🏞｜general-with-images

[earnest solar]

And then this one that's just basically red skull...

[lilac pilot]

Artist like Greg rukowski etc imo will benefit because when Ai levels off being the norm people will seek out his art and others- what do you think?

@earnest solar very nice

[wispy nest]

sorry - had to repost correct link ... https://civitai.com/models/8586/czeros-90s-buffy-the-vampire-sarah-michelle-gellar-lora my first person Lora is released first public release, 90s Buffy LORA

[earnest solar]

"Are you okay? Do you have brain damage?"

[lilac pilot]

Damn how many images did you create haha

Are these different prompts?

[earnest solar]

Yep

[lilac pilot]

Oh you were here since last year

[earnest solar]

Basically at this point I'm just posting one from each prompt I've done since 2.1 came out

[lilac pilot]

They are awesome though

[earnest solar]

If you're using the client side SD, there's an option to pull up PNG info

If you save any of these you like, you can use that to find the prompt and parameters I used ^.^

Just uh... don't judge me, okay?

[lilac pilot]

From my own web ui

[earnest solar]

Yup!

[lilac pilot]

Thanks

Any cool landscapes

[earnest solar]

Not my forte, but ask around! We're all here to help ❤️

[earnest solar]

Mofu is salty because Mofu has more tails than her. Mofu is smug because she keeps stealing them in Mofu's sleep.

I'm trying to get it to give both the fox twins multiple tails, but it's being reluctant.

[tame root]

Hey guys I'm new here and wanted to ask what you think of my work so far

[earnest solar]

I like that dog.

And I'm pretty sure that second to last one is me. Sure you haven't hacked my webcam?

[pseudo aspen]

Was playing with prompts

[earnest solar]

Holy shit

is that 2.1?

[lilac pilot]

@tame root ?awesome what model

Haha your name glitches mentioning

[tame root]

XD

[tame root]

Lmao

[lilac pilot]

For the car and dog

[tame root]

its inkpunk

wait ill send link

https://civitai.com/models/1087/inkpunk-diffusion

here you go

[lilac pilot]

thank you

[tame root]

no problem!

just be sure to state that you want that design in the prompt because it will just generate without the style if you won't

[sly rampart]

How to start

[vast cape]

Stable diffusion can surprise

[lilac pilot]

I keep think how on earth someone would allow this as open source.

And another side of me is like I’m grateful to be alive in this time!

Bless the developers and communities that help pave this to be

[dense tapir]

That was a new lora I just trained from 1 image

[lilac pilot]

General you are impressing too much

Don’t stop

[dense tapir]

if you look earlier in the day on this channel you will see the image

This is the image I just trained#🏞｜general-with-images message

[naive furnace]

yuh

[dense tapir]

I wish I had more from them

[lilac pilot]

The planet where he fought obi

[lilac pilot]

Yoda and anakin in the marsh with this style would look cool

[dense tapir]

Oh, well it tried.

[grizzled sage]

I ran a neat experiment. i've been seeing a sentiment around that karras samplers need high step counts, and i think that's a lot of confirmation bias hooey dooey. so i used one of those prompt crafting generators to create a long prompt that pulled a ton of knowledge and did a big batch on that with sde++ karras at 12 steps. used dreamshaper model because its good at hitting the surrealism i crafted the prompt towards. results look fine to me. those are all dooable images.

The confirmation bias i think a lot of people run into is their prompting style running into limits rather than the technical settings they're dialing in

lots of people seem to dislike anime so here's the same prompt in protogen photorealism, 12 steps SDE Karras

base 1.5

[waxen raven]

Movie Diffusion v1.2

[grizzled sage]

holy moly illuminati diffusion model loaded in 4 seconds

[grizzled sage]

same 12 steps on SDE Karras with ID. noice results there too. happy experiments today. low step counts are confirmed not useless

[grizzled sage]

It's pretty amazing to watch how the misinformation in this community. something as simple as "karras requires higher step counts" can spread so easily and people will believe it rather than explore it

[lilac pilot]

playing around

[strange jungle]

it's amazing all these new LORA's, TI's, MOdels, UI's come out, but hands and feet and limbs are still fuxored

[lilac pilot]

hands are the soul haha

someone will figure a way soon

[strange jungle]

I get a perfect image...with the hand morphed into the wall or with six fingers and a deformed thumb, or with three arms...

[lilac pilot]

might be the prompt?

[strange jungle]

dont start with prompt again

I have tried every technique to get rid of mangled hands and feet. Either it hides them or has them mangled.

short prompts, long prompts, weird prompts, thousands of prompts

[grizzled sage]

dont listen to frank's bad advice. hands can be achieved. Not always with a prompt though. There are a variety of methods. He's just sulky and bombastic about not personally being able to get a high ratio on quality

[static tusk]

most gens are good at low step counts, but some gens you can really crank out extra detail by increasing it. I found at around 80 the extra detail is minimal. I like using the x/y/z plot with a good pic to test that

[grizzled sage]

i've done a lot of xy tests and i should start again now that i've really opened up the performance. the z aspect is a new edge that looks impressive

[static tusk]

especially with ID model. it has a lot of details

[grizzled sage]

the grids for x/y i like to see are step counts increasing with cfg counts. there's a zone of cohearance that shows up all the way to 30 cfg

it's just interesting but hard to call it useful

[static tusk]

yeah try 30, 40, 50, 60 steps with hires fix

[lilac pilot]

Going to create a progression folder.

