File: ComfyUI_03778_.png (1.71 MB, 1216x832)

1.71 MB PNG

churchillposting was too powerful

File: flux1-dev-fp8_01085_.png (1.19 MB, 1024x1024)

1.19 MB PNG

>>101984382

daily reminder no one here has a clue how to achieve this:

https://www.virtualstagingai.app/

or even furniture an empty room from scratch whole keeping the same perspective

go no, I'll wait

>>101984401
>whole
*while

>go no
*go on

File: ComfyUI_temp_mkpud_00063_.png (3.09 MB, 1360x1600)

3.09 MB PNG

>>101984401
the real holy grail of AI is consistency, no one cares about that AGI crap

flux behaves A LOT different than SD
and I'm not going to tell you how

>>101984418
oh god my coomer instincts are kicking in, flux healed me. i was out

>>101984418
No one cares about celebrity slop either yet here you are

>>101984418
the fuck are you talking about, retard
that was solved years ago, not to mention there's endless ways around it

what I'm asking is a much greater challenge

>>101984438
the furniture thing? i googled it and i found
https://www.virtualstagingai.app/, it might help

>>101984390

norton front plate missing, are they going to fine you

does anyone use the CLIPTextEncodeFlux node? The clip_l prompt is meant to be tags and the t5 prompt is meant to be natural language. Do you think it makes a big difference vs the conventional way? I haven't noticed much of an improvement.

File: 00063-1191024999_cleanup.png (2.97 MB, 1280x1920)

2.97 MB PNG

>>101984401
Was gonna respond but thread got archived.

Since you have been crying for the entirety of last thread I'll take a shot. I have some experience working with facial detection software so this is the approach I would take.

Playing around with the tool seems like there's a few things that are done:
>detect empty, open floor areas in a room
>distort furnitures based on center point of room
>place furnitures in room based on outline of floor, maybe?
>place paintings on wall area and fans on ceilings after detecting areas based on outlines

I'm 99% certain they use preexisting images of furniture and distort them slightly based on the centerpoint of the room. The furnitures are neatly tilted and angled in a way that points to the center of the room from what I'm seeing. If you're savvy enough I think you can develop some sort of detection software to find the edges/outlines of the room and find the open area in a floor/wall/ceiling and put furniture or whatever in accordingly. I can't get into the nitty gritty details of the software since that's out of my wheelhouse.

>>101984446
the don't even have decent social media content, looks like a pajeet scammer site but it works perfectly

how THE FUCK can they have such an impressive results?

>Developed at Harvard Innovation Lab
sounds legit, but still

File: 1712985453754518.png (1.15 MB, 1024x1024)

1.15 MB PNG

check em

>>101984463
>>101984471
i'd say it's a model finetuned on furniture and segmentation/a series of masks

>>101984418
are you using that playboy lora?

File: ComfyUI_00035_.png (2.04 MB, 1024x1024)

2.04 MB PNG

>>101984382
>flux1-dev-bnb-nf4
pluses and minuses, vs full dev?

File: ComfyUI_temp_mkpud_00067_.png (2.75 MB, 1360x1600)

2.75 MB PNG

>>101984438
there is no real consistency in AI moron, thats why hollywood/ai companies struggle with video generation and studios keep using deepfakes, you can change a little character or in your prompt/workflow and you will get a different result

>>101984463
this coomer is at least trying

honestly, I don't even care about the segmentation, which is impressive, but who gives a shit

the real question is how are they inpainting the furniture with that level of accuracy, matching perspective, and without distorting (too much) the rest of the room

>I'm 99% certain they use preexisting images of furniture and distort them slightly based on the centerpoint of the room.
thought so at first, but clearly not, if you try it out all the results are different

>If you're savvy enough I think you can develop some sort of detection software to find the edges/outlines of the room and find the open area in a floor/wall/ceiling and put furniture or whatever in accordingly.
i srsly doubt some shitty furniture website developed some epic AI model to achieve this
this has to be some kind of shitty comfyworkflow hooked to some furniture loras, it can't be anything else

File: ComfyUI_00893_.png (1001 KB, 1280x720)

1001 KB PNG

>>101984505
Nerfed down so it can run on less vram

>>101984482
smartest /g/ user

>>101984511
i srsly hope you're trolling
altho celebrity coomers are unironically the lowestIQ on this site

More celeb
Preferably teenage and in compromising positions
Thx

>>101984534
and they call me a pedo, kek

>>101984401
here's how it may work.

train ai to guess the depth map

train ai to produce images that suit depth maps

File: 1712113183622057.png (241 KB, 959x292)

241 KB PNG

might as well ask in the new bread
any tips/workflows for outpainting with Flux?
specifically outpainting portraits > full body shots if that matters
searching doesn't really yield any useful results, getting it to stick to the source material is a real pain
denoising makes the grey padding affect the outpaint and controlnets like canny (leaving the padded area black) with denoise at 1.0 are a partial cope but still shit as they don't hold much relevance to the source material (picrel)
any advice would be appreciated

>>101984532
actually it might be segmentation/a series of masks and a model finetuned on furniture

More anime
Less 3DPD
Anime website

thank you for the bumps kind stranger ;)

>>101984529
Thanks, I'll skip it.

