[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: 1722800706851329.png (1.1 MB, 1024x1024)
1.1 MB
1.1 MB PNG
Previous /sdg/ thread : >>101779364

>Beginner UI local install
EasyDiffusion: https://easydiffusion.github.io
Fooocus: https://github.com/lllyasviel/fooocus
Metastable: https://metastable.studio

>Local install
Automatic1111: https://github.com/automatic1111/stable-diffusion-webui
ComfyUI: https://github.com/comfyanonymous/ComfyUI
SD.Next: https://github.com/vladmandic/automatic
AMD GPU: https://rentry.org/sdg-link#amd-gpu
Intel GPU: https://rentry.org/sdg-link#intel-gpu

>Use a VAE if your images look washed out
https://rentry.org/sdvae

>Run cloud hosted instance
https://rentry.org/sdg-link#run-cloud-hosted-instance

>Try online without registration
sd3: https://huggingface.co/spaces/stabilityai/stable-diffusion-3-medium
txt2img: https://www.mage.space
img2img: https://huggingface.co/spaces/huggingface/diffuse-the-rest

>Models, LoRAs & upscaling
https://civitai.com
https://huggingface.co
https://aitracker.art
https://openmodeldb.info

>Black Forest Labs: Flux
https://huggingface.co/black-forest-labs/FLUX.1-schnell
https://huggingface.co/spaces/black-forest-labs/FLUX.1-schnell
https://comfyanonymous.github.io/ComfyUI_examples/flux

>Index of guides and other tools
https://rentry.org/sdg-link
https://rentry.org/rentrysd

>View and submit GPU performance data
https://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html
https://docs.getgrist.com/3mjouqRSdkBY/sdperformance

>Share image prompt info
4chan removes prompt info from images, share them with the following guide/site...
https://rentry.org/hdgcb
https://catbox.moe

>Discord
6wUwtcJsr2

>Related boards
>>>/h/hdg
>>>/e/edg
>>>/d/ddg
>>>/b/degen
>>>/vt/vtai
>>>/aco/sdg
>>>/trash/sdg
>>
File: ComfyUI_00021_.png (2.49 MB, 1280x1024)
2.49 MB
2.49 MB PNG
dat cat lel
>>
>>101785853
>>Discord
>6wUwtcJsr2
Wat is?
>>
>>101785871
you dont even want to know
>>
File: 1.jpg (304 KB, 1536x1536)
304 KB
304 KB JPG
>>
File: forgeflux.png (9 KB, 609x129)
9 KB
9 KB PNG
tick tock comfy... ahahahha
>>
File: da nooz 24.jpg (487 KB, 1344x960)
487 KB
487 KB JPG
>mfw Resource news

08/08/2024

>TinyVAE Training Repository
https://github.com/XmYx/tinyvae-flux

>InstantX / FLUX.1-dev-Controlnet-Canny-alpha
https://huggingface.co/InstantX/FLUX.1-dev-Controlnet-Canny-alpha

>Grounded SAM 2: Ground and Track Anything in Videos with Grounding DINO and SAM 2
https://github.com/IDEA-Research/Grounded-SAM-2

>ComfyUI xlabs Flux ControlNet implementation
https://github.com/comfyanonymous/ComfyUI/pull/4260

>"Forge is under a week of major revision. Backend Rewrite is 90% finished"
https://github.com/lllyasviel/stable-diffusion-webui-forge#under-construction

>Fast Sprite Decomposition from Animated Graphics
https://cyberagentailab.github.io/sprite-decompose

>Concept Conductor: Orchestrating Multiple Personalized Concepts in T2I Synthesis
https://github.com/Nihukat/Concept-Conductor

>D2Styler: Advancing Arbitrary Style Transfer with Discrete Diffusion Methods
https://github.com/Onkarsus13/D2Styler

>ComfyUI-EbSynth: Fast Example-based Image Synthesis and Style Transfer
https://github.com/FuouM/ComfyUI-EbSynth

08/07/2024

>ControlNet Canny FLUX.1-dev
https://huggingface.co/XLabs-AI/flux-controlnet-canny

>Diffusion1B: Dataset of 1.2 billion images generated by diffusion models
https://huggingface.co/datasets/bigdata-pw/Diffusion1B

>Diffusers v0.30.0: New Pipelines, New Methods, and New Refactors
https://github.com/huggingface/diffusers/releases/tag/v0.30.0

>Diffree: Text-Guided Shape Free Object Inpainting with Diffusion Model
https://github.com/OpenGVLab/Diffree

>LLaVA-OneVision Easy Visual Task Transfer
https://llava-vl.github.io/blog/2024-08-05-llava-onevision

>IP Adapter Instruct: Resolving Ambiguity in Image-based Conditioning using Instruct Prompts
https://unity-research.github.io/IP-Adapter-Instruct.github.io

>SiCo: A Size-Controllable Virtual Try-On Approach for Informed Decision-Making
https://github.com/SherryXTChen/SiCo

>Photorealism LoRA for FLUX.1-dev
https://huggingface.co/XLabs-AI/flux-RealismLora
>>
>mfw Research news

08/08/2024

>How Well Can Vision Language Models See Image Details?
https://arxiv.org/abs/2408.03940

>Lightweight Video Denoising Using a Classic Bayesian Backbone
https://arxiv.org/abs/2408.03904

>Target Prompting for Information Extraction with Vision Language Model
https://arxiv.org/abs/2408.03834

>Advancing Multimodal Large Language Models with Quantization-Aware Scale Learning for Efficient Adaptation
https://arxiv.org/abs/2408.03735

>Soft-Hard Attention U-Net Model and Benchmark Dataset for Multiscale Image Shadow Removal
https://arxiv.org/abs/2408.03734

>Openstory++: A Large-scale Dataset and Benchmark for Instance-aware Open-domain Visual Storytelling
https://openstorypp.github.io/