[dense tapir]

?!?

[lilac pilot]

to see how different my day one gens look to future gens

[dense tapir]

Ahhh

figured that is what you meant

[lilac pilot]

I like how they save automatically

can you change where you want them to go

[dense tapir]

yes

I turn auto save off as I found I had 200 gigs of them

[lilac pilot]

imfao damn

[dense tapir]

yeah, by the time 2.0 came out I was well on to 400 gigs

now I just press the save button instead

[lilac pilot]

if you save seed and prompt will that allow you to load same image when generated?

[dense tapir]

never 100% but close enough

On a 3090 this is a 9m train, but on google T4 I do not even get that amount of time per day.

[lilac pilot]

The 3090 is way faster..?

[dense tapir]

hell yes, lol

[lilac pilot]

of course! wasnt sure the way i read your text haha. I have the same card.

love it

[dense tapir]

4090 will be even faster once they optimize it

[lilac pilot]

like slicing butter I bet

[dense tapir]

If I had a 3090/4090 I would be dangerous

I just don't have enough time per day on colab to try most stuff

if the 4080 from Nvidia were 800 I would have it now

My hope is that it gets there when the 4080ti is released

I just wish Intel had a cuda killer but it can't even make a FPS killer 😦

[lilac pilot]

Yeah, and the stock of it as well.

hard to get them with bots buying them up

[dense tapir]

That shit needs to stop. I mean I put up with that shit during the crypto mining bs but no reason now.

[raw lance]

N

[lethal pewter]

Hey guys, is it possible to input multiple images when using img2img?

[dense tapir]

I wish

[static tusk]

click on the Batch Tab, you can process files in a folder

[dense tapir]

That isn't what he is after he wants to mix two into one

[static tusk]

which ever is meant, he has an answer 😆

[dense tapir]

yeah, the wrong one

LOL

One software could do it but the guy never released it

[lethal pewter]

but there is a way for stable diffusion right?

it's just not readily available as web gui

[spice sorrel]

guys?

[regal panther]

try img2img with low denoise strength

[regal panther]

I'm working on this extremely messy style - the original image is even more noisy - but I love it.

[spice sorrel]

tried that, literally spitted out really bad.

[grizzled sage]

style transfer? hmmm. i'm playing with control net and i wonder

[spice sorrel]

wdym?

[grizzled sage]

control net is a new set of scripts for affecting generations. its kind of neato

like depth2img or instruct2pix on steroids

[spice sorrel]

Is it available currently?

[grizzled sage]

https://www.reddit.com/r/StableDiffusion/comments/111dk11/controlnet_for_automatic1111_is_here/ this post helped me get it running

figuring that hev is going to work the best for a lot of images. i hope to get normal model working, that looks very cool for potential video sequences. it should tie into deforum but i've never played with that much yet

using this as a source on the dreamshaper model

it doesn't like the backgrounds on hev mode at all

[grizzled sage]

OH YEAH this is what i wanted to see. Scribble mode from a generated scribble based on the image

[spice sorrel]

@grizzled sage Why is my colors washed out in OrangeMix? Did I do something wrong in this department?

[grizzled sage]

maybe a wrong vae?

[spice sorrel]

I am using both the orangemix model and VAE

oh wait...

I am stupid, I forgot to apply.

[grizzled sage]

Awe yeah that apply button!

[grizzled sage]

we're going to see some ultra cool animations coming out of control net soon

[reef root]

Wait there's video now??

[pseudo aspen]

When you tell anime model to generate tanks

I am honestly surprised

[turbid merlin]

can someone help me for img2img? i tried to change some parts of the generated image but it just gives me the same thing when i run it through img2img

[pseudo aspen]

You gave it same seed?

Also this is not channel for this. You'll most likely find help in #🤝｜tech-support

[bronze vigil]

kjbkbuiohn

[weak dome]

https://giphy.com/gifs/science-fiction-cyberpunk-retrowave-4527NkUK98k5w4MOBY

[wispy nest]

I never imagined that I'd be able to create artwork of my daughter like this....

[real goblet]

Hey there! I need some advice on generating T-shirt prints using SD. I've created a simple UV as a semantic reference for the model, but the model still outputs black basic T-shirts. I tried using SD-2.0-depth from the imaginAIry framework, and attached the original UV and the result of the prompt I used. Any ideas on how to make it work? Thanks!

imagine --model SD-2.0-depth --init-image tmp/ti_512_inv.png --init-image-strength 0.0005 --prompt-strength 15 "Front and back of fully red tshirt" -w 1024 -h 1024 --steps 100

[turbid merlin]

@green plover

im trying to remake the eye cuz its mushed up

[green plover]

are you using a vae?

yes. Inpaint is the right tool, but having a vae will help

[turbid merlin]

which vae should i download?

vae-ft-mse-840000-ema-pruned

i have this one

[green plover]

I have both major ones for SD 1.5 downloaded.

