File: FDG_News_000014_.jpg (800 KB, 1344x960)

800 KB JPG

>mfw Resource news

08/04/2024

>SimpleTuner v0.9.8 - flux 24 gig training has entered the chat
https://github.com/bghira/SimpleTuner/discussions/639

>FastSD CPU v1.0.0-beta.35: Adds Aura SR v2 support
https://github.com/rupeshs/fastsdcpu/releases/tag/v1.0.0-beta.35

>SimpleTuner now supports Flux.1 training (LoRA, full)
https://github.com/bghira/SimpleTuner#flux1

>Civitai Model Manager: CLI tool for managing AI models from CivitAI
https://github.com/regiellis/civitai_model_manager

>Skimmed_CFG: Powerful ComfyUI anti-burn allowing much higher CFG
https://github.com/Extraltodeus/Skimmed_CFG

>ComfyUI-AdvancedLivePortrait
https://github.com/PowerHouseMan/ComfyUI-AdvancedLivePortrait

08/03/2024

>ComfyUI/Forge Implementation of Smoothed Energy Guidance
https://github.com/pamparamm/sd-perturbed-attention

>TryOnDiffusion: A Tale of Two UNets
https://github.com/fashn-AI/tryondiffusion

>Nvidia reportedly delays its next AI chip due to a design flaw
https://www.theverge.com/2024/8/3/24212518

>ComfyUI Frontend Modernization: Transitioning to a New Era on August 15, 2024
https://github.com/comfyanonymous/ComfyUI/issues/4169

>CEO of Invoke says Flux fine tunes are not going to happen
https://www.reddit.com/r/StableDiffusion/comments/1eiuxps

>ComfyUI-FLUX-fal-API
https://github.com/gokayfem/ComfyUI-FLUX-fal-API

08/02/2024

>Optimizing Diffusion Models for Joint Trajectory Prediction and Controllable Generation
https://yixiaowang7.github.io/OptTrajDiff_Page

>UniTalker: Scaling up Audio-Driven 3D Facial Animation through A Unified Model
https://github.com/X-niper/UniTalker

>Smoothed Energy Guidance for SDXL
https://github.com/SusungHong/SEG-SDXL

>Mitigating Multilingual Hallucination in Large VLMs
https://github.com/ssmisya/MHR

>GalleryGPT: Analyzing Paintings with Large Multimodal Models
https://github.com/steven640pixel/GalleryGPT

>The Manga Whisperer: Automatically Generating Transcriptions for Comics
https://github.com/ragavsachdeva/magi

>mfw Research news

08/04/2024

>Comprehensive Survey of Complex-Valued Neural Networks: Insights into Backpropagation and Activation Functions
https://arxiv.org/abs/2407.19258

>Detached and Interactive Multimodal Learning
https://arxiv.org/abs/2407.19514

>Take A Step Back: Rethinking the Two Stages in Visual Reasoning
https://arxiv.org/abs/2407.19666

>ActivityCLIP: Enhancing Group Activity Recognition by Mining Complementary Information from Text to Supplement Image Modality
https://arxiv.org/abs/2407.19820

>More precise edge detections
https://arxiv.org/abs/2407.19992

>From Flat to Spatial: Comparison of 4 methods constructing 3D, 2 and 1/2D Models from 2D Plans with neural networks
https://arxiv.org/abs/2407.19970

>FedDEO: Description-Enhanced One-Shot Federated Learning with Diffusion Models
https://arxiv.org/abs/2407.19953

>FiCo-ITR: bridging fine-grained and coarse-grained image-text retrieval for comparative performance analysis
https://arxiv.org/abs/2407.20114

>OmniBal: Towards Fast Instruct-tuning for Vision-Language Models via Omniverse Computation Balance
https://arxiv.org/abs/2407.20761

>DMESA: Densely Matching Everything by Segmenting Anything
https://arxiv.org/abs/2408.00279

>Resilience and Security of Deep Neural Networks Against Intentional and Unintentional Perturbations: Survey and Research Challenges
https://arxiv.org/abs/2408.00193

08/03/2024

>Image Super-Resolution with Taylor Expansion Approximation and Large Field Reception
https://arxiv.org/abs/2408.00470

>Localized Gaussian Splatting Editing with Contextual Awareness
https://arxiv.org/abs/2408.00083

>Hierarchical Conditioning of Diffusion Models Using Tree-of-Life for Studying Species Evolution
https://arxiv.org/abs/2408.00160

>SynthVLM: High-Efficiency and High-Quality Synthetic Data for Vision Language Models
https://arxiv.org/abs/2407.20756

>Real Face Video Animation Platform
https://arxiv.org/abs/2407.18955

File: schFLUX_00014-Australian,(...).png (948 KB, 640x1536)

948 KB PNG

Wallpaper Quokka

>>101728927
Nice
What model is this?