File: 1714088842965636.png (1.14 MB, 1024x1024)

1.14 MB PNG

>there will never be open source dall-e
>it will be restrictive
>there will be no loras
>there will be no finetunes
>it will cost money to generate because an API is needed
YOU ARE HERE

>>101984534
This, but preteen. Preferably a fetus.

why is this guy swallowing elons cum in public?

File: Flux_00886_.png (867 KB, 1024x1024)

867 KB PNG

this flux fine tune is so good fuck

>>101984545
bro, the site isn't new, the model is clearly 1.5 or SDXL, these retards didn't train shit, there must be a way to:

1. match room perspective
2. detect windows and door frames
3. inpaint with high denoise while considering 1 and 2

>>101984582
could be a a series of masks or segmentation combined with a model finetuned on furniture

>>101984579
>pic unrelated

>>101984582
Oh man, they could call that an AuthoryWeb or something.

>>101984582
not to fucking mention, it renders THREE images in 1 minute, and they're ALL perfect

any advice for prompts?

>dude

dude yourself faggot

File: 1696968333238555.png (1.06 MB, 1024x1024)

1.06 MB PNG

>>101984573
>mocking a politician with AI = blowing them
libtard

>>101984592
i seriously doubt there's some kind of openAI level geniuses behind that site, desu

File: 00060-289294753.png (2.87 MB, 1280x1920)

2.87 MB PNG

>>101984482
Can they do that with a pre-existing image though? The model would somehow have to detect the angles of the room and angle the furnitures accordingly and detect where the walls, floors and ceilings are.

>>101984516
>inpainting the furniture
>thought so at first, but clearly not, if you try it out all the results are different
I don't think they are inpainting the furniture, I'm nearly certain they are using pre-existing images after taking a closer look at them. I think they have a large dataset of individual furniture photos for floor/ceiling/walls that are in many different angles. I suspect the images are tagged with what degree they are angled in and for where to use the furniture. It would be extremely difficuilt to inpaint images that are angled as neatly as they've done.

>i srsly doubt some shitty furniture website developed some epic AI model to achieve this
It's also possible they just slapped the name AI on there when in fact there is very little to no AI actually being used. There are quite a few companies out there that use the AI branding for hype and branding sake.

This is all my speculation though and I would be entirely wrong.

File: 00699_.png (1.94 MB, 960x1280)

1.94 MB PNG

>Help step-anon, I'm stuck!

>>101984582
Wouldn't ipadapter with segmentation do exactly what you're describing? Sometimes it feels like you guys forget half of these tools exist

>>101984647
>openAI
>geniuses
they have a literal whore whos only credit is "wife of famous person" on the board.

>>101984653
>It would be extremely difficuilt to inpaint images that are angled as neatly as they've done.
EXACTLY

don't think they are inpainting the furniture, I'm nearly certain they are using pre-existing images
impossible, because the results match any room you throw at it, respecting windows and door frames, it's way too organic to just be some kind of system that overlays .png furniture sets combined with SD

>It's also possible they just slapped the name AI on there when in fact there is very little to no AI actually being used.
nah, it's clearly AI, it's 100% free can try as many as you want
you can post any empty room and the fucking thing works every single time
I'm getting "decent" results with flux, but gening 3 PERFECT results in 1 minute is fucking insane

>>101984656
sometimes I wonder if you retards can read
segmentation is not the issue at all
the main issue is inpainting ROBUST furniture sets PERFECTLY why matching the exact perspective

>>101984733
>why
*while

File: 1713766828101984.png (41 KB, 563x283)

41 KB PNG

Still a little stuck on running flux right now. Updated ComfyUI + Python dependencies (there was some updates, so I got my hopes up), but I still get errors.

>flux-dev-fp8
Can run without problems, but it melts my computer.
>flux1-dev-bnb-nf4-v2
I get a long chain of
>size mismatch for img_in.weight: copying a param with shape torch.Size([98304, 1]) from checkpoint, the shape in current model is torch.Size([3072, 64]).
Numbers inside Size() change, and so does the x.weight. ...I really wish the update suggestion worked, I really wanna use this.

File: ComfyUI_32299_.png (1.02 MB, 1280x720)

1.02 MB PNG

>>101984654
The help is coming

>>101984747
try q4 or q8 with the gguf node

>>101984733
Ok, that site is ridiculous.
They could use that to compose just about anything. Why are they only using it for furniture? LOL

>>101984733
So? I bet it can't generate the same room and exact furniture with different perspectives (aka consistency)

>>101984801
SEE
now you know why i've been sperging for 4 hrs lmao
WHAT FUCKING SORCERY IS THIS

File: 1.jpg (68 KB, 1000x667)

68 KB JPG

>>101984680

>>101984809
why would you even want that?
talk about missing the point

>>101984733
you're apparently the one not reading the words ipadapter, have you just never used controlnet before...? all of the things you're describing are very doable and have been for some time

>>101984821
lmao, you fucking imbecile

File: Untitled.jpg (1.32 MB, 3201x1067)

1.32 MB JPG

>>101984733
Yeah, im not exactly impressed. Took a messy room from google images and on a not-controlled input it catastrophically failed in 2/3 gens and even the 1st has all kinds of issues.