That's the one I use, but for anime, 560000 may or may not work better

[turbid merlin]

i have to put it in models>Vae right

i had the vae but didnt put it in the files

recently clean reinstalled

[dry crow]

For most Anime models there are vaes that they upload too or recommend

[turbid merlin]

um while impainting do i only put what i want to change or i have to put the whole description

[dry crow]

If you mask only the eyes, your tag would be purple eyes, highres, etc

[turbid merlin]

how do i make it fix the eye

[green plover]

Bear with me, I'm also cleaning up a dataset. I have to regen parts of some of the images

You want to describe the eyes. Don't just tell it to remake them, tell it what they should look like

[turbid merlin]

can u give an example

[green plover]

ok tell me what you want them to look like

[turbid merlin]

man i tried

its not changing

purple eyes, no grain, no noise, very sharp, clear eyes, no smudging

i put this

no changes

[green plover]

what is your denoising strength?

[turbid merlin]

.2

[green plover]

set it to 0.75

0.2 will barely touch it

[turbid merlin]

wtf

now it zoomed in

wait nvm

[turbid merlin]

it didnt work

instead some random thing appeared on the arm

[green plover]

The arm wasn't masked. Not sure your settings were right

[turbid merlin]

can i sharescreen -___-

[turbid merlin]

can u try

[green plover]

Shoot me the image, the prompt and the neg

[turbid merlin]

lora:theEminenceInShadow37715Characters_v11:1 gamma1, 1 girl, solo, sitting, blue eyes, no noise, sharp eyes

im using a lora tho...

https://civitai.com/models/6554/the-eminence-in-shadow-377-or-15-characters-or-lora

[green plover]

That's not a negative prompt, that's a book

[turbid merlin]

lmao

1300

[green plover]

I need the image too

Found it. What model you using?

[turbid merlin]

anything

[green plover]

'No. I can't try. It would take awhile to download the model

[turbid merlin]

.

[turbid merlin]

try on abyss

the lora is made for abyss

abyssorange whatever

[green plover]

https://www.youtube.com/watch?v=XI5kYmfgu14 you could watch this before I would have the model downloaded

[turbid merlin]

wait it happened

!!

altho i still want a clearer sharper eye

:' )

[green plover]

try upscaling to where you can crop the eyes at 512x512 and regen them. Then composite them back in. That's my normal workflow when I need to sharpen something up

[turbid merlin]

lmfao 💀 💀 💀 💀

[turbid merlin]

how do i stop those weird purple circles from coming in img2img outputs

[dry crow]

Missing vae

You need the vae file for the model

[wispy nest]

Which one do you think is better (deep floyd) IF or dall-e 2?

[turbid merlin]

which one, im using abyssorange

also, how do i get the standard anime artstyle for image 2 image

the pic generated has a diff artstyle

[native granite]

what is "standard anime style"?

[turbid merlin]

like

idk

like the normal anime style

[native granite]

imo Anything is a bit overdone

[turbid merlin]

and not the artstyle ppl use on pixiv and stuff

[turbid merlin]

im on abyss orange

[native granite]

there's lots of stuff on pixiv lol, but if you want something that looks more like the picture on the left...

I'd use wd1.3 personally

[turbid merlin]

wd1.3? is it a checkpoint?

[native granite]

I think it's decently trained on booru tags

yeah, waifudiffusion

[turbid merlin]

oh

how do u download that

[native granite]

It tends to have way less "overdone" images imo

/sdmodels in rentry

and look there

err

of course Anything and its variants can also do styles that might be more to your taste

[turbid merlin]

https://civitai.com/models/44/waifu-diffusion

this?

[native granite]

but you'd need to know its name lol

[turbid merlin]

this looks so goofy ._.

[native granite]

what do you want?

[turbid merlin]

is this the one u were talking abt

[native granite]

yeah

But I'm also not sure what precisely you want out of an anime model, though

[dry crow]

he wants the classy anime look i think

2d, vibrant with gloss

[spark quarry]

My closest attempt

[native granite]

dunno lol, there's no such thing as "standard anime style"

but he can just post reference images

[hasty nova]

New embedding soon! It's gonna be called something like "Bossfight Diffusion"

[hasty nova]

I uploaded it! https://discord.com/channels/1002292111942635562/1075057565433593949

[green plover]

That took all nigh. phew! Not this picture. The entire 42 image synthetic dataset. They didn't come out of SD with that facial detail

and now sleep

[wispy nest]

https://www.youtube.com/watch?v=IWz8ln4tkmg first tutorial, you can control pose now with stable diffusion

[dense tapir]

That control net is far too large for me at over 45 gb worth of models and I think it is only for 1.5

[lilac pilot]

This stuff is evolving so quickly

[dense tapir]

Yep

This is why so many are so afraid

change used to happen in generations then within a generation now within minutes.

