blessed thread of frenship

the official bigma 2 and lumina waiti

the official pixart bigma and lumina 2 waiting room, also hunyuan finetune

>>101639278
Nice gens

File: Sigma_11753_.jpg (3.61 MB, 2688x1536)

3.61 MB JPG

Another day, another blessed bred

File: Sigma_11757_.jpg (3.29 MB, 2688x1536)

3.29 MB JPG

If you know, you know

>>101639562
is this meant to represent a cervix?

File: Sigma_11765_.jpg (3.29 MB, 2688x1536)

3.29 MB JPG

>>101639637
kek kaleidoscope but good guess?

Can anon harness the power of the 2K model on a shitty card?

File: Sigma_11771_.jpg (1.04 MB, 2688x1536)

1.04 MB JPG

>>101639683
It's slow on a 4090, but maybe if you offload T5 to CPU and use VAE tiling (it might do that automatically after running out of VRAM) if you're under 16GB

is there any model similar to pony that works with ipadapter?

>>101639786
pony doesn't?

>>101639841
only with a very low weight
animagine works but I don't like it

>>101639854
have you tried the mad scientist IPadapter? the additional control might give you better results
>animagine
that was going to be my suggestion kek

File: Sigma_11789_.jpg (3.18 MB, 2688x1536)

3.18 MB JPG

>>101639786
Never used Pony, sorry

>>101639786
>>101639854
derivatives like autism mix maybe

File: MJ6.1.png (56 KB, 1004x393)

56 KB PNG

holy shit this is an insane upgrade. shame what happened to local

File: Sigma_11793_.jpg (3.2 MB, 2688x1536)

3.2 MB JPG

>>101639913
>shame what happened to local
Oh No! Anyway..

>>101639913
bigma will put the mid in midjourney just you wait

>>101640004
It's only 1.3B and held back by the captions LLava 1.6 and Florence can do so don't hold your breath.

>>101640058
i have 2 nostrils so i can hold my breath for both pixart bigma (model being trained by a /lmg/ anon) and pixart bigma (the official successor currently in training) each at the same time

is SD3 still dead

>>101640371
yea

>/ldg/
>/sdg/
why the schism?

>>101639278
Sirs. What is an scheduler and why we have so many of them (Im using invoke)? And why Juggernaut-XL-v9 is raping my PC?

File: Sigma_11832_.jpg (1.91 MB, 2688x1536)

1.91 MB JPG

>>101640456
Welcome to ldg. If you post non-SD gens they get more toxic than usual at sdg. Plus the sketchy discord in their OP, no matter how much people protest. You can post where-ever obviously

>>101640456
they were strangely hostile to the discussion of alternative models to sd, which is odd because i see them allow dalle gens now. that and to distance ourselves from the stable diffusion name so more people realize that local is more than just sd.

File: 1716639610276698.png (1.6 MB, 1024x1024)

1.6 MB PNG

>>101640501
ah okay

File: b4a88907-5346-4809-b904-5(...).png (424 KB, 512x512)

424 KB PNG

What is a CFG scale?

>>101639913
cloud can't stop improving while we're still stuck with SDXL, fuck my life man...

>>101640479
the scheduler says how much of denoise is performed in the early steps compared to the later steps
>>101640559
CFG Scale is how much grounded on your prompt your gen will be, a lower CFG Scale don't let your gen drift away too much from your prompt a high CFG Scale lets the model drift away from your prompt

File: Sigma_11836_.jpg (801 KB, 2688x1536)

801 KB JPG

>>101640479
Diffusion models draw representations of your image by undoing noise addition in steps. The schedulers decide how "far along" each step is.