Not sure why everyone is hyping it up. I'd be impressed if it could take a non-perfect input, but this isn't even that unreasonable of one.

File: GGUF_T5_Q5_K_M.png (1.19 MB, 1024x1024)

1.19 MB PNG

There's a 50% chance this'll actually work first try.
https://github.com/city96/ComfyUI-GGUF/pull/49
https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf

DATSH I-PEEE ADOOPTURD AN CUNTROOL NETHS

lmao

jesus fucking fuck

>>101984840
OH GOD YEEEEEEEEEEEEEEEEEEEESSSSS

>>101984840
OH SHIT, LET'S GOOO

CITY IS A MANIAC

>>101984401
Oh cool with this I can img2img furnitured rooms into my doujin and I don’t have to paste em in anymore
Ty handy link

>>101984838
room supposed to be empty....
why would you even do that
are you even real humans?

File: 00008-3773149587.png (3.1 MB, 1280x1920)

3.1 MB PNG

>>101984707
Oh, I just read this on their website.

>Our AI virtual staging doesn't work with a predefined furniture library. It analyzes every room and based on the style and room types generates furniture that best matches the room. Therefore no two rooms will ever get the same furniture.

I still think they use pre-existing images. Also, it appears they can remove and change furnitures as well.

https://www.virtualstagingai.app/virtual-staging-ai

Maybe they do use inpainting of some kind. I don't know.

>>101984853
only living rooms unless you pay, tho

>>101984582
depth maps.

File: file.png (49 KB, 204x248)

49 KB PNG

>>101984855
if it can't do what it says it can do on the website then it's trash. why are you sperging out over me testing something they've advertised is possible?

>>101984401
It says on the website it uses 3d rendering

>>101984869
Oh, also they are also 100% using depth maps. Bet my life on it.

File: delux_ci_00071_.png (1.98 MB, 1536x968)

1.98 MB PNG

>>101984733
I'm guessing this guy got hired/contracted to make an "AI powered" app for a discount furniture store and has no idea how to make it. now he's trying to bait anons into showing him how to do it, lol

File: file.png (619 KB, 1127x682)

619 KB PNG

>>101984875
and the mask works, it just doesn't generate very well with anything but ideal circumstances. so again, i don't know what the big deal is when it clearly sucks at non-optimal inputs

>>101984840
customary "which should I download?" post

>>101984828
sorry you're confused by it? it's really not that mind blowing, try controlnet properly sometime.

File: file.png (515 KB, 882x512)

515 KB PNG

>>101984896
>>101984855

the "cleaned" photo just to truly hammer home that it just doesn't generate well without optimal inputs.

do you understand now? it's bad. it sucks.

>>101984869
>I still think they use pre-existing images.
that was my first thought too, but then it wouldn't nail weird angles as good as it does

>>101984875
>autism

>>101984879
so it matches the perspective, generates furniture 3d models matching the perspective, and then inpaints the textures with SD?

unironically, that seems to be what is doing, but how the fuck do you even achieve such thing?

File: 1708051286628065.png (105 KB, 1359x1011)

105 KB PNG

What am I doing wrong?
Trying to use Advanced ControlNets with Flux, error happens on other samplers such as KSampler too.

File: 1697485993512516.png (1.06 MB, 1024x1024)

1.06 MB PNG

ps1 style miku

>>101984916

shame that city's incredible feature (https://github.com/city96/ComfyUI-GGUF/pull/49) just got completely overshadowed by this retard having an autistic meltdown over a subpar implementation of inpainting

>>101984896
>>101984913
I don't give a shit about removing furniture, that's not an issue for anyone ever, the main requirement is adding furniture to empty places

>>101984820
>why would you even want that?

Because if I wanna see it from different perspectives, it would give me a different room

Also you're making a big deal about a nothing website, it can be done in many ways, replace items in a image can be done with any detection model even one trained by you (i.e. there are yolo models that detects nipples, i'm pretty sure there are ground models that detect items in a type of image that you can setup easily)

Adding items to an empty room its pretty simple also, you can add a controlnet image with a mask or generate a different gen and remove its background with new tools such as layerdiffuse

Hell, they may be using pre-rendered images as assets too, something that has been done before

Imagine getting excited at cherry picked results and a controller environment (aka pictures of rooms)

>>101984873
reminder that retard is sperging out because he doesn't understand controlnet lol

File: ComfyUI_13428_.png (3.68 MB, 2560x1536)

3.68 MB PNG

>>101984905
Q5_K_M worked fine for me, Q8_0 is the safe bet I guess.

File: ComfyUI_00904_.png (946 KB, 1280x720)

946 KB PNG

>>101984919
You have to update comfyui

>>101984930
Yes, and what aren't you getting? It doesn't work with a non-optimal input, even of an empty room.

>>101984838
>>101984838
>>101984838

File: ComfyUI_32308_.png (1.22 MB, 1280x720)

1.22 MB PNG

AI-generated jokes suck.

>>101984929
>text encoder
do you, literal subhuman retards, know that, you can actually, add any text you want with photoshop?

>>101984838
Empty house is the use case tho.

>>101984916
>but how the fuck do you even achieve such thing?
this really isn't that complex anon, you could probably have chatgpt explain how to set it for you if you asked nicely..