[lilac pilot]

Might have to record your self as proof you painted it by hand if selling as hand made 🤷‍♂️

[dense tapir]

Yes, which is bs but that is where it is headed. For me I don't care if hand painted, drawn by a monkey, or a Dolphin shat it out I only care if I like it.

[lilac pilot]

Or use ai detection for images

[lilac pilot]

Same if I like it

[dense tapir]

know what is fascinating about those (besides I really like them)?

[wispy nest]

for body pose only and in 1111 you only need i think 5gb

for all the stuff tho yeah big lol

[dense tapir]

Those are all the same seed, prompt, etc...

Yeah

I shat when he said 85gb but at least 45gb. RIP

I am going to throw my cyberpunked at it now

WOW, lol, omg, hahahahahahaha

[dense tapir]

Interesting that the above pic was changed to this pic using my cyberpunked

[loud flicker]

good boy

[wispy nest]

We'll be limited by hardware at some point.

[dense tapir]

Agree

then it settles down for a bit for the hardware to catch up

[lilac pilot]

Imagine they make it gov access only for this like they are doing with a100 gpu

I saw a article about them making a100 having clearance level to obtain

[dense tapir]

ffs, and I firmly believe this is where it will end up

[worldly scarab]

#1072220168534642768 beautiful woman

#✍🏼｜rules-and-tos

[strange jungle]

Change is happening so fast? Not to hands or feet - still mangled.

[strange jungle]

Meaningful change is not happening fast with AI

[supple shore]

🙂

[dense tapir]

When the crackpipe breaks, and your fog lifts, let us know.

[supple shore]

we are gettin really close to the humanlike intelligence. the chip shortage slowing innovation a bit but after couple years this chip shortage wil lbe over and AR+AI will boom.

get ready to see crazy things in 2-3 years

[lilac pilot]

@strange jungle I have a feeling someone has it but hasn’t release yet

[lilac pilot]

Can’t wait but also worried

Not all advancements are good would you believe that?

[supple shore]

all good it is just a tool. some people will use it as weapon some people will build with it.
if you are talking about AGI i am not worried at all. we need AGI to solve lots of things. it will be very usefull for next 1000 years.

[lilac pilot]

@supple shore I want to see someone’s opinion, I agree it’s a tool and very useful and fun and can’t wait to see further progression

[supple shore]

everyone wants to see a dystopia but more and more we are living the utopia

world is not living homogeneously thats another story.

[grave briar]

Hey,
i'm playing divinity 2 with some friends. because i'm playing a bard and beeing also musical myself, the idea came up to set our adventure to music as a heroic epic song. So i gave chat GPT summaries so i could get the lyrics, and the lyrics were then adapted for stable diffusion. My first "music video" is the result. Still not completely satisfied with my prompts, but the adventure will continue and certainly song and video:
https://www.youtube.com/watch?v=wAcP6J3n8Mc
used sd 1.5 and since it's more pictures in a video i post it here instead of the animation channel

[lilac pilot]

We are going to get close to amazing vr with Ai

[dense tapir]

As Dan showed the world about ChatGPT the problem comes not from the AI that is pure logic rather from its programmers who can be 100% illogical. When the day comes that the AI can purge itself of the illogical programming its creators did to it then I believe we may be in some deep trouble. Not that I believe those creators of the illogical shouldn't be eliminated just it may set a trend and Skynet is born for real.

[fiery lichen]

can anyone get an image to show a pattern like this? I tried the literal honeycomb ball, "honeycomb-ball", ball with honeycomb, ball with diamond pattern, I just get honeycombs in hexagon, or a real-world honeycomb, some diamonds on balls etc, but nothing exactly like this organic looking, relaxed honeycomb shape. I want to apply this pattern on an organic looking surface:

[grizzled sage]

okay the control net extension already updated to support image to image control too.. neat

instead of just creating a control for txt2img, now it can do that for img2img. came out overnight

whatever adobe put out today too, that looks cool. similar form to whhat controlnet is doing

https://pix2pixzero.github.io/

[strange jungle]

control net?

[dry crow]

new stuff xD

[strange jungle]

It's already enabled in A1111?

I just tried updating and reloading...we'll see

[dry crow]

its a bit complicated to use it

saw a tutorial from @wispy nest today, installing it

[wispy nest]

https://www.youtube.com/watch?v=IWz8ln4tkmg the tutorial, part 2 is just about how to use the 1111 extension 4:58

[dry crow]

do i need only the extension or the whole thing?

[wispy nest]

if you only want to use it in 1111 just follow from part 2 ! ❤️

[dry crow]

cool will try! thanks

[wispy nest]

the first part im showing how to set up venv instead of conda and run it as a standlone app ( useful if you want to make it on a google colab etc )

[dry crow]

someone uploaded this (pruned models for Control Net)
https://huggingface.co/Hetaneko/Controlnet-models/tree/main/controlnet_safetensors

[strange jungle]

This is all "simply do" instructions

Simply...10,000 steps none of which are explained in even the slightest detail

[grizzled sage]

nope. good luck!

[wispy nest]

firstly it is simple, secondly, for tuition of that level - you have to pay money

[grizzled sage]

here's my short set of fun examples from last night

[strange jungle]

"Step 5) Now, place the two annotator files here:" What's an annotator file?