File: 1708377421910164.png (1.66 MB, 1024x1024)

1.66 MB PNG

this sigma thingy is quite good

File: Sigma_11845_.jpg (2.6 MB, 2688x1536)

2.6 MB JPG

>>101640570
>Local models right now

>>101640643
Sigma_11845_.jpg anon has a recommendation for you

File: Sigma_11850_.jpg (2.27 MB, 2688x1536)

2.27 MB JPG

>>101640643
My favorite new model so far

File: Sigma_11851_.jpg (2.7 MB, 2688x1536)

2.7 MB JPG

>>101640663
I've only posted my model when people asked for it

File: ComfyUI_01226_.png (1.85 MB, 1280x1024)

1.85 MB PNG

Totally not what i had in mind, but still...

Post your best scheduler equation RIGHT NOW

>>101640101
>pixart bigma (model being trained by a /lmg/ anon)
ldg <3 lmg

File: Sigma_11854_.jpg (2.12 MB, 2688x1536)

2.12 MB JPG

>>101640709
>Post your best scheduler equation RIGHT NOW
MFW I use built-in schedulers (beta 0.5 alpha/beta)

>>101640725
>a /lmg/ anon
fug..

File: Sigma_11848_.jpg (2.85 MB, 2688x1536)

2.85 MB JPG

>>101640725
>ldg <3 lmg

sdg is like our aicg

>>101640818
No I'm born and raised from when SD used to be Python command line only.

File: Sigma_11864_.jpg (2.12 MB, 2688x1536)

2.12 MB JPG

>>101640831
256x256 or it didn't happen

IMO the best scheduler is the one that works well with your model, there's no better or worst

https://stable-diffusion-art.com/samplers/

File: file.png (117 KB, 256x256)

117 KB PNG

>>101640858
I've been playing with AI models ever since the early GAN days, I even trained a basic art style transfer one which is why I bought my 3090.

>>101640907
based and imggen pilled

>>101640907
whatever that thing is looks oddly delicious

>>101640783
Beta scheduler doesn't allow you to define denoise tho :/

>>101640942
>a galaxy in a glass bottle sitting on a table

File: Sigma_11872_.jpg (1.79 MB, 2688x1536)

1.79 MB JPG

>>101640907
HYYYYYYYPE

File: Sigma_11874_.jpg (2.23 MB, 2688x1536)

2.23 MB JPG

>>101640907
>>101640950
>HYPE IN AN IMAGE

File: Sigma_11877_.jpg (1.36 MB, 2688x1536)

1.36 MB JPG

>>101640949
>Beta scheduler doesn't allow you to define denoise tho :/
All my gens are pure Sigma, one shot, with no upscale/SD.. so it doesn't matter to me

File: Untitled.png (1.33 MB, 1840x1318)

1.33 MB PNG

I killed civitai myself. About 20 minutes ago I mass downloaded a bunch of style loras. Basically a mini DDOS

i miss meme anon

>>101641595

LORA downloaded 40152/165611. 404 1491
Training data downloaded 5348/27855. 404 21957

du -hs training_data/
431G    training_data/

du -hs models/
6.5T    models/

>a bunch of style loras
rookie numbers

>>101640101
Imagine having 3 nostrils and being able to wait even more

File: Sigma_11887_.jpg (2.26 MB, 2688x1536)

2.26 MB JPG

>>101641595
>>101641720
RIP. Are LoRA's the new pokemon?

>>101641771
>3 nostrils
could be the butthole if you're skilled

loras rape the models weights thodesu

If my images are looking 'overcooked' but the most emphasis I have is a 1.1, what else could it be?

File: IMG_7965.jpg (175 KB, 1024x1024)

175 KB JPG

Unfortunately due to circumstances i an too retarded to learn stable or any diffusion. Oh well, that‘ll do.

File: Sigma_11890_.jpg (2.16 MB, 2688x1536)

2.16 MB JPG

>Bigma

>>101641933
Loras compound, so the more you add the more they cook. Also different keywords have more impact so different "strong" keywords being turned up will cook your image.

>>101641946
brewing

File: Sigma_11891_.jpg (2.93 MB, 2688x1536)

2.93 MB JPG

>>101641943
Because reasons?