>>101984949
that's not what a text encoder means retard

>>101984932
you haven't even read any of the posts, everything you just said is wrong, lmao
lowIQ imbecile, lmao

>>101984929
Tl;dr - he ggufed the T5?

>>101984951

>>101984944
>>101984944
>>101984944

File: ComfyUI_temp_mkpud_00077_.png (3.33 MB, 1360x1600)

3.33 MB PNG

>>101984869
> Therefore no two rooms will ever get the same furniture.


Ha, you see people I was right all along 8)

>>101984944
it works perfectly with empty rooms, like it's designed to
i don't even, you retards are so stupid it's like nature's trolling

>>101984970
>he ggufed the T5?

he ggufed the T5.

>>101984929
Yes

>>101984986
absolute madlad, I'm PUUULLLIIING

File: ComfyUI_32310_.png (952 KB, 1280x720)

952 KB PNG

>>101984970
He ggufed EVERYTHING!

>>101984969
ok go get excited with more shitty furniture gens, go anon run! don't waste your high IQ

any comfy nodes/tools similar to civitai manager? just for browsing/thumbnails

>>101984952
>ask chatgpt nicely
lmao

>>101984981
>was not right about anything
>celebrity coombait

>>101984984
You're going crazy for technology that only works in optimal environments and acting like it's wasted potential (>>101984811), when anything outside of that constrained input of (empty rooms) is garbage.

So why do you care so much about something that can only be applied in optimal conditions? Just seems like you're having an autism fit for no real good reason, just completely unfounded desu

>>101984998
Will also have to do a pull on comfy since some stuff had to be changed around for it to work.

>>101984999
City96 please convince kohya to make SD scripts work with your ggufed models so my loras don't have to train overnight anymore please

>>101985002
I'm rubbing it in that no one here has a clue how to mimic those results, not even the inpaint part

>>101984940
Huh, nope, same error.

>>101985012
I'm pretty sure its some guy with spamming his shitty business project

>>101985012
i don't even

>>101985017
>kohya
I don't think I've ever talked to him outside making like one github issue on his repo kek

>>101985017
you can just @ him on twitter. he's addicted and replies to everything

>>101985034
yeah, it's my site and i'm trying to develop a free workflow so no one has to pay me because i'm a genius

do you retards even hear yourselves? LMAO

>>101985034
With the way he's sperging it's probably the resident schizo farming (You)s. At this point we should collectively ignore it, he has been answered and refuses it.

>>101985037
Explain why you care so much about something that has zero potential outside of it's very specific application, if your only response is "i don't even" then it sounds like you have no way of justifying this weird rant you've been going on for the last 3 hours.

>>101985019
Oh, we know. We just don't let you in on it, because you are Turkish.

>>101985049
you haven't answered shit you fucking imbecile
you have no clue how to achieve what site does, no one here does, or even half of it

File: ComfyUI_00905_.png (985 KB, 1280x720)

985 KB PNG

>>101985024
Well take a screenshot of your workflow and let's see what you're doing wrong

>>101985041
No time like the present... I asked him if he'd implement training for optimized flux models and he said he thought it'd degrade things too much. I'm assuming this is not the case with gguf though..

>>101985052
because i need to do that at a massive scale and don't want to pay that shitty site
also, there should be a way to emulate that process with either comfy or A1111

>>101985080
that can't be AI, damn!

>>101985088
this board need posters ID

>>101985017
>>101985017
what would even need to be changed in the code to support it? surely it can't be that hard?

>>101985082

Oh okay, as long as you're going mental over an extremely specific application over the tool - however you basically gushed over >>101984811 it's potential for things beyond furniture and I assumed living spaces.

"Large scale?" What's your usecase?
>>101984893 kek

Death to celeb sloppers

File: ComfyUI_32311_.png (889 KB, 1280x720)

889 KB PNG

>>101984840
Works perfectly with the default clip_l, but I'm getting this error with ViT-L-14-BEST-smooth-GmP-ft

Error occurred when executing DualCLIPLoaderGGUF:

'GGMLTensor' object has no attribute 'tensor_shape'

File "x:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 316, in execute
output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "x:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 191, in get_output_data
return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "x:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 168, in _map_node_over_list
process_inputs(input_dict, i)
File "x:\AI\ComfyUI_windows_portable\ComfyUI\execution.py", line 157, in process_inputs
results.append(getattr(obj, func)(**inputs))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "X:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF\nodes.py", line 224, in load_clip
return (self.load_patcher(clip_paths, clip_type, self.load_data(clip_paths)),)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "X:\AI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF\nodes.py", line 193, in load_patcher
if clip.cond_stage_model.clip_l.transformer.text_projection.weight.tensor_shape == None:

>>101984840
JESUS CHRIS I'M SWIMMING IN FREE VRAM NOW

>>101984893
you're not white

>>101985120
don't give a shit, just want to achieve the same thing

>>101985153
I don't think any 4chan anons are gonna help you replicate the workflow of a business for free lol. I suppose you could just pay for their API if your large-scale usecase is that important.

>>101984840
Is this useful in any way for people who aren't VRAMlets aka (2x3090s)? I'm assuming it's slower than just using the comfy clip offloader.