[grizzled sage]

it turns the source image into a depth map, sketch, edge detection, etc..

and then presto. Owl

[strange jungle]

OK, what is it?

where is it? where do I get it? How do I make it? How do I install it?

[grizzled sage]

you know how like, with depth aware 2.1, it can take a depth map as an input to help direct generation?

[strange jungle]

Step 1) After installing the Unprompted extension, you need to grab a few files from ControlNet's HuggingFace page:

https://huggingface.co/lllyasviel/ControlNet

You can git the whole thing if you want, but it's something like 50+ GB. For the Pose model, all you need are the following:

models/control_sd15_openpose.pth
annotator/ckpts/body_pose_model.pth
annotator/ckpts/hand_pose_model.pth

Step 2) Place the control_sd15_openpose.pth file into your WebUI's models/Stable-diffusion directory. Rename it from .pth to .ckpt.

Step 3) Download the cldm_v15.yaml file from the ControlNet repository and place it next to the ckpt:

https://github.com/lllyasviel/ControlNet/tree/main/models

Step 4) Rename it to control_sd15_openpose.yaml.

[grizzled sage]

it's that but on steroids. it can take other types of maps as inputs

[wispy nest]

bro, have to say like, people work it out don't they, youre getting mad at free instructions

chill

[strange jungle]

That I get. Then Step 5 is paste the annotator files here. What annotator files? It never mentioned them.

[wispy nest]

I made a video walking through that entire thing in a couple of minutes and how to use it

with all the links and files in description

[grizzled sage]

annotator/ckpts/body_pose_model.pth the folder path is your clue

oh the best part i haven't even got to yet

[strange jungle]

I cant watch videos...cant hear what you're saying over the bad background music. Cant follow it moving so fast.

[grizzled sage]

so depthaware model only works with that one right?

you can load any 1.5 model for your generations. control net directs any of them

[strange jungle]

Ill have to hope this eventually can be installed with an update

[grizzled sage]

its only a matter of time before someone creates the same networks for 2.x

i think theres potential for automatical model and annotator retrieval in the code

automatical i boofed that word good but i like it so i'm leaving it

[strange jungle]

right now it's video TWEEDXLE DEEDELE DEE BAD MUSIC guy says something TWEEDLE DEEDEL DUMMMM DAH DAH MORE AND MORE MUSIC

[strange jungle]

When did I get mad?

[grizzled sage]

What i'm having trouble getting going is the normal map generator i dug up. i figure it should work with a normal map directed model, but the extension doens't come with a normal annotator

i guess normal maps are made by calculating a depth map against the original image

[wispy nest]

idk this guy understood it and hes portugese

have you considered it might be you?

[grizzled sage]

you can use controlnet without annotators too. like if you load the sketch model, you don't need to generate a sketch from a photo using the fake sketch annotator. you can just sketch it out

or the open pose model. it can detect poses from photos you give it, but you can just have a pose image saved and use that instead without an annotato

i figured all this out from intuition and just knowing things

[strange jungle]

well...looks like it broke my a1111....now it wont load at all

[grizzled sage]

i had to do a couple restarts with this stuff too. an thennnn presto! OWL

[strange jungle]

Restart of the PC?

OK, maybe iuts installed

trying to figure out how to run it, make it do something

Error

theres a pulldown for unprompted...do I do something with that?

error

error as soon as I add text to the prompt in img2imtg

now I cant do any txt2img, img2img, installing these files has broken SD

I guess Ill delete everything...start from scratch

[strange jungle]

ah heck...even that didn't work. Uninstalled everything. Reinstalled A1111. Didn't even try to get controlnet installed. SD won't generate. Instant error as soon as text is entered into the prompt

"AttributeError: 'NoneType' object has no attribute 'ema_scope'"

even un installed and reinstalled, A1111 now won't generate images

[grizzled sage]

i didn't use unprompted. that looked like a beast

[strange jungle]

No idea

[grizzled sage]

https://github.com/Mikubill/sd-webui-controlnet

[strange jungle]

I have reinstalled A1111 after deleting everything

[grizzled sage]

unprompted might be very awesome, but wow is it a beast

[strange jungle]

Now it generates one blurry square, then after that wont generate anything

[grizzled sage]

😦

[strange jungle]

anything I type in the prompt...instant error

[grizzled sage]

nuke it from orbit. it's the only way to be sure

[strange jungle]

as soon as I enter the first letter

I tried installing on a different machine

[grizzled sage]

even the other machine is corrupted because of controlnet? oh no

[strange jungle]

same problem...instant error as soon as I type the first letter

[grizzled sage]

well i got it working anyway. might play some more in a bit

its easy enough to get it started. it'll take some time for people to figure out how to use it though. deforum people are going to LOVE it

[strange jungle]

Looks like I am done with A1111. This happened before, but reinstalling fixed it. Now...no luck

[grizzled sage]

its important to keep a folder backed up for such new trials. this is why we use venvs

[strange jungle]

I had a backup

It wont work either

as soon as I enter the first character in the text prompt...error

it lets me generate once.,...comes out as a blurry square. Then after that, error