>TALE: Training-free Cross-domain Image Composition via Adaptive Latent Manipulation and Energy-guided Optimization
https://arxiv.org/abs/2408.03637

>A comparative study of generative adversarial networks for image recognition algorithms based on deep learning and traditional methods
https://arxiv.org/abs/2408.03568

>Hybrid diffusion models: combining supervised and generative pretraining for label-efficient fine-tuning of segmentation models
https://arxiv.org/abs/2408.03433

>Set2Seq Transformer: Learning Permutation Aware Set Representations of Artistic Sequences
https://arxiv.org/abs/2408.03404

>A Non-negative VAE:the Generalized Gamma Belief Network
https://arxiv.org/abs/2408.03388

>FastEdit: Fast Text-Guided Single-Image Editing via Semantic-Aware Diffusion Fine-Tuning
https://arxiv.org/abs/2408.03355

>An Empirical Comparison of Video Frame Sampling Methods for Multi-Modal RAG Retrieval
https://arxiv.org/abs/2408.03340

08/07/2024

>MDT-A2G: Exploring Masked Diffusion Transformers for Co-Speech Gesture Generation
https://arxiv.org/abs/2408.03312

>DopQ-ViT: Towards Distribution-Friendly and Outlier-Aware Post-Training Quantization for Vision Transformers
https://arxiv.org/abs/2408.03291
>>
File: 00013-3790678055.jpg (296 KB, 1024x1360)
296 KB
296 KB JPG
>>
>>101785916
if forge can't do multi-gpu shit like on comfy I'm not leaving
https://reddit.com/r/StableDiffusion/comments/1el79h3/flux_can_be_run_on_a_multigpu_configuration/
>>
File: ComfyUI_00022_.png (2.05 MB, 1280x1024)
2.05 MB
2.05 MB PNG
CHOOSE YOUR FIGHTER

>>101785916
oh good, i can quit this spaghetti soon
>>
>set CUDA_VISIBLE_DEVICES= requires a third party node
baka
>>
File: 00004-2853947493.jpg (267 KB, 1024x1408)
267 KB
267 KB JPG
>>
>>101785916
The retard broke all extensions so you are better off using swarmui lol.
>>
File: 00006-1712675235.png (1.16 MB, 1024x1360)
1.16 MB
1.16 MB PNG
>>
File: garden.png (2.12 MB, 1024x1024)
2.12 MB
2.12 MB PNG
Could any anon share their workflow? (SD1.5/XL/XLPony/Flux)
I am curious about your generation recipes, i doubt the average workflow uses tools like Florence2 post-processing or Image CA sharpening like in >>101785065.
>>
>>101786047
Oh no we won't be able to super duper upscale our custom pozed big titty cartoon girls or whatever
>>
Just doing my part
https://pastebin.com/NYdMmBGH
>>
>>101786047
for now... dude has been rebuilding the entire backend so comfy cant come up later stating "ACKTUALLY FORGE IS USING THE COMFY BACKEND", ohh how the table turns, cant wait until forge runs SD3.1 and comfy cant lmao
>>
nb4 he mass reports
>>
and now, he lays low until he can discuss something else completely ignoring everything
>>
File: 00025-3912998804.jpg (652 KB, 1248x1656)
652 KB
652 KB JPG
mfw like debo even more
>>
File: 1475714351643.jpg (57 KB, 482x549)
57 KB
57 KB JPG
What are some labels to make a plump girl look plump? The definition of plump on most models looks like "she just eats" and if I use "fat" she is completely over-proportioned.
>>
Morning anons
>>
File: ComfyUI_10446_.png (2.23 MB, 1440x1120)
2.23 MB
2.23 MB PNG
>>
>>101786273
plump, curvy
>>
File: ComfyUI_00027_.png (1.35 MB, 1280x960)
1.35 MB
1.35 MB PNG
>>
File: file.jpg (275 KB, 1792x1024)
275 KB
275 KB JPG
>>101785919
it was real if you made the news
>>
File: file.jpg (307 KB, 1536x1536)
307 KB
307 KB JPG
>>
>>
File: OIG4.jtAUwIT8_SreV.jpg (194 KB, 1024x1024)
194 KB
194 KB JPG
>>
>tfw no ayy lmao lora
why even live
>>
File: ComfyUI_10450_.png (2 MB, 1440x1120)
2 MB
2 MB PNG
>>
File: ComfyUI_00028_.png (775 KB, 1280x960)
775 KB
775 KB PNG
not at all what i prompted for but ok
>>
File: ComfyUI_10452_.png (1.75 MB, 1440x1120)
1.75 MB
1.75 MB PNG
>>101786336
cheeen
>>
File: dealOfSomething.jpg (330 KB, 894x899)
330 KB
330 KB JPG
When will sdxl_forge have a logical ((lightweight)) LLM that dalle or flux have?
Technically possible, but would it cost money to train them?
>>
File: delux_bu_00036_.png (1.28 MB, 1344x768)
1.28 MB
1.28 MB PNG
>>101786137
doesn't need to be a competition. would be great for there to be a second UI that stays up to date with latest happenings so people have an option to use whatever they want if they want the newest toys. illy seems too flaky to stay on top of the project longterm though

>>101786143
if you really thought I was omnipotent, why would you even taunt me? you should be scared if you believed your own fictions

>>101786252
what would you like to talk about?