File: 1697564600127829.png (1.54 MB, 1336x1376)

1.54 MB PNG

>>101728822
how do I achieve this

>>101728960
LoRA probably, i would suggest Broly Culo

File: 00088.png (1.12 MB, 968x1088)

1.12 MB PNG

>>101728960
Pony

File: ComfyUI_0024.jpg (2.37 MB, 1550x2000)

2.37 MB JPG

>>101728960
it was made with a pony model

>>101728943
juggernautxl 8

>>101728977
Is this an earmuff-anon production? Looks great either way
>>101729015
Tristram... home...

>>101729017
>>101729010
yeah someone shared these, but any way to check if the image really came out of one of these? I thought the .png files kept the AI metadata or something but I see nothing

https://civitai.com/models/413497?modelVersionId=468372
https://civitai.com/models/611828/nakayama-tooru-style-rockman-zero-xl

>>101729003
kek

File: 1720786687406274.png (948 KB, 768x576)

948 KB PNG

>>101729047
>Tristram... home...
It's actually Hell, but same thing

>>101729017
Thanks anon, any reason you're not using v9?

File: 1712151124533833.png (922 KB, 768x576)

922 KB PNG

>>101729072

>>101729055
It should. This one has the metadata and it's similar enough you can probably play around with tweaking the prompt to get what you want
https://civitai.com/images/10757434

File: delux_tru_00036_.png (1.2 MB, 1344x768)

1.2 MB PNG

>>101729060
she doesn't look the same without her hair going wild

File: 102103-tmp.png (3.12 MB, 1536x1728)

3.12 MB PNG

>>101729117

File: delux_tru_00037_.png (1.2 MB, 1344x768)

1.2 MB PNG

>>101729180
this is to go even further beyond: ssj3 delphi

File: ComfyUI_0029.jpg (2.18 MB, 1550x2000)

2.18 MB JPG

no reason

File: delux_ggf_00037_.png (1.22 MB, 896x1152)

1.22 MB PNG

>>101729237
teenage angst has paid off well

File: delux_bu_00005_.png (1.18 MB, 1344x768)

1.18 MB PNG

I yearn for a model that knows how bongs work

File: 00092.png (1.3 MB, 968x1088)

1.3 MB PNG

>>101729284
I had trouble trying to make emo girls though

File: 00044.png (1.24 MB, 1024x1024)

1.24 MB PNG

>>101729333
Nice, I tried to do something similar last night but gave up. Either the face was busted or there was some kind of body horror, and I couldn't get full body

File: ComfyUI_0003.jpg (2.44 MB, 1550x2000)

2.44 MB JPG

>>101729402
with flux? try changing the aspect ratio, and use full body shot in the prompt, also add the footwear in the prompt, that should work

File: ComfyUI_0007.jpg (1.83 MB, 1400x1800)

1.83 MB JPG

silly porn models

Flux looks a bit overcooked all the time. Even at 1.5 cfg.

File: 1714312699376556.png (159 KB, 267x284)

159 KB PNG

>>101729492
kek

>>101729590
You're supposed to do 1 CFG

>>101729492
Women that cute don't wear that kind of shit and whenever they do is because they used to be men.

>>101729600
it's already pretty washed out at 1.5, but I'll give it a shot. I'm 80% sure you're joking but it's worth a try anyway.

File: delux_ggf_00038_.png (1.05 MB, 896x1152)

1.05 MB PNG

>>101729590
I've seen it overcook randomly but I don't think its consistent or pervasive

File: 1709257753071987.png (1.44 MB, 1024x1024)

1.44 MB PNG

>>101729617
I've been doing it all at 1 after people told me to, picrel, not exactly washed out

File: delux_bu_00010_.png (1.09 MB, 1344x768)

1.09 MB PNG

flux is so shit. it doesn't even know how a magic game is supposed to look

File: 00100.png (1.41 MB, 968x1088)

1.41 MB PNG

>>101729616
>Women that cute don't wear that kind of shit
Now they can because AI

>>101729709
I just don't think it's been trained to show hot girls playing one

File: FLUX_59856118968592_00001_.png (626 KB, 768x640)

626 KB PNG

>>101729629
That image looks very cooked to me.