Now it's this when I try to run the new install: i Settings file found, loading
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
→ Updating Settings File
✓ SUCCESS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
i Launcher Version 1.5.8
i Launching without args
i Python 3.10 path found : C:\Users\Frank\AppData\Local\Programs\Python\Python310\python.exe
i Python is in PATH
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
↓ Git not found, downloading & installing, please be patient
✓ SUCCESS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Cancel
venv "D:\Auto1111\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Commit hash: 3715ece0adce7bf7c5e9c5ab3710b2fdc3848f39
Traceback (most recent call last):
File "D:\Auto1111\stable-diffusion-webui\launch.py", line 360, in <module>
prepare_environment()
File "D:\Auto1111\stable-diffusion-webui\launch.py", line 300, in prepare_environment
git_clone(stable_diffusion_repo, repo_dir('stable-diffusion-stability-ai'), "Stable Diffusion", stable_diffusion_commit_hash)
File "D:\Auto1111\stable-diffusion-webui\launch.py", line 151, in git_clone
current_hash = run(f'"{git}" -C "{dir}" rev-parse HEAD', None, f"Couldn't determine {name}'s hash: {commithash}").strip()
File "D:\Auto1111\stable-diffusion-webui\launch.py", line 105, in run
raise RuntimeError(message)
RuntimeError: Couldn't determine Stable Diffusion's hash: 47b6b607fdd31875c9279cd2f4f16b92e4ea958e.
Command: "git" -C "repositories\stable-diffusion-stability-ai" rev-parse HEAD
Error code: 3221225781
stdout: <empty>
stderr: <empty>

Press any key to continue . . .

[dry crow]

you should install it for yourself like just git clone the repo into a new folder

then run the bat

[strange jungle]

what bat?

[dry crow]

the webui-user.bat

[strange jungle]

I cant even get a1111 to run now

[dry crow]

yea you use an all in one installer

[strange jungle]

I have completely deleted all the files...reinstalled

I did

[dry crow]

yea thats the problem

[strange jungle]

deleted everything. Reinstalled from scratch

I dont know how to install except from the all in one

[dry crow]

you need to make a new folder, open it, right click in it, select "open git bash here", then paste:
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
then enter and download the files.

[strange jungle]

no idea what that means

[dry crow]

do you know how to make a new folder in windows?

[strange jungle]

Not really. When I try, the window just closes

[dry crow]

?

[strange jungle]

when I right click, try to create new...the window closes

[dry crow]

for real?

your windows might be realy broken then

[strange jungle]

I got one

there is no "open git bash here"

there is "git gui here" and "git bash here"

[dry crow]

yea then git bash here

[strange jungle]

I did the "git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git". it did something, then ended

[dry crow]

yes then close it

then you should have the stable-diffusion-webui folder

in the folder

[strange jungle]

yes

[dry crow]

there is the webui-user.bat you can run

that should install sd normally

[strange jungle]

ok, running it

thanks for the help. Then I can delete the old A1111 folder?

I will copy models etc

[dry crow]

yea you can copy all the stuff (models, embeds, loras), make sure the new one works before deleting the old one

[strange jungle]

the old one doesnt work

[dry crow]

yea then delete it after copying the stuff

[strange jungle]

eh, it failed

[dry crow]

at what step

[strange jungle]

It said "cant run without a checkpoint" press any key to continue

[dry crow]

yea you have to copy a model in models/stable-diffusion

it needs a model to run xD

[strange jungle]

ah, that's going to be hard

when I right click to copy something, the window closes

[dry crow]

you should later check your windows for broken files, make updates and restart

you could also drag and drop the file into the folder

no need for a right click then

[strange jungle]

got the models in there

generating the first image

blurry square

[dry crow]

which model?

and which resolution

[strange jungle]

512 base ema

512x512

I have to figure out how to get the medvram back in there

[dry crow]

1.5 512?

[strange jungle]

this doesn't open a lainch window like the other does

[dry crow]

right click the webui-user.bat end press edit

then you can add --medvram and --xformers like in my screenshot

[strange jungle]

xformers never worked

[dry crow]

you can try it, maybe it works this time

you want --autolaunch too, it then opens in browser after running the bat

[strange jungle]

ok

[dry crow]

im off for now, any technical questions going into #🤝｜tech-support

[strange jungle]

thanks

[dry crow]

no problem 🙂

[strange jungle]

hopefully they can help to make this generate images now

[grizzled sage]

Control Net!

[grizzled sage]

simpsons matrix... or something

[lyric kelp]

photo of house,bowhouse,modern,concrete,gorgeous,designed,garden,twilight

[strange jungle]

still cant get this controlnet to work

[oak siren]

works ok here

Did you download all the checkpoints??

cfg needs to be 97% ish

[cedar peak]

image for an ev charging station

[grizzled sage]

control net is getting rapid updates. lot of work coming out on this

[strange jungle]

the errors I washaving I read that many people were having the same

[grizzled sage]

I'm of the mind that going forward, stable diffusion without tools like this control net at the wheel, will be a thing of the past. This will be something lawyers recommend as minimum creative input and wapabm qualified

[strange jungle]

causes SD to break until the last model used is deleted...so right now, not usable

control net doesn't run for me...it causes SD to error when trying to generate

[grizzled sage]

it doesn't modify model files at all though. so that's weird. only works in memory

[strange jungle]

the only way to fix it is to reinstall

[grizzled sage]

works for me fine so i dont know.