File: IMG_7975.jpg (243 KB, 1024x1024)

243 KB JPG

>>101641946
Very nice

File: IMG_7980.jpg (202 KB, 1024x1024)

202 KB JPG

>>101641965
Yeah exactly. A tragedy.

File: Sigma_11895_.jpg (2.86 MB, 2688x1536)

2.86 MB JPG

>>101641961
HYYYYYYYYPE

>>101641968
ty, classy 1girl's

>>101641957
Does cutting down the emphasis of a lora help or does it need to be removed completely?

>>101642001
You find the balance by turning them down.

File: 1721403137238533.jpg (6 KB, 238x212)

6 KB JPG

retard here, why is SDXL or any SD for the reason unable to generate decent backgrounds without the details there being a huge mess? It's able to do decent pictures of a say, chair or table as long as it's the main focus on the picture, but once a chair or table is just some background object and the focus is on something else it gets all fucked up and requires manual fixing with inpainting etc

File: IMG_7926.jpg (157 KB, 1024x1024)

157 KB JPG

>>101642020
Man i love greenhouses

>>101642032
That hair texture is mega realistic

>>101641933
turn down cfg

File: Sigma_11906_.jpg (2.68 MB, 2688x1536)

2.68 MB JPG

>>101642029
The VAE is a lossy compressor and it eats up background faces/text at 4 channels. SD3 added 16ch to overcome this but RIP anything but text out of that thing.

>>101642040
They rock

>>101642053
ty, 2k model picks up more details

>>101642029
To add to >>101642075 latent space is a compressed version of pixel space therefore tiny details will become a mess unless upscaled and thrown through another sampler.

>>101642132
this rocks

File: Sigma_11919_.jpg (2.5 MB, 2688x1536)

2.5 MB JPG

>>101642207
kek

File: 00_07.jpg (358 KB, 1552x1200)

358 KB JPG

>>101642132
Amazing

>>101641946
im astounded at how good the 2k version looks

File: ComfyUI_01388_.png (1.75 MB, 1280x1024)

1.75 MB PNG

pony and oil painting lora rocks

File: Sigma_11922_.jpg (2.24 MB, 2688x1536)

2.24 MB JPG

>>101642309
ty, you got a nice set you're working with yourself

>>101642323
ty. I want more 2k models on civit!

Good SDXL models for softcore NSFW? I'm training a realistic concept model (girls peeing, yes I know I'm weird) and I'm trying to find the best base model / mix to train on. I have ~2000 well tagged images. The model needs to be able to do lewd, know what a pussy looks like, good at poses / anatomy, and ideally be pretty flexible / not overfit. Some things I've tried or will try soon:

Base SDXL: nope, mediocre anatomy, and lack of pretrained NSFW makes it shit for fine tuning on my dataset.
Juggernaut: not good enough at NSFW. Seems weirdly overfitted too. Training on this gives terrible results.
Halcyon: similar to juggernaut in limitations, but the 1girls look really good IMO.
Iniverse: pretty good actually.
bigASP: extremely flexible base model, probably best results of all I've tried. A bit difficult to wrangle, need to play with quality tags a lot.
Aphodite: haven't trained on it yet but it does softcore NSFW way better than most of the popular realistic models.
Anteros XXXL: haven't even downloaded it yet but looks promising. It's just SDXL trained on 1M+ photos (similar to bigASP in that regard) so might work really well.
Literally all hardcore models: way too fucking overfit on hardcore porn. Anteros might be the same way actually, we'll see.

Anything I've completely missed that meets my requirements?

>>101642389
i think you'll have better luck asking this in the >>>/h/, >>>/d/ or >>>/aco/ generals. i think theres an /aco/ general dedicated to realistic gens

>>101642389
Huh, why no Pony? Pony is for anatomy.

>>101642424
ok thanks I was not aware /aco/ had a realism thread.