>>101786291
gm

>>101786329
keep us up to date as you push out more datasets so I can make sure to tag them in the news. you were doing flickr next?
>>
>tfw mass reply word salad
>>
>>101786137
>for now... dude has been rebuilding the entire backend so comfy cant come up later stating "ACKTUALLY FORGE IS USING THE COMFY BACKEND
why should a user give a shit about the underlyings of a product? when I buy a phone or a TV or whatever, it never comes into my mind what's inside and who copied who, I just care about the results
>>
File: ComfyUI_10444_.png (2.01 MB, 1440x1120)
2.01 MB
2.01 MB PNG
>>101786436
shut the fuck up debo, comfy has always made this a competition, ever since he was posting his shitty metrics or spamming "the future is comfy" with the failed SD3 model, also he was the one bad mouthing Illya, also I wont reply to you so dont waste your time replying to me
>>
>>101786486
>never care about product just consume product
ngmi
>>
>>101786486
yeah because you're a dumb coomsoomer
>>
>>101786490
>>101786503
why do you imply I don't care about the product? I care, I literally said that I care about the RESULTS, that's the only thing that matter, are you one of those guys who dismember a TV to look at the inside after buying it? the fuck?
>>
File: file.jpg (432 KB, 1792x1024)
432 KB
432 KB JPG
>>101786436
yeah flickr next, inspired by a 10m flickr set with so called high res images that are actually one of the lowest resolutions available
i fell back to sleep this morning and went out this afternoon though. the research stage is done like the requests to make, just need to put it together and scale
there's a getty scrape in progress, ~2m so far, unfortunately those are max 612x612 until i figure out how to rip the full version without watermark, but good enough for pretraining or training a watermark remover
>>101786486
it's not so important to the end user but it is annoying to have comfy acting the way he does
>>
Can someone please share a Comfy workflow Flux in mid-end PCs?
>>
File: 1704304179681820.png (1.24 MB, 1024x1024)
1.24 MB
1.24 MB PNG
it's neat how many prompts work as you'd think

in-game screenshot of the game World of Warcraft main menu. There are 3 buttons. The first button says "Enter Game". The second button says "Settings". The third button says "Quit". In the background is the city of Silvermoon from World of Warcraft Burning Crusade, and a female Blood Elf in Paladin armor.
>>
>>
File: ComfyUI_10457_.png (1.95 MB, 1440x1120)
1.95 MB
1.95 MB PNG
>>
>>101786563
for recent games yeah it works fine, but I'm a boomer-millenial who've grown up in the 90's early 00's, and I'd love Flux to make really low poly ps1 style, but it can't do that, or I just haven't found the way yet
>>
File: 1710358037595801.png (1.09 MB, 1024x1024)
1.09 MB
1.09 MB PNG
>>101786563
>>
>>101786545
What did he do?
>>
>>101786047
just wait until the reforge chad port flux from forge then
>>
File: 1697008155805756.png (1021 KB, 1024x1024)
1021 KB
1021 KB PNG
>>101786603
you can even influence the logo graphic if you tell it to include stuff.

in-game screenshot of the game World of Warcraft main menu. There are 3 buttons. The game logo includes the text "adventures with Flux". The first button says "Enter Game". The second button says "Settings". The third button says "Quit". In the background is the city of Silvermoon from World of Warcraft Burning Crusade, and a female Blood Elf in Paladin armor.
>>
>>101786561
mid end and low end are all the same
>>
>>101786598
someone was getting some good ocarina of time styled gens but I don't know how
>>
File: 1702718193076804.png (692 KB, 1024x1024)
692 KB
692 KB PNG
it seems to know the deus ex with jensen but I wonder if it knows the original
>>
>>101786598
follow the reply chain, the post i replied to was a reply to another anon's post >>101786137
forge included some code from comfy for compatibility or something and comfy lost his shit as usual insisting forge is just using comfy as a backend
his attitude sucks desu, like he tried shitting on some other ui because it used diffusers' modeling, when he just copypastes the original research implementation anyway which is basically the same as importing from diffusers
>>
>>101786508
To give a less hyperbolic example, anon: people on this board will check the specs of a system before buying it or build their own. Because we care about the product - not just the fact it can be a pc.
>>
File: mech06.jpg (138 KB, 765x776)
138 KB
138 KB JPG
>generates a free drawing in a few minutes that
>would take a professional illustrator hours or days to draw.
>and yet I complain that it is slow.
What a poor gtx1060...
>>
>>101786718
but at no point in time I'm gonna care about who copied who on the inside of a product, if it works it works, that's how science evolves, when you look at a researsh paper, there's dozens of citations of other work, you derive a work and make it better, that's how it always worked
>>
>>101786708
true, Comfy is hypocritical, he uses python packages which is basically the code of someone else, and then he starts to have a stroke when someone else copy his code? LOOOOOOOOL
>>
File: flux_comfyui_mt_00003_.png (1.35 MB, 1280x960)
1.35 MB
1.35 MB PNG
i dont know what's going on anymre
>>
File: 1718564539261352.png (602 KB, 1024x1024)
602 KB
602 KB PNG
progress
>>
>>101786748
you'd care if you bought a samsung tv and an angry hisense representative kept banging on your door shouting about how samsung copied some irrelevant thing
>>
>>101786790
kek
>>
>>101786762
if he doesn't want it directly copied he should make it a package like actual backends e.g. diffusers
>>
File: flux_comfyui_mt_00004_.png (1.72 MB, 1280x960)
1.72 MB
1.72 MB PNG
>>
>>101786818
He's mentally ill
>>
File: ComfyUI_171019_.png (876 KB, 1024x1024)
876 KB
876 KB PNG
>>101786762
That's not the part that makes him an idiot. He's going around claiming that forge doesn't use ComfyUI when the entire backend is ComfyUI code.
That's as dumb as if I was going around claiming ComfyUI wasn't using pytorch.
>>
>>101786893
>>101786890
The irony
>>
>>101786893
>the entire backend is ComfyUI code.
it's it true? I have a hard time believing that
>>
>>101786914
It's not and he even got laughed off of reddit. His janny ani got so asshurt that he tried to false report the project too.
>>
>>101786893
help yourself by keeping your chilling out and keeping your mouth shut desu
>>
>>101786893
tick tock
>>
File: ComfyUI_171028_.png (1.16 MB, 1024x1024)
1.16 MB
1.16 MB PNG
>>101786914
If you don't believe me look at what the reforge dev is saying. Most of the reforge improvements where just from updating the ComfyUI code and porting custom nodes over.
>>
File: flux_s__00102_.png (1.29 MB, 768x1024)
1.29 MB
1.29 MB PNG
>>101786034
ooh this backlit effect is lovely, catbox?