File: 00000-3086111536-Australi(...).png (850 KB, 984x984)

850 KB PNG

Hello

File: ComfyUI_00910_.png (831 KB, 1280x720)

831 KB PNG

>>101985094
What we need is a whole new AI board, with new features like workflow sharing and stuff. MOOT would have already done that if he was still here

>>101985175
no one here knows to, is not about the money
I could pay here anyone 5K right now and no one would have an answer either
you faggots have no clue how to achieve that
not just that, no one here can even furniture an empty room from scratch

File: 1714021120502380.png (507 KB, 2422x1093)

507 KB PNG

>>101985061
Sure, here.
Basically I'm trying to load an image, pad it with extra pixels for outpainting, then use Advanced ControlNets so I can mask out the existing portion and apply the ControlNet ONLY to that portion, not interfering with the new pixels. My hope is that by guiding the upper portion it'll help the model to generate something that actually fits in the new pixels, but I could be completely wrong. Either way, I can't even try it because of that error.

>>101985198
Okay that's cool dude, if no one knows how I guess there's no point you being here. Let us know if you make any progress.

>>101985192
LOL, keep dreaming
this site is dead af

>>101985210
>Let us know if you make any progress.
never, I hate these threads and everyone in them

>>101984963
cute, pixel art prompt or lora?

>>101985221
Sounds like a problem, you ever considered quitting?

>>101985230
quitting what?

File: ComfyUI_32312_.png (1010 KB, 1280x720)

1010 KB PNG

>>101985179
Made my s/it about 20% faster on 3060 compared to t5xxl_fp8_e4m3fn, even though I could fit everything in VRAM before.

>>101985235
browsing 4chan if it makes you mad

File: delux_hh_00005_.png (1.95 MB, 1024x1344)

1.95 MB PNG

>>101985192
>AI board
you're preaching to the choir

>>101985198
whats your end-game, exactly? just to keep yelling at people until someone randomly make your business app for you?

>>101985249
i'm not mad at all

>>101985256
If you hate these threads and everyone in them it sounds like they made you mad. Or is it irrational hatred?

>>101985229
https://files.catbox.moe/spbunb.png
No loras activated.

>>101985255
imagine being this stupid

File: 00023-2614232807.png (3.17 MB, 1280x1920)

3.17 MB PNG

>>101985221
>never, I hate these threads and everyone in them
Then why are you here then? Lol.

>>101985266
How's your upscale workflow? It any good?

File: 00001-1750971038-Australi(...).png (870 KB, 984x984)

870 KB PNG

>>101984940
>>101985061
>Agent Smith
Kek

File: delux_hh_00004_.png (2.13 MB, 1024x1344)

2.13 MB PNG

>>101985268
ok, I'm imagining. now what?

>>101985265
it's rational hatred
most people here are clearly autistic and don't quite understand human emotions

>>101985272
to see if anyone with a brain might come up with something
needless to say, hasn't happened yet

>>101985132
I'll test it in a bit and try to fix it.
>>101985179
Nah just pin it on cuda:1 together with the VAE, that's always gonna be faster than loading/unloading and it takes load off cuda:0.

>>101985287
If you hatred is irrational, and you identify elements of the people here that make you dislike them - is that not at least some form of anger? Which would make you mad? At the very least, it stems from a negative emotion.

>>101985283
nothing, keep making your negroes I guess, lmao

>>101985279
Not really, lol, I was going to improve it but couldn't find enough motivation.

>>101985298
>it's rational hatred
>If you hatred is irrational

why are two chatbots replying to each other

>>101985324
i'm gonna make one at some point just to shit these threads even more, is literally too easy

File: 00034-4047851182.png (2.77 MB, 1280x1920)

2.77 MB PNG

>>101985336
>i'm gonna make one at some point just to shit these threads even more, is literally too easy
No need, we already have a particular individual like that.

>>101985200
Is that the default controlnet loader? Maybe you should try the xlabs one and see if that solves the issue, also change the dtype too, I haven't tried it yet s I can't say for sure though

>>101985345
no idea, i only come here when it's extremely necessary

File: xlabs.jpg (154 KB, 1655x501)

154 KB JPG

>>101985200
Something like in pic related is what I mean

File: 1721641665719120.png (1012 KB, 1024x1024)

1012 KB PNG

another ps1 miku (ps1 aesthetic lora)

File: 1718326512324761.png (1023 KB, 1024x1024)

1023 KB PNG

>>101985393
ps2

File: ComfyUI_32314_.png (1.15 MB, 1280x720)

1.15 MB PNG

>>101985324
>two chatbots replying to each other

>>101985430
i'm not a gay trannime faggot, I actually work with SS, tho

any nodes like civitai helper? or that have lora thumbnails etc? I can tell most by name just curious

>>101985422
takkun

>>101985132
>>101985292
Okay no clue why it does that lol, I'll look into it tomorrow.

>>101985451
social security?

>>101985522
*SD

File: ComfyUI_00042_.png (1.68 MB, 1024x1024)

1.68 MB PNG

Why can't Flux understand basic layout?
>a river runs left to right at the bottom of the photo. It is a grasslands, devoid of features otherwise.

>>101985525

>not a gay trannime faggot

>works with SD

hmmm

>>101984963
Is the adventure game genre coming back, bros? This would make it so easy.