Maybe it's because my monitor has larger than average pixels or maybe it's because I've been a base model prompter since way back but I'm really sensitive to "the AI look"

>>101729645
That does look vibrant and high contrast. I'll try taking it down a notch. But this first gen at 1.0 still looks a bit washed given how much I tried to emphasize blue in the prompt.

>>101729724
overcooked

File: 1703936990158848.png (1.05 MB, 1024x1024)

1.05 MB PNG

based on current news

File: delux_ggf_00058_.png (1.22 MB, 896x1152)

1.22 MB PNG

>>101729616
>>101729724
cute girls always wear band merch. its what they substitute for not having a personality

>>101729770
>That image looks very cooked to me.
I think you might be confusing 'cooked' with 'saturation'. cooked isn't just a saturation or contrast quality, it has a very specific affect on hues and lightness balance

File: FD_00613_.png (2.69 MB, 1024x1536)

2.69 MB PNG

I have a couple for you polfags.

File: FD_00610_.png (2.29 MB, 1024x1536)

2.29 MB PNG

>>101729834

File: FD_00615_.png (2.39 MB, 1024x1536)

2.39 MB PNG

>>101729843

File: FLUX_354731784297966_00001_.png (559 KB, 768x640)

559 KB PNG

>>101729770
Some of these 1.0cfg gens show obvious signs of denoising failure, where things don't finally become coherent shapes/objects but remain indistinct fields of visual noise. This one has clear artefacts of cfg too low.

Maybe a few failed gens in this way is desirable, not in itself, but as a side-effect of striking the optimal balance. But I suspect the ideal cfg might be a little higher.

>>101729806
I've been doing this as long as you Debo, don't try to lecture me on what cooked does or doesn't mean. Everyone knows overcooking is a generic term for artefacting from too-high cfg, and it manifests in many different ways. One such way is the classic "slimy" look often associated with finetunes, which your image subtly exhibits

File: delux_ggf_00044_.png (1.16 MB, 896x1152)

1.16 MB PNG

>>101729857
I'm not trying to lecture you or suggest I have a better opinion or anything. I'm just suggesting you may have a bias influencing your perspective. its ok to just disagree though, not trying to make it personal. I'm just saying I haven't felt flux to be more prone to over-cooking and haven't seen it in other anons' gens either

File: FD_00621_.png (2.55 MB, 1024x1536)

2.55 MB PNG

Remember 2B is all you need.

File: F_00102_.png (1.42 MB, 968x1088)

1.42 MB PNG

>>101729806
She looks like she would have been an MSI fan

THAT EASY DIFFUSION DL GIVE ME A VIRUS NOTICE!

>>101729834
style prompt? it's an older style of advertisements

File: delux_ggf_00057_.png (1.29 MB, 896x1152)

1.29 MB PNG

>>101729935
msi girls were as unhinged as msi

>>101729853
>Hiter
One job

>>101730022
A worn vhs cover for a movie.
>>101730076
Don't worry, Nazis can't actually read. It's pure pattern recognition.

File: delux_bu_00034_.png (1.22 MB, 1344x768)

1.22 MB PNG

>new pizza who dis

File: FLUX_317995639452902_00001_.png (628 KB, 768x640)

628 KB PNG

>>101729903
You're a tedious bore who talks down to anons whenever you don't understand what they're saying. I am not interested in your suggestions that I might be "biased" or might not understand basic concepts. Maybe you could try using your eyes to see the very obvious things I see and point out. E.g. that there is a large chasm between this image and yours in terms of cookedness which is easily discerned by the human eye.

I am not saying this to be insulting; initially I pointed out the cooking only because it was relevant to the question being discussed; but since you have more or less accused me of not knowing what I'm talking about for suggesting this, I think it's fair for me to point this out more forcefully. I am not offering this gen as a perfect or compelling image, but it has that naturalness of texture, light, etc. which we associate with real images; if it weren't for some evident errors, it could almost fool us. Yours on the other hand can be clocked as AI at a glance, it has that AI look. Can you seriously not see it?

Well, if anyone could fail to see it, it would be you. You have ever been, and still remain, the worst /sdg/ poster.