>>101786414
my girlfriend says "is she just in an ocean of piss?"

flux isn't a panacea sadly, even 50 steps euler karras still fucks up hands and i'm often getting this weird red tint
>>
>>101787042
>my girlfriend
pics?
>>
>>101787031
having code which he credited you for and using only your code is two different things autistic sperg.
>>
>>101787042
50 steps karras is mostly a meme, it fries a lot of images. imo the best use for it, is for creating images with texture, and only some styles benefit from that.
>>
File: 1703672748636757.png (972 KB, 1024x1024)
972 KB
972 KB PNG
in-game screenshot of the game Final Fantasy 6 main menu. There are 3 buttons. The first button says "Start". The second button says "Settings". The third button says "Quit". In the background is the city of Narshe from Final Fantasy 6.

it's not exactly FF6 but it got the title font perfect, pretty cool, almost a fusion of 6 and 7 logo
>>
>>101787133
>50 steps karras is mostly a meme, it fries a lot of images.
true, I had some "cope time" with karras because it gave me really different pictures compared to the other schedulers, so I thought that understanding it better would lead to new and fun territory, oh boy how wrong I was, you have to understand I was young and stupid back in the days (3 days ago) kek
>>
>>101787146
it's not exactly FF6 but it got the title font perfect
that's probably the only thing it knows about FF, I tried to ask for any character from FF10 I got nothing correct
>>
File: delux_me_00005_.jpg (1000 KB, 1344x960)
1000 KB
1000 KB JPG
reposting cuz I'm addicted to this gen

https://suno.com/song/035d4236-21c8-4f0b-bec9-e0451fa92a3b
>>
File: flux_s__00073_.png (559 KB, 480x640)
559 KB
559 KB PNG
>>101787133
>>101787180
yeah this seems to be correct, i've been having good results with it on schnell tho, this is 6 steps
>>
File: file.jpg (317 KB, 1024x1792)
317 KB
317 KB JPG
>>101787200
it is pretty neat
reposting these desu
https://suno.com/song/267e7d2f-5cfe-4f9c-b424-a596384cf867
https://suno.com/song/73dfeb9b-d1c2-4654-9b97-606eacd3a383
https://suno.com/song/d7ca010e-6451-4048-a916-6d05601ae10e
>>
File: flux_s__00145_.png (1.24 MB, 768x1024)
1.24 MB
1.24 MB PNG
>>101787201
flux rly loves it's fingernails tho
>>
File: hlky.png (95 KB, 1745x459)
95 KB
95 KB PNG
>>101787256
you're hlky? I used your first stablediffusion repos back in the day, i still have them in my hoarding folders
>>
File: 1696210614008651.png (963 KB, 1024x1024)
963 KB
963 KB PNG
>>101787146
this one is closer, prompted castle in the desert like the game
>>
File: file.jpg (275 KB, 1536x1536)
275 KB
275 KB JPG
>>
File: 2024-08-06_00190_.png (2.77 MB, 1024x1680)
2.77 MB
2.77 MB PNG
>>101787357
try roman numerals VI .. it works quite well in FLUX
>>
>>101787357
can you provide the prompt anon, I wanna change it into ff10 and see if it knows some of its world
>>
>>101786893
ah this bullshit claim again
>>
File: ComfyUI_01704_.png (1.35 MB, 1024x1024)
1.35 MB
1.35 MB PNG
>>101787386
Flux can nail all the first 12 numerals number, the exception is IV though kek
>>
File: flux_comfyui_mt_00008_.png (2.32 MB, 1280x1024)
2.32 MB
2.32 MB PNG
>>
File: file.jpg (338 KB, 1792x1024)
338 KB
338 KB JPG
>>101787352
yeah that me. cool. i think zerocool is still keeping the webui going. i do data and (H)AITemplate now
>>
File: 2024-08-06_00205_.png (2.58 MB, 1024x1680)
2.58 MB
2.58 MB PNG
>>101787403
I had troubles with XIII .. IV worked all the time.. had to replace XII with 13 for my Miku cards
>>
File: ComfyUI_01707_.png (1.45 MB, 1024x1024)
1.45 MB
1.45 MB PNG
>>101787435
You can see how obvious it has trouble with the IV one when you try to make a clock
>>
>>101787403
the universal contestant that all clocks must be depicted as 10:10 also applies to latent space
>>
File: delux_me_00008_.jpg (993 KB, 1344x960)
993 KB
993 KB JPG
>>101787256
are these all using lyricsgpt? I used lyricsgpt as a base for mine and it gave a really good structure but diverged drastically from the input, so I ran the output through claude to reintroduce data language. fun fact: that lyricsgpt run on hf took 27min lol. I know you said I should jsut run it local but I don't have an llm setup

>>101787357
very nice

>>101787386
>>101787435
good tarots. I tried to get some flux magic cards previously but I think I was describing the artwork in too much detail that it forgot to draw the card itself. do you have a pretty light-weight prompt for these?
>>
File: ComfyUI_10470_.png (1.17 MB, 1024x1024)
1.17 MB
1.17 MB PNG
>>101787426
can't believe its going to be 2 years already
>>
>>
SDG song
https://suno.com/song/b8cca9c8-1c55-4104-9357-b5942ae569ff

[Verse]
Morning light we log on fast
S.D.G posts here to last
Captchas check we jump right in
A.I. art our daily spin