>>101985537
yeah, because making gay trannyme shit here 24/7 somehow generates revenue?
if so tell me how, because it would be easy as fuck

File: delux_hh_00007_.png (1.9 MB, 1024x1344)

1.9 MB PNG

>>101985528
describe the river last

>>101985561
oh you're one of *those* people

why don't you just get a job?

>>101985567
I just said i work with SD...

File: ComfyUI_32319_.png (1.24 MB, 1280x720)

1.24 MB PNG

>>101985547
Soon.

>>101984927
what lora anon?

>>101985584
clearly not very good at your job if you have to come begging to us for help with a basic workflow LOL

We're so back. Now I should be able to fit both Flux and an LLM into VRAM at the same time. I don't have to load and unload.

>>101985600
>basic workflow LOL
none of you have a single clue how to achieve what I asked, all you can do is sit here and make trannyme trash 24/7
you will never work in this field

>>101985617
you wont be working in it for much longer if you can't get your head around a *very* basic application lmao

>>101985622
>lmao
you're laughing at yourself, you have no clue to do what I asked

>>101985594
ps1_style_flux on civitai (might just be called ps1 or ps1 game style)

it does the low poly look nice so it's a good style in general

>>101985629
Yet you're the professional you're PAID TO DO THIS and yet you're on the level of some hobbyists dicking around with generating some waifus.

I really, really fail to see how I lose in this situation.

What's the best way to scale up flux images? I only have a 12gb GPU.

>>101985637
>I really, really fail to see how I lose in this situation.
I work with SD, you sit here all day making anime trash
I just explained this...

>>101985650
And again, you wont be working with SD much longer if you can't even do your job lmaoooo

>>101985657
>lmaoooo
you are literally retarded and it's very amusing

>>101985666
kek, im not the one who can't even do their job LOL. all you have is calling others retards when you can't even do the thing you're paid for. bet you're getting real nervous wagie

>>101985568
nice and unironically its real.
apple is joke now

>>101985677
>LOL
bro, you're the joke here

>>101985688
"Bro" i'm not the one employed to beg on 4chan because you're incapable of doing your job hahahahahaha

>>101985688
is your job on the line or something? is that why you're hella desperate?

File: delux_hh_00009_.png (2.58 MB, 1024x1344)

2.58 MB PNG

these guys are funny to watch

>>101985324
just report it for off-topic

>>101985694
>hahahahahaha
you're obviously having some kind of meltdown for being a loser

>>101985695
how am I desperate?

I don't see why T5 couldn't be loaded in q8. We can load it in fp8 but that is not as accurate as q8... any progress has been made yet?

>>101985703
it seems like you are dead set on getting any kind of help you can and we're not sure why

>>101985703
no you literally just don't have a response to the fact that you can't do your job. it's just funny that you're here calling others retarded when you're literally begging for others to spoonfeed you.

How did you get the job? Did you show bossman your 1girl slop?

>>101985709
>>101984840

>>101985709
>t. least perceptive anon
>>101984840

>>101985697
>he likes watching shirtless negroes?
manlier poster in these threads

>>101985714
I srsly doubt i'll get any help at all from a bunch of talent-less retards, desu

>>101985725
you're obviously having some kind of meltdown for being a loser
and it's very funny, plz keep replying

>>101985740
funny that you shut down when someone calls you out desu

>>101985709
anon... >>101984840

>>101985740
yeah because it's impossible to tell what they're doing exactly

you're better off directly reaching out to them instead of asking coom brained anons on 4chan for help oif all places lmfao

>>101985744
shut down how?

>>101985747
>HI HELLO, COULD YOU PLEASE EXPLAIN EXACTLY HOW YOUR BUSINESS WORKS SO I CAN COPY IT, THANKS!

you stupid retards are hysterical, unironically

>>101985732
>>101985738
Nice

File: b_00264_.png (926 KB, 1280x720)

926 KB PNG

Flux

>>101985767
it's pathetic how you can't do the job you're paid to do, despite bragging about working with SD. then when I called you out on that you shut down and just kept calling me a retard, i'd like to see you actually come up with some kind of retort that isn't just some personal attack - because that tells me my words are actually true.

File: DSSSXX.png (859 KB, 1280x720)

859 KB PNG

>>101985774

>>101985774
WAY too skinny to be trump

>>101985767
you obviously do not know how to talk to others lol

I bet I could reach out to them and get their entire workflow if I really wanted to

>>101985800
this guy is probably just hoping that someone on 4chan will do that for him anyway, kek.

>NONE OF YOU RETARDS KNOW ANYTHING ABOUT ASKING THIS WEBSITE FOR THEIR WORKFLOW, NOBODY HERE IS CAPABLE OF THAT

File: b_00268_.png (971 KB, 1280x720)

971 KB PNG

>>101985793
true

>>101985781
you're an imbecile, your brain doesn't work
we work with SD
my job is not figuring out how to furniture empty rooms
it would be cool to add that service, tho
hence why I'm here
I told everyone you stupid retards probably wouldn't have a clue because you're just a bunch of basement dwelling losers that spend your pathetic lives here 24/7 generating pictures of furries and trannyme girls, but I am being paid to do this research, so I've basically being working for the past 4 hrs and everyone is having a laugh here at the office with this interaction

File: ComfyUI_32330_.png (984 KB, 1280x720)

984 KB PNG

Copypasting anime synopses as prompts can give nice results.