(Here, for contrast, is an anon who helpfully pointed out how I was wrong: >>101729645)

>>101728822
Whom?

Debo containment thread

File: delux_ggf_00045_.png (1.16 MB, 896x1152)

1.16 MB PNG

>>101730183
>you have more or less accused me of not knowing what I'm talking about
I said the opposite. I respect your input
>Can you seriously not see it?
I said I could, but that I don't think its outsized compared to other models

sorry if I said the wrong things to you anon, or said things in the wrong way. I just wanted to converse because you raised an interesting perspective

>>101730261
Alright fine. I don't think this is sincere but it doesn't matter whether it is, I appreciate the decorum of it anyway. It is an adequate performance.

I apologize for my outburst of temper. And admittedly there are pedophiles and schizophrenics who are much worse posters, so even ignoring your virtues I still couldn't call you the worst.

Let's forget all this unpleasantness.

you know we're back when anons are actually being civil to each other in these threads

File: F_00105_.jpg (273 KB, 953x1047)

273 KB JPG

>>101730067
I first learned about them from a girl that was

File: 1717637875176204.png (861 KB, 768x576)

861 KB PNG

>>101730336
I took like a 6 month hiatus and came back for Flux

File: Flux_00013_.png (1.2 MB, 904x1104)

1.2 MB PNG

>>101730336
Too busy genning to shitpost. Sorry.

File: file.png (1.87 MB, 1024x1024)

1.87 MB PNG

>>101730356
so did i, actually. seems using (brackets) no longer works for emphasis on parts of a prompt in flux. does anyone know, was that an SD-only thing?

File: Flux_00172_.png (1.46 MB, 1128x904)

1.46 MB PNG

This is a cool volcano but where's my bunger king? Also has weird shit around the border that I haven't seen before.

File: FLUX__00034_.png (1.02 MB, 1024x1024)

1.02 MB PNG

>>101730356
I was into LLM's for a bit, nemo had just come out which probably has some decent finetunes by now
I'm loving my 3060 purchase. it was the most recent and expensive gpu I've ever bought by far, but pretty worth it

>>101730386
I'm not sure that emphasis works at all for Flux desu

>>101730404
You are doing flux on a 3060? How long does it take?

>>101730479
about 140 seconds for 1MP, considerably more beyond that

>>101730502
You can drop it as low as 512 and still get coherent results from flux btw

File: file.png (2.08 MB, 1024x1024)

2.08 MB PNG

>>101730438
i'm fooling around with random punctuation and this bacon and eggs prompt. oddly ***eggs*** and bacon makes the bacon into a ring shape lol, but that's about the only interesting thing that i've seen yet

File: FLUX__00037_.png (88 KB, 256x256)

88 KB PNG

>>101730522
I know, I plan on making some icons

>>101730502
Similar here with a 3060ti. The difference between dev and Schnell was barely 10~20 seconds so I just said fuck it and went with the added quality from dev.

>>101730502
>>101730579
460 seconds on my 2060 12GB. Funny how AI today is like video games in the late 90s/early 00s, every next generation yields double or even greater performance, love to see it.

>>101730598
>460
Is that after queuing up multiple images? Sometimes my first image after switching models takes over twice as long as all the ones that come after.

File: FLUX_978635544937398_00001_.png (627 KB, 768x640)

627 KB PNG

after extensive testing I've concluded that 1.25cfg is ideal for FLUX

>>101730640
Does it make everything look all washed out like that

>>101730640
doesnt look denoised

File: FLUX_565039417897099_00001_.png (739 KB, 768x640)

739 KB PNG

>>101730762
Sometimes. Not always. That's partly from the prompt.

Every CFG value is a compromise. I think 1.25 strikes the best balance.

File: schFLUX_00018-Australian (...).png (977 KB, 896x1088)

977 KB PNG

Last one from me, good night anons

>>101730813
have u seen this anon
https://www.reddit.com/r/StableDiffusion/comments/1ekgiw6/heres_a_hack_to_make_flux_better_at_prompt/

>>101730844
Night anon

>>101730729
Box?

I think trani is shit (as a human)

these all look like the typical instagram AI account.

Plastic skin, weird eyes, dead hair that looks like a wig

and why do all these pictures look like they been 'enhanced' by someone who can barely use photoshop?

please stop using energy to generate this crap and go jump into a volcano

>>101730813
see
>>101730787
>>101731920
>>101731931

>>101730313
>>101730183
now kith

some have said the beta scheduler works better, true or placebo?