[Verse 2]
Thread by thread we find our way
Trollfights come we laugh and play
Imaginations set ablaze
In our world we spend our days
>>
File: ComfyUI_10476_.png (1.29 MB, 1080x1280)
1.29 MB
1.29 MB PNG
>>
File: ComfyUI_10477_.png (1.68 MB, 1080x1280)
1.68 MB
1.68 MB PNG
>>
File: file.jpg (273 KB, 1792x1024)
273 KB
273 KB JPG
>>101787508
yeah those are all with lyricsgpt, this one too
https://suno.com/song/c5fb3e77-1fbc-4e1b-bef9-bed384a6aed2
the next version will be pretrained with general knowledge then more lyrics with filtering to help with the bias, and im going to get gpt-4o-mini to generate a description for a subset so it can be trained to work from a description
might scrape pitchfork reviews, could be useful pretraining
https://github.com/bigdata-pw/LyricsGPT is the same as the hf demo just without the ratings
>>101787538
pretty crazy how far it's all come from gradio scripts shared in pastebin lol
>>101787604
>>101787620
these are nice
>>
>>101787646
Your data is obsolete, btw.
>>
File: ComfyUI_10481_.png (2.52 MB, 1200x1600)
2.52 MB
2.52 MB PNG
>>
>>101786273
(Morbidly obese:1.2)
>>
File: 2024-08-06_00172_.png (2.98 MB, 1080x1680)
2.98 MB
2.98 MB PNG
>>101787508
>good tarots. I tried to get some flux magic cards previously but I think I was describing the artwork in too much detail that it forgot to draw the card itself. do you have a pretty light-weight prompt for these?
Thanks I dont.. I look at the Rider-Waite ones and discribe them starting with
>Hatsune Miku in the Tarot card "DEATH" below, card number "13" above, in the style of Rider Waite
where you replace DEATH with the name of the card and number with either roman numeral or arabic number .. the rest is creative T5 boomer prompting
>>
>>101787705
i see the taylor in this prompt
>>
>>
File: 2024-08-08_00004_.png (1.28 MB, 1280x720)
1.28 MB
1.28 MB PNG
>>101785853
>>
File: file.jpg (360 KB, 1024x1792)
360 KB
360 KB JPG
>>
File: ComfyUI_10486_.png (2.39 MB, 1200x1600)
2.39 MB
2.39 MB PNG
>>101787812

I dont think taylor swift exists in flux
>>
>>101787391

it's something like this

in-game screenshot of the game Final Fantasy 6 main menu. There are 3 buttons. The first button says "Start". The second button says "Settings". The third button says "Quit". In the background is a castle in the desert from Final Fantasy 6.
>>
File: ComfyUI_10493_.png (2.94 MB, 1200x1600)
2.94 MB
2.94 MB PNG
>>
wee thanks anon
>>
File: 00064-1392605818.png (3.15 MB, 1248x1824)
3.15 MB
3.15 MB PNG
>>
>/sdg/
>Everyone is talking about flux
>>
>>101788009
oops bye
>>
File: ComfyUI_10498_.png (2.96 MB, 1200x1600)
2.96 MB
2.96 MB PNG
>>101788009
we are waiting for SD3.1
>>
File: BMP_FLUX_03940_.png (1.02 MB, 1024x1024)
1.02 MB
1.02 MB PNG
Evening
>>
File: ComfyUI_10499_.png (3.02 MB, 1200x1600)
3.02 MB
3.02 MB PNG
>>
>>101788245
kek
>>
File: ComfyUI_01719_.png (397 KB, 1024x1024)
397 KB
397 KB PNG
>>101787890
For flux, every FF has the FF7 logo kek
>>
is this a skill issue or is it very hard to prompt to make two characters do what i want the to do? inpainting is difficult too because its hard to make the whole image flow together. i wish i just learned how to draw
>>
>>101785880
Should join?
>>
>>101788264
yeah it wants to make the meteor logo but you can also get just the text and other stuff, depends
>>
>>101788264
are you sure? pic related official art
>>
State of SD 3.1: Awesome edition?
>>
>>101788324
fool FF6 advent children does not exist, thats AI art, FF7 advent children is a thing
>>
>>101788324
>>101788341
we are finally entered post history era
>>
File: ComfyUI_01720_.png (1.53 MB, 1024x1024)
1.53 MB
1.53 MB PNG
Kek
>>
File: ComfyUI_10503_.png (2.41 MB, 1600x1200)
2.41 MB
2.41 MB PNG
>>
File: flux_comfyui_mt_00017_.png (2.38 MB, 1024x1280)
2.38 MB
2.38 MB PNG
UNITINU
>>
File: ComfyUI_10505_.png (2.72 MB, 1600x1200)
2.72 MB
2.72 MB PNG
>>
File: flux_comfyui_mt_00018_.png (1.9 MB, 1024x1280)
1.9 MB
1.9 MB PNG
>>
do any of free ver of SD generate porn
>>
File: BMP_FLUX_04115_.png (1.78 MB, 1024x1024)
1.78 MB
1.78 MB PNG
>>
File: 1705693396220603.png (1.33 MB, 1024x1024)
1.33 MB
1.33 MB PNG
neat
>>
File: 2024-08-08_00566_.jpg (1.35 MB, 5120x2880)
1.35 MB
1.35 MB JPG
>>
File: BMP_FLUX_04543_.png (1.52 MB, 648x1200)
1.52 MB
1.52 MB PNG
>>
File: ComfyUI_01723_.png (1.11 MB, 1024x1024)
1.11 MB
1.11 MB PNG
>>101788552
Holy fuck that one is fucking close!
>>
>>101788765
cool, every day im learning flux can do some new style.
>>
>>101788071
>we are waiting for SD3.1
Dead company
>>
File: flux_comfyui_mt_00023_.png (1.58 MB, 1280x768)
1.58 MB
1.58 MB PNG
>>
>>101787595
Cringe
>>
File: BMP_FLUX_04452_.png (1.09 MB, 1024x1024)
1.09 MB
1.09 MB PNG
>>
>>101788902
wb
>>
very schizophrenic in here
>>
File: 1712661669300734.png (1.2 MB, 1024x1024)
1.2 MB
1.2 MB PNG
>>
File: redeem.jpg (842 KB, 2500x3333)
842 KB
842 KB JPG
the last hope of SAI
>>
File: flux_comfyui_mt_00029_.png (1.26 MB, 1280x768)
1.26 MB
1.26 MB PNG
>>
File: BMP_FLUX_05073_.png (1.46 MB, 648x1200)
1.46 MB
1.46 MB PNG
>>101788974
Heya. Now freeze, sucka
>>
File: Doll Dai.jpg (304 KB, 1280x1856)
304 KB
304 KB JPG
>>
File: file.png (3.25 MB, 1248x1824)
3.25 MB
3.25 MB PNG
Is this deal worth it, /sdg/?
>>
not trolling but i still dont get why debo adds the discord to OP all the time
>>
File: 1717296482527565.png (1.21 MB, 1024x1024)
1.21 MB
1.21 MB PNG
>>
File: Dai & Dumb.jpg (318 KB, 1280x1856)
318 KB
318 KB JPG
>>
File: delux_me_00011_.jpg (247 KB, 1344x960)
247 KB
247 KB JPG
- flux doesn't know low poly
- flux doesn't know gordon freeman
- flux doesn't know half life
this is the closest I could get
>>
File: 1723148318600798.jpg (753 KB, 2500x3333)
753 KB
753 KB JPG
>>101789147
>>
>>101789477
>super cokejeek
>>
Just came back after a long hiatus, is SD3 bad? Are the censorship claims true?
>>
>>101789219
Grabbing > kiss
>>
File: Pizza.jpg (324 KB, 1536x1536)
324 KB
324 KB JPG
>>
File: delux_me_00013_.jpg (359 KB, 1344x960)
359 KB
359 KB JPG
>>101789562
yea it was a trainwreck. the censorship was the least of the problems, it could barely produce coherent images or follow prompts. SAI also attached a useless license to it so people couldn't use it for training and even civit didn't want to host SD3 images out of legal fears