>>101985824
Take your shit laced project somewhere else pajeet.

>>101985800
I mean, there's a really nice market for it
we might reach out to them to offer the service from out site, but first we're going to try to figure it out ourselves
it's not like the business depends on it or anything
what's curious is that there's many other sites offering that service but they're all utter shit

>>101985824
>but I am being paid to do this research, so I've basically being working for the past 4 hrs and everyone is having a laugh here at the office with this interaction

you don't live in reality. lmao. this is the most ridiculous attempt at cope i've ever seen.

>hahaha yea me and the guys at the office are laughing at you on 4chan.org, as i call you all faggots and retards

are you genuinely deranged? did you genuinely think anyone would believe such a transparent lie?

jesus christ, you know what - you have my pity. I dearly hope you get some help, you clearly need it

File: delux_hh_00011_.png (2.23 MB, 1024x1344)

2.23 MB PNG

>>101985774
is this real?

>>101985842
no, sirs
show bob and vagene

File: ComfyUI_32325_.png (1.08 MB, 1280x720)

1.08 MB PNG

Boku to Roboko

>>101985823
even you democrats hate faggots.

File: ComfyUI_32326_.png (1.21 MB, 1280x720)

1.21 MB PNG

Shikanoko nokonoko Koshitantan

>>101985851
>subhuman 4chan poster retard can't even conceive people actually have jobs
kek, I love it

>>101985824
Wow share curry recipe

File: b_00089_.png (988 KB, 1280x720)

988 KB PNG

>>101985863
I think both democrats and republicans are niggers

>>101985875
no, you have my pity because you genuinely thought people would believe that your "job" is sitting in an office browsing 4chan for four hours while laughing with your colleagues about calling people faggots and retards

that is incredibly sad, and not based in reality at all - which genuinely makes me feel bad for you

>>101985880
Your friend literally shot at Trump. We all know that Crooks, who was the 7/13 shooter, was a 4chan mod.

File: ComfyUI_32328_.png (961 KB, 1280x720)

961 KB PNG

Yuru Camp (>>101985723 is Yama no Susume, >>101985838 is Starship Troopers, >>101985844 is VTuber Nandaga)

>>101985877
i cant sirs, our business depends on it

>>101985882
I mean, just for today or so, it's clear you stupid retards have no clue how do replicate the process
now we're just trolling waiting to see if anyone suggests anything, but it probably won't happen

File: file.png (771 KB, 2337x1057)

771 KB PNG

>behold, the billion dollar workflow

took me 10 minutes.

kek, as many people in this thread have said. time and time again, it's just an application of segmentation, masking and inpainting. literally what the website shows and can be seen with basic testing

can you fuck off now?

File: ComfyUI_32334_.png (1 MB, 1280x720)

1 MB PNG

Shuumatsu Train Doko e Iku

File: delux_hh_00012_.png (2.19 MB, 1024x1344)

2.19 MB PNG

>>101985900
I can tell this is fake because he's not shidding and farding himself

>>101985907
1) lighting doesn't make sense
2) there's no point in having a cuck chair if there's no bed

>>101985897
>now we're just trolling waiting to see if anyone suggests anything
>we

im sorry but again this is just a massive cope, and i pity you. if you and your colleagues are all in on this then surely you can attach an image with a timestamp and everyone in the office? you know, cause you're all just trolling

File: ComfyUI_32327_.png (1006 KB, 1280x720)

1006 KB PNG

Yuu Yuu Hakusho

>>101985907
finally a little something
the perspective is pretty good, try to give it more complexity, put a rug, a couch, a coffee table, fill up the space

File: ComfyUI_32336_.png (1.02 MB, 1280x720)

1.02 MB PNG

Shigurui

>>101985922
yeah right
bro you're a fat autist with no life, we don't want some loser sending us 100 pizzas everyday for a year because I troll you in some shitty thread
i don't give a shit if you believe or not, kek

>>101985907
thanks btw, at least it's something, kek

File: 1694290008912501.png (1.22 MB, 1024x1024)

1.22 MB PNG

I knew you were a comrade when you walked in...

File: ComfyUI_32338_.png (959 KB, 1280x720)

959 KB PNG

Unexpectedly, Boku no Pico

File: 00005-2937671911.png (1.16 MB, 896x1152)

1.16 MB PNG

Is there a reason that Pinokio isn't in the OP? That shit is awesome, it's a 1-click installer for SVD, ComfyU1, Automatic1111, Bark, and so many more. It's so convenient.

>>101985969
Buy an add

File: file.png (654 KB, 2134x966)

654 KB PNG

>>101985927
wait, this is actually what you couldn't pull off?

there's no way. i was just trolling you with this piss easy workflow expecting you to get mad. fucking hell you actually are an imbecile, and you want more?

man, im so satisfied right now. you're a braindead idiot

>>101985938
>>101985922

me btw, and i was churchill, and basically half the other guys while my LoRA was training. so this is incredibly satisfying for me to know that you really are a loser who knows NOTHING

>>101985944
it's so fucking funny that you folded at the smallest amount of help. i hope you keep your job, because your last few posts were genuinely, very, very pathetic and the only reason i made this workflow is because i genuinely pitied you.

good night and sweet dreams you little spoonfed retard

>>101985911
I haven't seen anyone else mention that show since it aired kek.