File: 1691990810696413.png (1.2 MB, 1024x1024)

1.2 MB PNG

>call it Twitter one more time, bitch

>>101732203
>its not "to tweet" its "to xcrete"

>>101730183
Just filter him dude

>>101732022
see
>>101732168
>>101732227

File: Screenshot 2024-08-05 214604.png (304 KB, 1475x314)

304 KB PNG

>>101732163
What is the beta scheduler?
These are the only ones I have tried. Only uni_pc got a visibly distinct and useable result.

>>101732328
I guess it's like picking between euler or karras, not 100% sure but some have said it works a bit better for gens

File: Screenshot 2024-08-05 215322.png (3.23 MB, 2112x1054)

3.23 MB PNG

>>101732350
>>101732163
Feels like placebo. If you can guess which is Beta and which is Default then you have a better eye than me.

>>101730183
dude its debo there's a reason he gets ignored and filtered by 95% of all anons

File: Screenshot 2024-08-05 215641.png (2.96 MB, 2116x1061)

2.96 MB PNG

>>101732398
2nd example.
This is even more subtle.

File: image - 2024-08-01T033326.195.png (1.9 MB, 1488x840)

1.9 MB PNG

derp shark

>>101732163
I tested on my slow rig with with euler 10 steps to check how soon different schedulers start to converge, normal, simple and beta results looked very close to each other.

>>101732423
>subtle
nta but there's a pretty clear difference, what's your monitor resolution?

File: FD_00658_.png (1.24 MB, 1024x1024)

1.24 MB PNG

>>101732436

>>101732458
4k HDR. Obviously there's a difference but it's not as changed as >>101732398, where the props and jewellery changed, i.e more subtle.

File: image - 2024-08-01T032652.781.png (2 MB, 1488x840)

2 MB PNG

>>101732463
Street sharks

File: FD_00744_.png (1.23 MB, 1024x1024)

1.23 MB PNG

>>101732500
Thank you for being ~40. I loved this shit as a kid.

File: file.jpg (751 KB, 1024x1267)

751 KB JPG

leonardo scrape is finished except a few days that were missed for some reason, tasks are per hour (their pagination uses timestamps and an offset, using day ranges slowed down as offset increased)

>>> TASKS.count_documents(query_eq("complete", True))
18110
>>> TASKS.count_documents(query_not_exists("complete"))
116

>>> IMAGES.estimated_document_count()
953839291

checked distinct prompts last night, will be slightly off from the final count

>>> count_distinct(IMAGES, "generation.prompt")
228941006

shame they don't have 1B by themselves

>>101732600
Why did you do it?

File: file.jpg (389 KB, 1024x1792)

389 KB JPG

>>101732636
for research, the last big dataset of generations is diffusiondb from late 2022, it only has 14m and they're all sd1.4/1.5
also for the likes

>>101732683
Has anyone liked your data yet?

How to prompt a comic book cover with title in Flux?

File: file.jpg (492 KB, 1024x1792)

492 KB JPG

>>101732841
no >:(
one set from a math forum has a few downloads and some chinese researchers copied some of that data into their own dataset which has over 100 likes, they didnt credit me but i guess you could say i have likes via proxy from that one

what's the difference between autismmix and AAAAutism and other similar finetunes? Isn't autismmix already pony but with all the furry shit cut out and focused on anime?

>>101732958
it's a more anime focused checkpoint, that one is my go-to ponyXL model for making anime gens, regular and the confetti variation of autismmix. even without loras you can get good results, just use the booru tag of whatever character.

File: 00067-TFT_124021204.jpg (977 KB, 2048x2560)

977 KB JPG

bunny

>>101733302
have you had a play with Flux yet?

File: BMP_FLUX_03099_.png (1.58 MB, 1024x1024)

1.58 MB PNG

>>101733302
Kek

File: file.jpg (490 KB, 1792x1024)

490 KB JPG

>>> count_distinct(IMAGES, "generation.custom_model.name")
1049

neat

>>101733317
yeah I see uses for it especially as a first step before using SD.
No LoRAs yet limits it heavily though, ie I can't make accurate anime girls, if I could start training effectively I would give it a go.