>>101642446
Forgot to mention pony, I've tried several realism pony models and none are realistic enough. There's way too big of a "style shift" between what the model outputs by default and my dataset. Training results are always bad, like the model has work too hard to change the overall style during training.

>>101639913
>Things should look "generally more beautiful" across the board
hah

>>101642335
pretty cool gen, great background

File: 801.png (1.86 MB, 1456x816)

1.86 MB PNG

>>101640004
the eternal local wait grows ever longer

>>101642748
Badass. Sigma?

File: Sigma_11937_.jpg (2.73 MB, 2688x1536)

2.73 MB JPG

>>101642748
Nice

>>101642759
i think that's midjourney

eternally coping

File: Sigma_11938_.jpg (2.95 MB, 2688x1536)

2.95 MB JPG

Does midjourney even go above 1 megapixel?

File: ComfyUI_01468_.png (1.68 MB, 1280x1024)

1.68 MB PNG

Finally i'm starting to get results on arm and hand poses,with careful phrasing.

>>101642825
Local is always going to be limited compared to models which require >A100s to run. But what local has is uncensored.

File: ComfyUI_01466_.png (1.74 MB, 1280x1024)

1.74 MB PNG

Coherent results, but a pain in the ass.Like describing a 5 year old details of a photo.
They should work on more natural prompt comprehension by ai.
Btw i've tried auraflow,i have good hopes about that

>>101642892
>But what local has is uncensored.
Not only uncensored outputs, but also the ability to do /whatever/ you want with it. I don't care for services which only allow the user to define positive and negative prompts, no matter the additional bells and whistles they add (like -sref or whatever).

>>101642937
i remember the pony guy saying he's going to use natural language captions for the next model

File: Sigma_11955_.jpg (2.16 MB, 2688x1536)

2.16 MB JPG

>>101642892
>Local is always going to be limited
When you think about it, what you can gen with local is unlimited. Whereas SaaS is limited to what your ruler decides.

>>101642937
models trained on booru tags work best with booru tags innit

>>101642759
>>101642792
midjourney 6.1 lulz

File: Sigma_11966_.jpg (3.22 MB, 2688x1536)

3.22 MB JPG

>>101643195
It really do be like that sometimes

>>101643331
Your daddy gave you a good gen

>>101639718
my 30 series is crying

amgib

>>101641614
Where did he go?

File: Sigma_11977_.jpg (3.53 MB, 2688x1536)

3.53 MB JPG

>>101643487
Sorry to hear that. Does it work at all? How much VRAM?

best gens

File: Sigma_12013_.jpg (2.22 MB, 2688x1536)

2.22 MB JPG

>>101645185

File: Sigma_12020_.jpg (1.97 MB, 2688x1536)

1.97 MB JPG

>>101645582

File: 7742d7c2-97ef-49a5-9e25-c(...).png (1.46 MB, 1024x1024)

1.46 MB PNG

>>101639278

File: Sigma_12032_.jpg (2.16 MB, 2688x1536)

2.16 MB JPG

>>101645185

File: Sigma_12034_.jpg (2.51 MB, 2688x1536)

2.51 MB JPG

>>101645303

File: Sigma_12042_.jpg (3.26 MB, 2688x1536)

3.26 MB JPG

>>101645303
>>101645966

>>101644717
>>101645265
>>101645619
>>101645678
>>101645966
>>101645988
>>101646053
so good

File: Sigma_12047_.jpg (2.36 MB, 2688x1536)

2.36 MB JPG

>>101646306
ty

>>101646459
great

>>101646584
ty, a little cliche summoning banksy from the latent realm

>>101646629
nice

File: Sigma_12060_.jpg (2.16 MB, 2688x1536)

2.16 MB JPG

>>101646686 (You)
With gen

>>101640907
>>101640950
>>a galaxy in a glass bottle sitting on a table
Nailed it.