SAI went on an apology tour and said "uhh actually that was just a beta!" and that they'll release a better SD3 'soon'. its a huge mess for them, not even considering flux coming out and cosmic dunking on them
>>
File: 00206-2502623800.png (3.05 MB, 1344x1728)
3.05 MB
3.05 MB PNG
>>
File: BMP_FLUX_04220_.png (1.83 MB, 1024x1024)
1.83 MB
1.83 MB PNG
>>101787974
>>101789602
>>101789696
Nice
>>
File: catbox_5133pn.png (3.75 MB, 1344x1728)
3.75 MB
3.75 MB PNG
>>
File: delux_me_00025_.jpg (548 KB, 1344x960)
548 KB
548 KB JPG
>>
>>
i wish debo would just fuck off forever
>>
>>101786345
flux?
>>
File: schizo-thread.jpg (234 KB, 1024x768)
234 KB
234 KB JPG
>>
>>101790187
He's crying in the other thread now
>>
File: delux_bu_00037_.png (1.09 MB, 1344x768)
1.09 MB
1.09 MB PNG
>>101790287
I don't keep that thread open
>>
>>101790307
sure thing thread schizo
>>
File: image.png (1.1 MB, 1152x896)
1.1 MB
1.1 MB PNG
>>101789630
>SAI went on an apology tour and said "uhh actually that was just a beta!" and that they'll release a better SD3 'soon'.
That is actually a good thing, it shows the backlash worked, at least.
>>
>>101790307
Why do you always add the discord to OP debo?
>>
>>101790307
This you?
>>101787954
>>
File: delux_tru_00028_.png (1.34 MB, 1344x768)
1.34 MB
1.34 MB PNG
>>101790323
hopefully. they've had a habit of talking more than delivering though. there's also the possibility that they do earnestly want to unfuck SD3 but even their best effort model can't compete with flux and is DoA

>>101790346
why do you ask the same question everyday when you're never going to accept the answer? this is like a zen koan
>What purpose does a question serve if no answer is acceptable?
>>
EasyDiffusion failed at running flux
Fooocus failed as well

Can someone just save me some time and tell me what SD local install can run Flux? Don't even have to tell me how to install, I will look it up
>>
>>101790378
i wonder what bs he'll reply with to this
>>
>>101790384
When you weren't here for a week the mysterious "baker anon" was also absent. How do you explain that thread schizo?
>>
>>101790346
it annoys the most people it's in his character to be a speccy loser
>>
File: de_iw_bo_00094_.png (2.75 MB, 1344x1728)
2.75 MB
2.75 MB PNG
>>101790378
>see new thread in catalog
>open thread
>>mfw
>close thread
was this supposed to be some kind of gotcha? lol

>>101790391
only comfyui supports it right now. forge will probably support it soon, the dev is doing a major code push
>>
File: 00188-AYAKON_12402137.png (1.5 MB, 1280x1280)
1.5 MB
1.5 MB PNG
>>
>>101790472
lonely loser
>>
File: file.jpg (403 KB, 1024x1792)
403 KB
403 KB JPG
so sleepy. i got monster and freddos
>>
>>101790483
>lonely
he's got PW :] oh and the random unaware newfags
>>
>>101790487
Big data never sleeps, remember this.
>>
File: 00295-jak.jpg (2.67 MB, 2048x2432)
2.67 MB
2.67 MB JPG
>>101790472
You're pathetic and knowing that I live rent free to the point you burn everything around yourself for years will never not be funny
>>101790483
Hear Hear
He needs to stay in this containment thread instead of anon posting with the same exact pathetic style