File: 000000_15962_.png (2.24 MB, 1024x1536)

2.24 MB PNG

>>101729047
>Is this an earmuff-anon production? Looks great either way
kek, it's me trying to get Sams helmet/Visor on her head and still have that blonde hair showing.... Diff. considering FLUX cfg/neg. creating a workflow/testing with a negative prompt atm

>>101733342
yeah anons here have mentioned you'll need 1xA100 to train a LoRA and 3x or more for a finetune.

Is Flux the next logical step up for Pony diffusion? Looks like a pretty solid model.
>not even gonna try it on my 1080 though

>>101733359
I have two 3090s, if flash attention is properly integrated for dual GPU training I think it will be viable

>>101733362
completly different target goal, pony is a comic oriented fetish porn model, FLUX is a censored general purpose model

is a1111 flux being worked on with all the features? I can't be bothered going through all these comfy nodes for inpainting and such if it is all going to be done in a1111

>>101733302
Is the holding the phone in mouth from a lora?

>>101733517
no, controlnet and inpainting

File: Screenshot 2024-08-06 000805.png (53 KB, 825x408)

53 KB PNG

>>101733385
Well the dude who makes simpletuner says specifically it likes multi-gpu training.

>>101733541
good news, I foresee this being possible given some time and effort from the community because letting people work on LoRAs and getting controlnet in is huge for the model becoming more than local midjourney

File: file.jpg (555 KB, 1024x1792)

555 KB JPG

>>> TASKS.count_documents(query_not_exists("complete"))
0

finished

>>> IMAGES.estimated_document_count()
957610978

File: 00107.jpg (641 KB, 2480x3120)

641 KB JPG

3rd epoch, a little sad the volcano token turns the ground into pumice.

>>101733541
Flux Dev/Schnell are trained. You'll need to rip out the guidance conditioning and finetune it back to behaving like a Pro model. Once that's done this re-Pro model can be finetuned and LORA just like normal. And that would restore the ability to use negative prompts. The negative prompting is baked in atm. Given the size you'll need A100 40GB+ GPU.

can someboy prompt a 16bit-like tileset with grass and water? thanks guys, you'll have my everlasting love

File: 29269125.png (1.14 MB, 1024x1024)

1.14 MB PNG

>>101733753
prompt: 16bit-like tileset with grass and water

File: 288115309.png (2.44 MB, 1800x904)

2.44 MB PNG

>>101733809
lul

File: 0.jpg (572 KB, 1024x1024)

572 KB JPG

>>101733753

File: BMP_FLUX_03129_.png (1.06 MB, 1024x1024)

1.06 MB PNG

Wtf I hate flux now

>>101734346
That flux? How do you prompt it to recognize characters?

>>101729709
>doesn't even know how a magic game is supposed to look
to be fair, neither would the n*rmie f*males depicted

What are this general preferred SDXL models? I'm looking for some recommendations, I've only used ponydiffusion so far.

>>101734382

(Samus Aran:1.5 from Metroid) wearing a tight (zerosuit:1.3) with  colors of Red, green and black hues,, She is wearing a Full protective Helmet on her head,. her boots have small jetpacks enabling high jumping abilities,

File: 1710126353540557.png (1.34 MB, 960x1280)

1.34 MB PNG

My coworker dressed like this today and I couldn't stop staring.

>>101734511
Thanks. Does that emphasis work or is it your guessing?

>>101734556
My guessing, I always use it...kek

>>101734390
>replying to the thread schizo
thanks for being part of the problem you faggot nigger retard

>>101734542
>yandere skin walker in background
I don't feel to good about this

>>101730400
Holy shit, flux?

>Thread schizo breaks filters faster and faster
Maybe take a hint?

>>101734769
he's a california skeet nigger that slurps up homo doodoo then comes here

>>101734769
>breaks filters faster
What do you mean?

File: moji.jpg (2.9 MB, 5153x1643)

2.9 MB JPG

drop your top 10 'mojis

>>101735101
You're using the emojis in the prompt?

This general should be deleted. Stable Diffusion is obsolete now. It's like having a GPT-J general for LLMs when Llama is available.

>>101735166
so what was SD replaced by?

>>101733589
Have you tought about making your own extensive image database by leeching every single image off shutterstock, getty images etc and even pinterest?

>>101735185
Flux.

>>101730183
how to spot the newfriend

>>101735162
yeah for example the hotel/skyline there is just

can't remember how 4chan deals with emojis we'll see

>>101735166
this isn't a Stable Diffusion general, its a stable diffusion general. we are not affiliated with the model

File: Screenshot 2024-08-06 003451.png (14 KB, 430x208)

14 KB PNG

>>101735446
great...