File: ComfyUI_temp_ktrrr_00025_.png (1.19 MB, 1152x832)

1.19 MB PNG

>>101646686
thanks, need to figure out (again) making ultimateupscale work - those 2k gens are really nice
>>101646697
prompt?

File: Sigma_12065_.jpg (3 MB, 2688x1536)

3 MB JPG

>>101646714
ty, love this gen!

>>101646697 (You)
>prompt?
Circular Eclipse: Concentric circles that gradually shift from full to crescent shapes, balancing positive and negative space.

https://lite.framacalc.org/4ttgzvd0rx-a6jf
what's your favourite 3 or 4 letter gibberish art style word for Pony? and why is this page formatting so hard to read holy shit

i guess people just ignore all of these and purely use style loras, don't they? how do lora styles compare with the built-in art styles?

>>101646758
Can't believe anons are still mining that model

>>101646770
are they? i can't even tell if the page has been updated, i just had it bookmarked

File: Sigma_12068_.jpg (1.84 MB, 2688x1536)

1.84 MB JPG

Good night anon

>>101646782
I hope not. Something needs to replace it already.

File: ComfyUI_temp_ktrrr_00035_.png (954 KB, 1152x832)

954 KB PNG

>...(abstract photography:1.5), ...

I use SD but I don't know what Sigma is. Can I run it on my 8GB GTX 1070 Ti?

>>101646975
Maybe, but you might have to use SD3s clip instead of T5. In OP.
>>101639278
https://github.com/city96/ComfyUI_ExtraModels

>Model Ranking
How accurate is that? And I see some cool sigmaballs gens, are the requirements different from SD? I got 8gb VRAM.

>>101647008
Fuck, I asked just as you replied to that anon. How much should I have for this T5 nonsense? I saw some files and they were bigger than the models themselves I think.

>>101646975
Yes, but not worth the effort. View it like one of those Youtube videos about can I run X on a raspberry pi and then celebrating their superior 8 fps.

You have to do annoying things to get T5 working.

>>101647063
use https://civitai.com/models/497255?modelVersionId=568405 instead of T5

>>101647099
thank you

File: Untitled Canvas ship.png (40 KB, 1728x842)

40 KB PNG

Hey I did this quick sketch and I'd like to see what Fooocus does with as a realist film sci fi model. What do you guys suggest.

>>101647228
Some cinematic model

File: 2024-07-19_21-43-24_8197.jpg (245 KB, 1408x704)

245 KB JPG

>>101647306
You mean write that in the prompt? Also my Fooocus isn't working, its saying not enough CUDA or some such.

Image is recreation I did several days ago when Fooocus was working.

>>101647342
yeah just write it in the prompt

>>101647228
https://civitai.com/models/359999/cinematicredmond-cinematic-model-for-sd-xl

>>101647401
>https://civitai.com/models/359999/cinematicredmond-cinematic-model-for-sd-xl


Is this a sort of plug in for Foocus that I can use with that program/ I'm really new with all this.

>>101647498
if you cant run that run this https://civitai.com/models/550737?modelVersionId=612841
it does more than 1girl

File: MJ_V6.1.jpg (120 KB, 888x592)

120 KB JPG

>>101639913
Doesn't look like anything special. You can literally do this with Sigma, it's like 6.1 is a preliminary bigma model.

File: EVkbbJzXgAU9IWq.jpg (137 KB, 819x536)

137 KB JPG

>>101639278
>txt2img: https://www.mage.space
>img2img: https://huggingface.co/spaces/huggingface/diffuse-the-rest
>sd3: https://huggingface.co/spaces/stabilityai/stable-diffusion-3-medium

these all junk. Level of 2019


Anything better on par with midjourney but still web interface and free?

>>101647739
I wonder if anyone here spends time looking at the images this model generates. Even 6.1 is still about a year behind Dalle. Every model is. Look at what is posted on the Dalle thread then come back to this image.