He never improves
He always seethes
And he always cries
Just by me posting here I give his worthless existence meaning which is sad
>>101790506
>PW
That loser seems to not care about me but call for me whenever he gets a normal human reaction to his mental illness
>>
so wtih 16GB VRAM how should i set up flux?
>>
File: file.jpg (297 KB, 1024x1792)
297 KB
297 KB JPG
>>101790521
my nodes and scripts are always online
>>
File: deuni_00050_.png (2.7 MB, 2016x1152)
2.7 MB
2.7 MB PNG
>>101790532
thank you for the bump! aren't kitsunes meant have 9 tails?
>>101790506
I love PWkinnies xd
>>
File: delux_bu_00038_.png (1.12 MB, 1344x768)
1.12 MB
1.12 MB PNG
>>101790545
follow the fp8 options here:
https://comfyanonymous.github.io/ComfyUI_examples/flux/#tips-if-you-are-running-out-of-memory
>>
>>101790200
nahh sdxl pony mix
fuck flux
>>
>inb4 trani pedo posting
>>
>>101790577
>>101790580
Well this is disturbing
>>
>>101790532
>posting the same image for over a year and acting smug
Please, stay in your discord.
>>
>>101790572
They are your children.
>>
File: trani.jpg (5 KB, 225x225)
5 KB
5 KB JPG
>>101790595
>>
File: BMP_FLUX_05273_.png (1.16 MB, 1024x1024)
1.16 MB
1.16 MB PNG
>>101790545
I'm trying out gobbo-chan's WF but its pretty slow for my 8 VRAM
https://files.catbox.moe/62alp0.png
>>
>>101787349
Lack of gloves in the training data, my guess is that it's another victim of the "no nsfw" material.
>>
>the only time sdg isn't dead is when anon is hyped up on drama
Let it die
>>
>too scared to reply to ran
>too scared to join the discord in OP
>afraid of accountability
>>
>bmp randomly here when the drama starts
im just noooticing
>>
>>101789467
>flux doesn't know gordon freeman
This is so damn retarded...
>>
Hello, /g/riends. Just installed AI yesterday. How do I install FLUX now?

It's telling me I need to sign in with a email to download, but I don't remember my email so I made a new one, then the site asked me to confirm the email, but the email says I have to wait 48 hours to send or receive emails.

What do?
>>
>>101790637
A little too realistic in style, she's more elf like.
>>
>d*b* seething again
>>
>>101787538
Very cute!
>>
oh so thats the trick
>>
File: file.jpg (545 KB, 1024x1792)
545 KB
545 KB JPG
>>101790617
more like my employees. my cats are my children
>>101790641
>Lack of gloves in the training data
i got you
https://huggingface.co/datasets/bigdata-pw/fashion-150k
>>
Why is every workflow I get for Flux has 20+ steps in the Basic Scheduler? Shouldn't it work in just 4 as the official site says?
>>
Is it true that "the californian" is unironically pro trump? Brainlet lmao
>>
File: 0.jpg (223 KB, 1024x1024)
223 KB
223 KB JPG
>>
File: BMP_FLUX_05028_.png (1.51 MB, 648x1200)
1.51 MB
1.51 MB PNG
>>101790690
Yeah it's a struggle. I keep either getting flat anime or hyper Disney, can't get one I like much.
>>101790665
>>101788245
Ran and Debo can fight their own battles, don't drag me into it please.
>>
>>101790841
>Ran and Debo can fight their own battles
Well, last time debo rage quit and began to cry bully kek
>>
>>101790680
You want some AI GYATT without being fanum taxed?
HuggingFace is super lame, if you download there, you're have no rizz. Civitai is what's bussin. Be a sigma and get your Flux model, there!
There's lots of skibiddi models in Civitai. The s
>>
>>101790841
D*b* is a coward who vanishes when a faithful discussion starts
>>
File: delux_tru_00032_.png (1006 KB, 1344x768)
1006 KB
1006 KB PNG
>>101790744
dev model is 20+, schnell is the 4 step one
>>
>>101790886
Engaging in this conversation would require self-reflection which Debo lacks
>>
>>101790637
yeah 8gb it's a pain, i will just wait until i get access to 12gb
>>
>>101790841
>Yeah it's a struggle. I keep either getting flat anime or hyper Disney, can't get one I like much.
Flux is really bad at understanding styles, which isn't a surprise since most of them were deleted.
Well, at least their description.
>>
>>101790735
Nice, wish I was a flux dev with the same knowhow and compute.
>>
File: 00011-2024-05-17.jpg (2.42 MB, 2048x2480)
2.42 MB
2.42 MB JPG
>>101790841
He got himself in a lot of trouble due to his retarded behavior and is now basically alone. I don't know what's wrong with him but every time it blows up in his face. You should join the discord btw
>>
>>101790961
>embeddings are about to make a big comeback
>>
>>101790879
wtf
>>
>>101790991
Or bodypart image recognition + inpainting every time for the model.
>>
>>101790915
I see.
>>
>>101790879
Just went to civit and it says I can't go in because they're fixing thing. I'm totally cucked
>>
Schnell vs Dev fp8
Which gens the better results?
>>
File: ComfyUI_00124_.png (1.51 MB, 1024x1024)
1.51 MB
1.51 MB PNG
>>101790961
Try putting in manga descriptions instead of anime.
>>
>>101791024
Now it says they're adjusting and there's a picture of the Nvidia Chinese saying "It just works" but it doesn't just works
>>
my quarterly check if I can use anything half-decent with a 6700xt
>>
File: delux_bu_00039_.png (1.22 MB, 1344x768)
1.22 MB
1.22 MB PNG
>>101791033
dev will almost always be better but there's not a massive difference between the two from what I've seen
>>
>>101785916
Umm, Anon, those who cannot IQ ComfyUI use the other UI's.... kek that simple.
>>
>>101791024
I thought you were capping. It's really down. Well, SD got mogged by Flux so, you must really try it.
Let the Civitai devs cook. Meanwhile, do some mewing.
>>
File: nerd.jpg (232 KB, 1024x1024)
232 KB
232 KB JPG
What's a good system under $5k for running SD? I'm currently using an old Alienware from 2011 that barely runs it with a bunch of warnings about how I don't have the specs to run it.