>>101730183
welcome newfren
you just encountered the "anon" called thread schizo / debo
he will waste your time as much as he can and argues in bad faith
dont worry, every newfag falls for him once
my advice would be to simply ignore or filter him, it will not get better

>>101735166
Well it should at least be modified. But autists gonna autist.

>>101734700
SDXL

File: 000000_15990_.png (2.45 MB, 1137x1516)

2.45 MB PNG

>>101735451
Correct, we diffuse latent space in a stable and coherent fashion.

>>101735487
Can you make her sexy

>>101735459
Why not just go to the other general where he's been kicked out of?
Before flux dropped the general mostly him and PW and a few other undesirable avatarfags cope posting talking about their day to day lives.

how do i prompt flux? does it benefit greatly from boomer sentences or is it fine to just give it lists of nouns and adjectives? is it strongly affected by positioning/grammar of words? i'd just learn by fucking around with it but that's a lot of 40 second gens

>>101736102
I had better results with "boring snapchat pic circa 2015" than anything long

>>101733361
sigh... *unzips dick*
>>101733640
>Welcome... to the desert... of the real

>>101736092
im schizo anon and /sdg/ is my home ESL friend

File: file.jpg (638 KB, 1024x1792)

638 KB JPG

>>101735186
real world images are planned like many other things desu
soon:tm: trust the plan:tm: two more weeks:tm:

State of SD 3 Awesome Edition?

>>101736172
>cope post
>cope phase
This is why people hate posting here.

i did some tests and the regular model and inpaint model produce identical inpaint images

>>101729003
can you use LoRAs on bing?

>>101734835
Asuka Aran is best Bounty hunter

File: delux_bu_00001_.png (1.02 MB, 1344x768)

1.02 MB PNG

>>101736320
two weeks

File: 1722789727652211.jpg (680 KB, 1664x2432)

680 KB JPG

what are some good artist styles to use with pony? Someone posted this and I really wish I knew how they made it, link is so cuteeeeeee

>>101736525
stop being gay

>>101736893
Please do not necro bump, thanks

>>101736907

File: delux_bu_00003_.png (1.1 MB, 1344x768)

1.1 MB PNG

>>101736907
wdym

>>101736907
nig*bo needs to his general is dying

>>101736907
bump (Y)

How come when I try to create hyper realistic images they look like complete shit? I'm using easydiffusion.

>>101736937
i know im repeating myself but the basedbin should be in OP instead of the discord

>>101736953
I'll add it next time I bake thanks!

File: delux_bu_00004_.png (1.34 MB, 1344x768)

1.34 MB PNG

>>101736950
"hyper realistic", "hyperrealism", "photorealism" are art terms applying to paintings, drawings, sculptures, etc. when you use terms like these as prompt tokens, you're ironically introducing non-realism vectors

File: ComfyUI__00001_.png (954 KB, 768x1280)

954 KB PNG

goo morning

>>101736953
discord shouldn't even be in the OP, why are we shilling offsite walled gardens

File: deflux_mo_00001_.jpg (612 KB, 1344x960)

612 KB JPG

>>101737074
gm

>>101737118
this question has been asked a lot but never answered

>As of 16:30 UTC, Our Engineering team is investigating an issue with processing of events in all regions. During this time users may experience failed events on Droplets. We apologize for the inconvenience and will share an update once we have more information.
typical

>>101737118
You're right. The fact that it's on the op is forcing me to join, despite the fact I have free will to completely ignore it. God damn it. Guess we can just continue this conversation in the discord seeing as you was forced to join it as well.

We can't talk in the discord, since debo won't be able to chime in :(

File: delux_bu_00006_.png (1.22 MB, 1344x768)

1.22 MB PNG

>>101737218
I'm surprised they didn't just delete the discord after I left. what's even the point?

File: ComfyUI_00141_.png (1.16 MB, 960x1088)

1.16 MB PNG

code monkey go to job...
>>101737074

File: liecon.png (1.26 MB, 1188x1070)

1.26 MB PNG

lykon seething

>>101737195
I said shilling, not forcing, why are we as a long-running thread pushing this community-splintering clique-building forum? Put it in comments or whatever fine but having it in OP is an endorsement which makes no sense since this is ALREADY the place to discuss and post

>>101737118
>>101737337
Did you already forget that SAIs official discord was in OP for a long time before it was replaced?