>>101647739
what's missing on MJ is a better prompt understanding... and the fact we have to go to discord to gen, that's fucking dumb

File: ComfyUI_temp_ktrrr_00102_.png (1.08 MB, 1088x960)

1.08 MB PNG

>>101647761
>these all junk.
the only online generators i use are the ones set up for new models when they release. you are correct those links are old.

Been away from imagegen for a loooong while (1-2y), what's the absolute best quality model we can run locally? Still SD or is there better stuff out now (or coming soon)?

>>101647789
That's not the only thing that's missing. Proper privacy unless you wanna pay the Jew $720 per year.

File: 4654565465621.png (85 KB, 1092x716)

85 KB PNG

>>101647864
Forgot pic

>>101647864
>Proper privacy unless
There's 0 privacy when you use an API, that's why I'm all for local yeah

>>101647739
>that shit tier gen
>MJ_V6.1
You're trolling

>>101647878
API is not too bad though, you may not be inclined for the entire world to see your prompt and steal it.

File: 4565644561.png (780 KB, 1571x850)

780 KB PNG

>>101647880
>You're trolling

I'm not kidding you go to their showcase. The filename confirms it's 6.1 I just renamed it. You can gen way better images with a XL/Pixart/Hunyuan finetune, nothing beats a local combo and complete control over style.

File: MJ_V6.1_2.jpg (211 KB, 888x592)

211 KB JPG

>>101647914

>>101647914
Seems to me they are using a completely different architecture/model, but it's a failure and calling it anything other than beta was a retarded move or MJ really looks worse than I thought.

>>101647914
>>101647966
>>101647977
yeah, something's wrong with those pictures, it looks like a random SDXL finetune now, they'll probably revert back to v6 that's for sure

>>101648005
>>101647977
I think it's the SD3 architecture. This is very slow for them and they should've caught up to Dalle by now (which they didn't). They might be completely overshadowed by Bigma. Obviously if they are set back a year because of Emad then Chinks have won.

>"sitting on a couch"
>they squat in front of it
I hate computers.

>>101646787
seeya

>>101646787
gn anon. your paper gens are my favorite

>>101647880
>>101647966
>>101648005

>>101642748

>>101642516
Zonkey is a great realistic Pony model

cough

>>101647569
What is the use case for more than 1girl?

File: 8989444566.png (556 KB, 1461x812)

556 KB PNG

>>101648304

File: 5849489456.png (362 KB, 676x729)

362 KB PNG

>>101649220

File: wdv-test-v0.1_00002_.jpg (1.16 MB, 1264x1848)

1.16 MB JPG

What a piece of shit

Spent 2 weeks taking photos of art pieces and now I should start training it. Just need to decide base model. Seems there's nothing new about next PixArt

>>101650622
waifu model? lul

>>101650691
It's far from ready. Reminds me of that Sigma Anime v0.1

>>101650691
Eternal bigma waiting room. And that one anon training base model

>>101650622
is this some sigma anime model you're training? what is that?

>>101651033
https://huggingface.co/waifu-diffusion/wdv-tests

>>101651141
>waifu-diffusion
oh.. i remember now. yikes.

File: 00_02.jpg (310 KB, 1552x1200)

310 KB JPG

Morning anons

>>101651428
morning, awesome gen. catbox?

File: 00_01.jpg (308 KB, 1552x1200)

308 KB JPG

>>101651473
I'm not at my computer right now but it was generally like this

hyperrealistic, photograph, a vintage television on a podium in the middle of a tropical rainforest, glass wedding arch, color gradient sky,  night time, Halftone pattern, dual color image, vintage, print pamplet illustration, snowy mountains in background, full moon, paper mache collage, detaied, paper cutouts

File: 1713930494799654.jpg (230 KB, 1000x1000)

230 KB JPG

is there an image editor that works well?
I wanna just remove her fascinator in this pic and make the eye symmetrical to her left one

File: Sigma_12069_.jpg (1.74 MB, 2688x1536)

1.74 MB JPG

>>101647892
>entire world to see your prompt and steal it
kek

>>101648084
gm

>>101648131
gm. Also amazed at how well it handles the idea

>>101651428
gm

File: Sigma_12075_.jpg (2.22 MB, 1536x2688)

2.22 MB JPG

>>101651698
There's lots of image editors that "work well" but many have huge learning curves. You can mess with inpainting or photopea.com is free online

File: Sigma_12077_.jpg (2.34 MB, 1536x2688)

2.34 MB JPG

Busy day

Hello, can the pixart generate obscene breasts yet? (bigger than head). Thank you in advance.

