602
this is a very confusing post. you refer me to me. it's distressing, to say the least.

what would people say?

>>101096607
Oh right on, well spazz on my friend

File: ia_00017.jpg (352 KB, 1256x2712)

352 KB JPG

>>101096621
i said what i said; and i meant it.

File: ComfyUI_temp_fqpzj_00005_.png (3.15 MB, 1616x1256)

3.15 MB PNG

>>101096637
I'd rather shit razor blades

>>101096662

>>101096662

>>101096662

File: ia_00018.jpg (314 KB, 1256x2712)

314 KB JPG

>>101096651
that's gonna be a major yikes from me, dawg. hard to believe there are still people as retrograde as the likes of (((you))). All hands on deck!

>>101096180
just checked and it's the most pants on head retarded thread I've ever seen in my entire life
thanks for the good laugh, read every single post in the current thread it was hilarious

>>101096676
yeah okay man

HALP ME OBAMA MAN! I'M GOING INSANE

File: ia_00019.jpg (477 KB, 1256x2712)

477 KB JPG

cycling through the tv approved opinions
non of them seem all that compelling but
there's this thing i can't ignore that makes me pick
one or the other

File: ia_00020.jpg (362 KB, 1256x2712)

362 KB JPG

i am red team
(but have you considered..?)
now i am blue team
(but have you considered ?)
now i am no team
i am i team
my team is team
team.
teamy team team team
team.

fuckin teams. shit's fucked man. RIPq

aww shucks, i'm still here
well that's just great

File: download - 2024-02-11T143(...).jpg (70 KB, 512x768)

70 KB JPG

Guys I've been using Fooocus. Is that any good? also I really want to get into animation the prompts. I look at the resources above but they are just Chinese to me. Anyone got any tips. Thanks

Is there a working solution for sdxl controlnet? I mean working like it's as good as in 1.5. I have seen the adapter-x stuff but is there any experience with that?

Theory question:
I understand that given dataset of different classes, we can use the forward process in diffusion to try and learn a backward diffusion.

But how is it implemented that novel classes are combined?
If I say "cat sitting on a Star-wars X-wing in front of the a status of Gauss" and give it a noisy image to work on, how does it apply both targets (cat, x-wing, Gauss).
The learning data only had cats and x-wings.
I understand they were also split in learning.
* In the generation step, does it alternate between the diffusion motions through the space that it had learned?
* Is the merging of classes already incoorporated in the learning step (that sounds like too many combinations)
* Something else?

Secondly, once you've already got a more or less okay image of a cat and Gauss behind it, how is it achieved that further refinement (e.g. to make the cat more crisp) does not mess up the Gauss statue? Is this automatic just by the cat and Gauss being in vastly different dimensions in laten space, so that the Gauss data isn't affected by any catty motion?

>>101098288
Ttext to image processes strongly rely on clip to generate image, formers diffusion models used clip directly with backwards propagation to modify images towards being similar to text for the clip model
stable diffusion models learn to use embeddings but all the text processing is done by clip.

>>101098975
Thanks, got a reference for clip?

This sound more like my "combinatorical explosion of data" option, but not sure yet.

>>101098196
it might just be a matter of waiting for good models. hetaneko replicate is really good at what it does

>>101099047
Sometimes it's better to look at the technical ancestors of something to understand the ways it works.
I was talking about diffusion models prior to stable diffusion.
Today it strongly relies on embedding and latent vectors instead of directly images and text

>>101091912
I don't know if this is the best place to ask, but is there any NCNN model for processing RAW camera images? aka digital negatives? stuff like debayering, denoising etc.

thanks.

>>101099061

Thanks anon I'll try it out.

>>101099816
You're looking at methods that aren't diffusion and more photo restoration or processing. But yes. The latest one I am aware of with open source code and model is https://github.com/megvii-research/NAFNet. The state of the art right now is https://github.com/Ascend-Research/CascadedGaze but they only released their paper last month.

>>101091979
How much did they pay for SAI? Their researchers are gone, right? So they've got the weights but how valuable are those really?

>>101092412
If you are so careless that you allow yourself to get doxxed you deserve it honestly

>>101101616
I'll check those out, thank you so much.

>>101096916
>Fooocus. Is that any good?
It's ok.

> I really want to get into animation the prompts
That's mostly done with comfyui and other more complex tools.

>>101092412
play a game of let act like a retarded douche and win retarded prizes. No sympathy.

>>101099816
https://www.phoronix.com/news/Darktable-4.8-Released

>Prem Akkaraju