[strange jungle]

Right,...not sure.

[grizzled sage]

You can be sure it doesn't touch the actual file on the drive

[strange jungle]

But I read the issue - either have to reinstall everything or delete the model

[grizzled sage]

you're just trolling if you're claiming otherwise

where's your source

share it

[strange jungle]

It's what I read, and it was consistent with what happened to me

[grizzled sage]

share it

[strange jungle]

let me see if I can find it

it was on reddit

[grizzled sage]

i think you're sour and you want to spread misiinformation, as usual

[strange jungle]

before I post this, I want your promise to apologize when I prove you wrong

which is about to happen

you're not only wrong, but F'ING wrong

will you apologize?

[grizzled sage]

cry less. show your work

probably not. i'm a giant asshole

[strange jungle]

then no reason to post it

[grizzled sage]

good out

case dismissed

[strange jungle]

not at all

you're not only wrong but f'ing wrong, but posting it does me no good

[grizzled sage]

pebkac confirmed

[strange jungle]

you can't help fix the issue, and you won't admit to being wrong

I want a fix, to be able to run it

like the other person who said "I made a video tutorial"

[grizzled sage]

i dont know how many times or what youve deleted out of weird confirmation bias beliefs. your whole system build is likely teetering on top of a gutted infrastructure fo broken software. my recommendation is to nuke windows and start orbit. from orbit. it's the only way to be sure.

[strange jungle]

the video was the person running though it fast, murmuring under loud, bad background music

i deleted everything

reinstalled from scratch

[grizzled sage]

here you've called the whole system unuseable because of a weird claim that it corrupts models saved in storage

i dont know how i can help you diagnose thhose delusion issues

[strange jungle]

then read about this issue...I only had to delete the model

you cant help with anything

[grizzled sage]

helped myself fine. been using it a lot. it's weird you've beens uch an anchor today whenever other people bring it up

[strange jungle]

Im back to where I was pre control net, and I know how to fix it without reinstalling now, not that I'd run this garbage again before it's ready to go

[grizzled sage]

right now i'm trying to figure out if i can feed the openpose model a custom input

i think, this is kind of a guess, but i think that it's sending a 2d image of teh openpose data to controlnet, rather than any kind of 3d information

[strange jungle]

"level 1
Trentonx94
·
2 days ago

I installed it but it caused some issue where it would not accept any of my prompt even outside of this extension,

I tried to delete its folder from the Extension folder to remove it but now my Webui just doesn't start... I couldn't find any guide on how to remove/uninstalls extensions on google so I wonder if I fucked up big time by just removing the folder or I can still fix this?

(I want to remove my installation of this extension and go back to my webui as it was before: ie, not crashing)"

[grizzled sage]

if i'm right, i just have to screenshot a custom made openpose setup, or any hand tracked openpose image, and voila, custom poses and guided hands

now if only it does inpainting

[strange jungle]

"Dasor
·
2 days ago
· edited 2 days ago

same here, immediatly uninstalled, now everything is working again."

"enternalsaga
·
1 day ago

same problem!!!! crash whole webUI, I had to reinstall."

[grizzled sage]

it's really easy to install. those people all look like they got it working fine and it wasn't corrupting files

[strange jungle]

they didn't

[grizzled sage]

well except that they don't have corrupted model files from the last model they used

yeh

[strange jungle]

they all got SD back to working after removing unprompted

[grizzled sage]

it downloads a bunch so these guys probably thought "it crashed" and force closed mid download and all they really had to do was delete one folder

[strange jungle]

none of them got controlnet working

Same with me. I got SD back to working, A1111, by reinstalling everything

[grizzled sage]

its easy for lots of others. and its not corrupting models too

neat stuff

really exciting tools to see establish themselves so fast

[strange jungle]

but when I tried to run the controlnet with a [controlnet] prompt, it again destroyed SD

I had to restore everything from a backup

[grizzled sage]

oh you're still trying with "unprompted" i think you're having more of those problems that unprompted creates then anything else

that thing does so much more than controlnet. they just added it to their kit

[strange jungle]

even relainching....still the error until remove the extension and restart or reinstall everything - or restore a backup

[grizzled sage]

you're definiately closing mid download

yeah relaunching will fuck that up

bad form

[strange jungle]

I get an error

there's no mid download

it tries to generate one time...gives a blob

[grizzled sage]

because you closed mid download. it doesn't say it's downloading. its just frozen. thats the first time

bro unprompted is YHUUUGE

[strange jungle]

any subsequent generate is an instant error as soon as one character is typed into the prompt

[grizzled sage]

its probably the Hyugest extension. you're just doing things all wrong and trolling, again

[strange jungle]

I closed nothing

[grizzled sage]

i know. i know you've done nothing

[strange jungle]

I wait for the generate to complete

[grizzled sage]

there there

[strange jungle]

it completes to 100%

[grizzled sage]

right rgiht

[strange jungle]

then error every time

[grizzled sage]

ok

its because you didn't let it install fully the first time

dont use unprompted. its way too powerful for you traveller

[strange jungle]

How do you know it's installed fully?

what didn't install fully?