File: Sigma_12063_.jpg (3.09 MB, 2688x1536)

3.09 MB JPG

>>101652152
Probably with this:
https://civitai.com/models/460998/vintageknockers-pixart-sigma?modelVersionId=517752
or this
https://civitai.com/models/522164/twistedreality?modelVersionId=652364

That guy is a boob maxxer

>>101651537
how does sigma perform with this prompt?

File: Sigma_12070_.jpg (2.79 MB, 2688x1536)

2.79 MB JPG

>>101652471
Moons mostly look like hot garbage in Sigma. It's similar, but I think worse. Really need to add moon photos to the dataset

>copy someone's exact model and tags to see how this shit works
>mine comes out so wildly different its not even recognizeable to the original
Alright, cool, I understand how SD works even less now

>>101652712
Resolution, seed, sampler and the GPU/UI you use also matters.

File: Sigma_12081_.jpg (1.67 MB, 2688x1536)

1.67 MB JPG

>>101652712

>>101652750
All of that was matched aside from the GPU and UI so rip

>>101652766
Why do you want an exact copy, so dumb.

>>101652752 (You)
>>101652712
lul forgot text. Only on a few hours sleep. You need same seed/cfg/sampler/etc. One sampler produces wildly different visuals than another

>>101652779
It was just a test to see if I could at least get remotely similar results since I clearly dont know what im doing otherwise.

>>101652806
You type in words and it makes images similar to the words you typed in. It's not complicated.

File: 00000-3887962902.png (390 KB, 512x512)

390 KB PNG

>>101652833
prease understandu
that when my result looks like me loading my ass with paint and projectile shitting onto a wall, that i'd want to copy another set of prompts and their settings exactly and then work my way down from there
and my confusion when it doesn't look "similar" whatsoever

>>101652874
What model are you using?

>>101652900
I just pulled the data out of this image since it looked standard enough >>101645846
with the automatic1111 ui (amd)

>>101652948
Does that image look 512x512?

File: 00006-2209909268.png (2.62 MB, 1024x1536)

2.62 MB PNG

>>101652957
the 512 image was just me trying to figure it out on my own, hence seeking out other images to copy off of to see where I went wrong
this one is the copy off that other image's tags

>>101653006
Write the positive and negative prompt here. Which model are you using?

>>101653029
it's already in the image I linked

>>101653035
Okay I don't give a fuck any more, thanks. You're too stupid for this.

>>101653042
Thank you for filtering yourself out of my replies when I already provided the link.

>>101653067
>why won't anybody help me I expect people who help me to spend more effort than me, I already have the information I can copy and paste but fuck anyone trying to help me, they should waste their time
Anon, your problem is caused by your laziness and stupidity. You are lazy and stupid which makes you incapable of following instructions. I am likely the only person to have helped you. Have a good day.

>>101653109
Don't worry i'll be back later pretending to be someone else asking for help and you'll never know.

>>101653119
Anon, you missed my point. You're lazy and fucking stupid. You will ask lazy and fucking stupid questions. You will be a faggot about it. No one will help you because you come off as lazy and stupid and anyone who tries to help will realize you're actually 80 IQ and lazy.

>>101653119
NTA but kekd

>>101652111
very busy day indeed

File: Sigma_12095_.jpg (1.52 MB, 2688x1536)

1.52 MB JPG

>>101653119
lul

>>101653181
Lots going on and lots to do

lmgigma status?