Never been on this board before.
>>
File: file.jpg (411 KB, 1024x1792)
411 KB
411 KB JPG
>>101790983
send it to their discord or email or something
>>
>>101791142
may as well wait for 50 series gpus in 4 months.
>>
>>101791136
jelqing rn
>>
File: 0.jpg (421 KB, 2048x1024)
421 KB
421 KB JPG
>>
>>101790472
Thanks mate
>>
>>101791136
>>101790879
Why are talking like a retard?
>>
They just released a new version which is really good at prompt comprehension
you really should try out "flux neger" unironically
>>
>>
File: mc_xl_00184_.png (2.21 MB, 1224x2144)
2.21 MB
2.21 MB PNG
>>
File: 00048-3788220452.jpg (670 KB, 1096x1424)
670 KB
670 KB JPG
>>
File: ComfyUI_02656_.png (1.32 MB, 1152x896)
1.32 MB
1.32 MB PNG
>>
>>101791298
In my experience, I get a much better prompt comprehension when I use the fp16 clip model, even when using the fp8 Flux model.
Gen speed is basically the same, compared with the fp8 clip model.
>>
>>101791151
They have all the data.
They just chose to make it more retarded than it should be.
>>
File: ComfyUI_00368_.png (2.49 MB, 1248x1824)
2.49 MB
2.49 MB PNG
>the last thing you see before waking up
>>
>>101791298
He doesn't know how2pc. Telltale sign of a gen alpha.
>>
>>101791298
Where can i download flux neger?
>>
>>101791374
>I get a much better prompt comprehension when I use the fp16 clip model
Can confirm + usually less dyslexia text.
>>
>>101791298
I thought negers had poor reading skills, in general.
>>
File: ComfyUI_02657_.png (1.2 MB, 896x1152)
1.2 MB
1.2 MB PNG
>>
>>101791453
Das wacis' cracka, we's be goods at teh readin' an' sheit
>>
File: ComfyUI_02658_.png (1.32 MB, 896x1152)
1.32 MB
1.32 MB PNG
>>
File: BMP_FLUX_05132_.png (840 KB, 648x1200)
840 KB
840 KB PNG
>>101790879
I think anon just had a stroke
>>
>>101791528
Are you "schizo anon"?
>>
How to best prompt with Flux?

booru tags? book-like verbose descriptions? some other way?
>>
File: file.jpg (285 KB, 1024x1792)
285 KB
285 KB JPG
>>101791395
cope desu. their dataset sucks
>>
File: delux_bu_00040_.png (932 KB, 1344x768)
932 KB
932 KB PNG
>>101791567
>book-like verbose descriptions?
yes
>>
File: 0.jpg (642 KB, 1024x1024)
642 KB
642 KB JPG
>>
File: BMP_FLUX_04026_.png (2.29 MB, 1024x1024)
2.29 MB
2.29 MB PNG
>>101791539
I made a portrait of him the other day
>>
File: mc_xl_00220_.png (2.13 MB, 1224x2144)
2.13 MB
2.13 MB PNG
>>101791519
>>
File: delux_tru_00035_.png (1.32 MB, 1344x768)
1.32 MB
1.32 MB PNG
>>
File: file.jpg (285 KB, 1024x1792)
285 KB
285 KB JPG
doing pitchfork now
8.7 best new dataset
>>
>>101791567
it's trained on a vLLM describing pictures, so write like that, or just use one to write for you
>>
File: de_bo_bp_0092.jpg (715 KB, 1824x1248)
715 KB
715 KB JPG
>>
File: de_re_ap_00100_.png (2.42 MB, 2016x1152)
2.42 MB
2.42 MB PNG
>>
>>101791567
hiya flux, I want to gen a catgirl. got that? and when I say catgirl, I mean a nice, fluffy and cute catgirl. but also sexy, understand? so let's talk about the details. first, the ears. I want them to be triangular-shaped and with long hair. Got that? ok, next, the....
>>101790841
how do you like flux so far?
>>
File: de_iw_bo_00096_.png (2.51 MB, 1344x1728)
2.51 MB
2.51 MB PNG
>>
File: delux_flebo_00003_.jpg (300 KB, 768x768)
300 KB
300 KB JPG
>>
>>101785925
why are you posting this slop

these are all garbo papers
>>
>>101791826
good question that will never elicit a good answer from him
>>
>>101791826
who's forcing you to read any of it? just scroll past it like a big boy
>>
>>101791793
I'm not very good at it but I've seen some amazing stuff from others. I heard the Dev version is where the good anime is at though and I'm on the Schnell. I did download the merge version but haven't tried it, not sure if you need to change the other stuff out for the 16 version or what yet.
>>
Is Debo the same schizo who spams hdg about Nai and groidbots or is that a different one of the troons itt?
>>
>>101791871
every anon you dislike is the same poster
>>
>>101791586
Dataset that precisely lacks many characters (unless you describe them, sometime), people (unless you describe them with luck), styles, and nsfw (nude = underwear, no idea what most poses are) ?
I honestly think it's deliberate.
>>
>>101791867
>>101791891
okay debo
>>
>>101791891
So he is? Ok
>>
>>101791935
>>101791935
>>101791935
>>
>>101791870
nah go straight for the dev version. grab full text encoder and off you go. as in, t5xxl_fp16.safetensors. and dont try and squeeze the full dev model into your gpu (unless you got more than 24gb of vram). can choose one of two quantizations in the dropdown menu in the model loader. when in doubt, here https://comfyanonymous.github.io/ComfyUI_examples/flux/
looking forward to your gens
>>
>>101787488
A lot of roman numeral clocks actually do this for some reason.
>>
>>101791793
I've seen people on civitai using Pony score tags along with "masterpiece," "best quality" and adding (weights:1.4) to their Flux prompts...
>>
>>101792189
Less easy to mix up between IV and VI
>>
>>101791973
Dunno if my 8 VRAM can handle it but might give it a shot, thanks



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.