>>101985975
I'm not him, I literally found out about it 2 days ago and suddenly it's been a breeze installing everything. You can literally just type in and one-click install everything in the Advanced UI section, provided you've got an NVIDIA GPU. It's genuinely useful and I've used it to find tons of other stuff I've wanted to try out.

File: delux_hh_00013_.png (2.13 MB, 1024x1344)

2.13 MB PNG

>>101985978
stop, he's already dead

File: ComfyUI_32341_.png (1.02 MB, 1280x720)

1.02 MB PNG

Sasakoi isn't yuri anymore
>>101985983
I still haven't finished it, but i's a fun ride.

File: ComfyUI_32343_.png (1.15 MB, 1280x720)

1.15 MB PNG

Princess Tutu

>>101985969
Catbox for the elf, if you don't mind? Thanks :)

>>101986018
cute, lora or pixel art prompt?

File: ComfyUI_00047_.png (1.41 MB, 1024x1024)

1.41 MB PNG

>>101985563
>there is an empty grassland. There is nothing but grass, save a river running left to right through the photograph.

File: 00004-2937671910.png (1.2 MB, 896x1152)

1.2 MB PNG

>>101986024
They're just images I made of Princess Zelda smoking weed to try out the SVD and video generation. Here's the other good one I got out of Flux, for some reason it wants to force anything video game related into a cartoon.

File: delux_hh_00014_.png (2.37 MB, 1024x1344)

2.37 MB PNG

>>101986027
welp, I'm out of ideas
if you really want a specific composition, you can doodle something in mspaint then i2i it but I get you're trying to do it with a prompt

>>101986025

pixel art, 8-bit, large square pixels:

In a fairy tale come to life, the clumsy, sweet, and gentle Ahiru seems like an unlikely protagonist. In reality, Ahiru is just as magical as the talking cats and crocodiles that inhabit her town—for Ahiru really is a duck! Transformed by the mysterious Drosselmeyer into a human girl, Ahiru soon learns the reason for her existence. Using her magical egg-shaped pendant, Ahiru can transform into Princess Tutu—a beautiful and talented ballet dancer whose dances relieve people of the turmoil in their hearts. With her newfound ability, Ahiru accepts the challenge of collecting the lost shards of her prince's heart, for long ago he had shattered it in order to seal an evil raven away for all eternity.

Princess Tutu is a tale of heroes and their struggle against fate. Their beliefs, their feelings, and ultimately their actions will determine whether this fairy tale can reach its "happily ever after."

pixel art, 8-bit, large square pixels:

clip_l is "pixel art, large square pixels" too.

File: ComfyUI_32345_.png (1020 KB, 1280x720)

1020 KB PNG

Mazinger Z

So can flux work on a gtx1070 or nah

File: ComfyUI_32347_.png (1.06 MB, 1280x720)

1.06 MB PNG

Mahou Shoujo Site

>>101986078
Me on the left, trying to ignore

File: ComfyUI_00023_.png (1.43 MB, 1344x768)

1.43 MB PNG

>>101986065

Works on my RTX 3060 with 12GB VRAM so I imagine with a 1070 and 8GB VRAM you'd be a bit slower than me but definitely able to do it.

File: ComfyUI_32349_.png (956 KB, 1280x720)

956 KB PNG

Mahou no Star Magical Emi

File: ComfyUI_00048_.png (958 KB, 1024x1024)

958 KB PNG

I cracked the code.

File: ComfyUI_32350_.png (1.04 MB, 1280x720)

1.04 MB PNG

Mahou no Princess Minky Momo

>>101986100
There is nothing to crack, it's all oboob source.

File: ComfyUI_32351_.png (1.11 MB, 1280x720)

1.11 MB PNG

Mahou no Angel Sweet Mint

>>101986100
tupperware tits

File: 1722348169713399.png (1.42 MB, 1024x1024)

1.42 MB PNG

guys I prompted eva braun, what happened?

>>101984401
You can train image + text to image models by adding input channels in the unet of SD 1.5.
I did that for a project at school for creating more perspective of one image one year ago (by further finetuning this https://huggingface.co/docs/diffusers/main/en/api/pipelines/pix2pix).
It's not new. Just have empty room + furnished room pairs (which u can synthetically make by inpainting tons of furnished rooms and the manually sorting those that are bad and have some real image pairs). Use a image segmentation model that identifies the objects and their location in the picture (I'm sure even gpt4 vision is enough for this if you tell it to give a strict representation). Then fine tune the model i linked with the instruction based on the image segmentation description text and the empty room image.

>>101985978
Thank you anon, for making that guy fuck off. Truly impressive you were able to replicate all that shit in comfy ngl.

You just know he's studying your workflow right now, lmao.

File: ComfyUI_32355_.png (1.02 MB, 1280x720)

1.02 MB PNG

Hello!! Kiniro Mosaic

>>101986162
What's cogflorence?