>>101653445
bigma anon do you frequent lmg?

>>101653567
I only go to check the news. Sadly training and doing captions uses my computer so I really don't do anything outside of watch Tensorboard.

>>101650622
I don't get how this guy hasn't given up yet, his finetunes were always total ass since the beginning in 2022 kek

>>101652780
Dude's probably got no negatives, copying a bunch of lora tags, and running with only 5 steps.

bigma my beloved

>>101654003
Png extract copies steps, dimensions, upscaling, and negatives too

>>101654316
have you considered he's incompetent and stupid?

File: Sigma_12100_.jpg (2.38 MB, 2688x1536)

2.38 MB JPG

>>101653605
Actually same except finetuning language models is fun af too. lmg is the best for breaking model news
>Sadly training and doing captions uses my computer
local anon curse


>>101654003
Maybe. He hasn't said much about the setup. Maybe he thinks the metadata comes through 4chan and doesn't know about catbox.moe in advance.

>>101652607
Other than the moon it looks really good. Do you happen to have any direct comparisons (same settings) of 2K vs 1024px versions?

>>101654348
Yes, because it's still me you're replying to.
ha ha gottem

>>101654409
I linked to the catbox in question already stating that's where I pulled it from.

File: ComfyUI_00821_.png (1.99 MB, 1080x1344)

1.99 MB PNG

>>101637628
The newbie again.

>>101653605
Post sexy tensorboard pics

File: Sigma_12105_.jpg (3.04 MB, 2688x1536)

3.04 MB JPG

>>101654809
Okay it was hidden behind a reply https://files.catbox.moe/r3dcen.png

And you run this on ComfyUI and can't get the same gen.. AMD?

File: Sigma_12117_.png (1.75 MB, 1344x768)

1.75 MB PNG

>>101654642
bunline1024 0.9, the other is 2k v0.4 in progress

>>101655200
AMD correct, but automatic1111 which is likely the bigger issue. Like I said I have no idea what i'm doing and came into this with a minimal understanding of how this all works and understand it even less now. But I guess that explains why theres so many nearly identical looking UI forks.

File: Sigma_12115_.jpg (3.36 MB, 2688x1536)

3.36 MB JPG

>>101655302
https://github.com/comfyanonymous/ComfyUI

But I'm nvidia and don't know how it does on AMD. You can drag the catbox image onto comfy and the workflow populates. Clean workflow too

File: Screenshot_20240731_133353.png (200 KB, 1546x723)

200 KB PNG

>>101655445
>>101655302
Picrel is what your referemced catbox looks like loaded on my comfy

>>101655445
>>101655464
I'll give it a shot. I'm not really attached to that catbox image in particular it was just one of the first links I ctrl-f'd in a thread to try as a baseline. I'm mostly just here whining that my understanding of the system wasn't quite what I thought it was, or that the output could differ so wildly from the smallest things.

Spent far too long fucking around with python and venv on debian for automatic1111 so hopefully this wont take quite as long.

>>101655302
>automatic1111 which is likely the bigger issue
nta but i believe you're right, automatic1111 and comfyui have different backends so they should generate different looking images even with the same seed. someone correct me if i'm wrong.

File: Sigma_12112_.jpg (2.72 MB, 2688x1536)

2.72 MB JPG

>>101655514
Bigger learning curve but more control

File: ComfyUI_00835_.png (3.02 MB, 1080x1920)

3.02 MB PNG

>>101655102

Here it is, warm fresh bread...
>>101655488
>>101655488
>>101655488

File: Sigma_12102_.jpg (2.96 MB, 2688x1536)

2.96 MB JPG

>>101655102
>>101655592
Hey, welcome back. Nice gens!

>>101656617
no way that's a gen, did you just pull out a random image to fill the thread? :P

>>101656617
>>101656755
https://desuarchive.org/g/thread/100435445/#q100435445
hmmm