File: file.png (803 KB, 1318x1010)

803 KB PNG

>Multiple Engineering teams are continuing to investigate the issue. Customers may also encounter errors when managing other services and using the API at this time.
but i need new nodes >:(
>>101737329
not defending lykon but the opinions of anyone with their face as their profile picture can be safely ignored, even more so when its an ai generated version of their face

>avatarfag opinion discarded

>>101737246
People are flooding to the discord because of you :]

File: delux_bu_00007_.png (980 KB, 1344x768)

980 KB PNG

>looks like you've activated my trap card, anon

>>101737514
that doesn't make sense. why would they flood the discord when I'm not even there?

File: schFLUX_00019-Pixel art s(...).png (920 KB, 832x1216)

920 KB PNG

Morning anons

My Python Lib folder is now 27.4GB, what can I safely remove (comfyui)

>>101737677
G'mornin Quakkas

>>101737329
move your company to a nonretarded country, fire all the useless safety retards, uncuck your models, take an easy win

File: FLUX_290344551039472_00001_.png (670 KB, 640x768)

670 KB PNG

>>101737329
Look how cooked his gens are. I feel like I'm going crazy. What % of the world population can't see it?

File: ComfyUI_00142_.png (1021 KB, 960x1088)

1021 KB PNG

>>101737324
snak time

File: 000000_16000_.png (2.08 MB, 1075x1434)

2.08 MB PNG

This was made with this Flux with negative workflow.https://files.catbox.moe/bcjkbn.json

>>101737329
SD3 has an API model too doesn't it? nobody mentioned it but he just assumed he's talking about it?

File: deflux_mo_00004_.jpg (202 KB, 1344x960)

202 KB JPG

>>101737677
gm

>>101737584
no one wants to deal with you kek

is sexy.ai the best AI porn generator online? are there any better options?

I had a good idea for a prompt when I went to bed last night, and I repeated it in my head so I wouldn't forget, but I forgot
I know it's two words, and they both start with S

>>101737914
Saint Satan?

File: delux_bu_00009_.png (1.23 MB, 1344x768)

1.23 MB PNG

>>101737836
thats not what your mom said
(she's very friendly and encouraging)

>>101737914
skibidi spongebob

>>101737876
>no loli
It's shit by default.

File: delux_tru_00018_.png (1.47 MB, 1344x768)

1.47 MB PNG

I still have more of these truespace gens too

File: Portrait pixel art style (...).png (631 KB, 832x1216)

631 KB PNG

>>101737829
Cute!

>>101737729
the foot tentacle

>>101738079
overcooked

File: Vramlet.jpg (370 KB, 1424x1024)

370 KB JPG

flux is so good

File: 2031-Little pot on the pi(...).png (1.2 MB, 1024x1024)

1.2 MB PNG

>>101738098
No. You can have food instead. Come eat.

>>101738090
>>101738090
>>101738090

File: delux_bu_00011_.png (1.24 MB, 1344x768)

1.24 MB PNG

look at this smug little fucker

>>101738204
me in the middle

File: delux_tru_00019_.png (1.41 MB, 1344x768)

1.41 MB PNG

>>101738297
lucky monke

File: 1722884527937_out-0.png (788 KB, 832x1216)

788 KB PNG

I'm helping!

File: 1835193247-flux1-schnell.png (1.43 MB, 1024x1024)

1.43 MB PNG

I thought my 12GB 3080 would be enough for flux dev, but fuck no. Not even close. Have to settle with Scnhell.
>pic of my dream when i'm getting a 4090 24GB

>>101737776
kittenbox?

so, how much room for improvement do you think there is?

>>101738528
>>101738061
>>101737677
damn these are pretty good. haven't been paying any attention to SD for a couple months. what's the process for getting nice pixel art like this?

>>101739443
It's a new model, it's called flux, you need over 20 gb of vram to run it properly but you can try it this way
https://replicate.com/black-forest-labs/flux-dev
Apparently 13gb vram is the minimum to get max quality out of FLUX dev locally
Btw for the pixel art, the prompt is in the filename of >>101738061 and >>101737677

>stable diffusion is already old news
>flux model can actually get human fingers and text right

its beyond fucking over

>>101739647
Is it 13gb or 20gb to run it properly?

>>101739877
if it cant generate boobs i'm not interested

>>101739877
24 is recommended, 13 is the minimum