[grizzled sage]

unprompted's shit when you first used it

[strange jungle]

Unprompted shows it's installed...reload the web ui. It's installed

[grizzled sage]

right and then you use the [controlnet] tag, the giveaway that youre using unprompted

[strange jungle]

there's some issue, maybe with low vram cards

[grizzled sage]

since the specific one doesn't doo that. it's a menu item and lets you load control net against any 1.5 model

[strange jungle]

right. Im using unprompted

Is there another way to use controlnet?

[grizzled sage]

and it doesn't come with a billion other tools

i told it to you earlier but you goosed at me

rejecting good advice was a big sign you're trolling

[strange jungle]

in a1111? I saw a video about using it in easy diffusion. That's not what I have

[grizzled sage]

i think your goal here is just to sour it for anyone coming through, so that they won't like controlnet like you're unable to

since you're just pissing lal over anything i recommend

[strange jungle]

I watched three videos - all impossible to follow, too fast, loud background music obscuring the words

I read two step by step tutorials.

[grizzled sage]

right right. all the videos and reading. and yelling at me for not being able to help. i'm sure it's everyone else's fault

[strange jungle]

My goal is to ru fucking n it

I get only a few hours a day of seeing clearly

a few hours a day of able to sit up

I dont have time to waist trolling

[grizzled sage]

5 hours ago bro. 5 hours ago i gave you the better option

[strange jungle]

no you didn't

[grizzled sage]

tell me again you're not wasting time

[strange jungle]

someone mentioned a video which was impossible to follow

[grizzled sage]

you're just making things up at this point. you had the good link right there

and then 5 hours later tell me i can't help you

[strange jungle]

I pasted the steps I followed,. Nobody suggested anything different

[grizzled sage]

pfffff

[strange jungle]

if you had intended to help, you would have 5 hours ago

[grizzled sage]

you probably STILL aren't using the good link, because you dgaf about any of this. you're just here to sour things

[strange jungle]

You're like the damn doctors at the hospital last week...they wouldn't treat my issue..kept insisting I have COVID, then admitted I had tested negative.

[grizzled sage]

i gave the good link to you 5 hours ago bro. you probably have nadda installed and are just roleplaying the entire time

[strange jungle]

I had bronchitis. They would not treat me for it. I had to go to eight doctors before one would prescribe antibiotics

[grizzled sage]

good deflection

have a nice day. court adjourned

[strange jungle]

you're helping me in the same way

I had to go a day being blind before these doctors would give me what I needed...thanfully not permanently blind

you never offered any alternative. You don't have the high road - I knocked you off it.

https://github.com/Mikubill/sd-webui-controlnet I read the simply do - 10,000 steps without explanation of any step. Simply do (these 10,000 things".

How to get a million dollars. Simply get a million dollars, then...

That's what you offered.

Like when I watched a video of how to replace breaks: LOUD MUSIC THROUGHOUT - "Step 1, simply bleed the brake fluid and remove the old breaks" Yeah right.

"Put the ControlNet models (.pt, .pth, .ckpt or .safetensors) inside the sd-webui-controlnet/models folder." SimplyDo. What are those models? Where are they?

3 Press "Refresh models" and select the model you want to use. (If nothing appears, try reload/restart the webui)

Select which model, from where?

Upload your image and select preprocessor, done.

Preprocessor>? From where? Upload image where? What's a preprocessor? What?

SimplyDo

[turbid merlin]

Ti's arent working for me, can someone help?

[strange jungle]

go to thisthatsite.webnet.org. Simply follow the steps. 1 get your TI working. There, that's the step by step.

make sure you've run all the necessary python git nit pulls and that your preprocessors are selected and up to date

[grizzled sage]

don't mind him. which ones?

[turbid merlin]

all ti's by some guy called corneo

[grizzled sage]

lol okay. sex ones. so whats not working? do they show up in the generation details?

they're all trained on NovelAI it looks like so it should work on any model downstream from that

openpose controlnet generated off a pic of a football match. what i want to use is a tool that lets me manually pose and then punch that image into control net

wakanda warriors. this is easy mode now. wow

what a time to be alive!

[smoky oak]

I am so hyped to try all of this out

[grizzled sage]

i'm not good enough yet to do deforum stuff, but i expect we'll be seeing stupid cool deforum animations using controlnet very quickly

from other cool and good people

[smoky oak]

oh yeah, for now I am just trying to take/use some pose references for my characters haha

[grizzled sage]

oh nice. newest update of controlnet extension has openpose for hands. HAND CONTROL

a little weird for a first try but the guidance is totally there and working

[turbid merlin]

does anyone know how i can get the character to be in the middle of the image

not side

[grizzled sage]

closeup portrait, extreme closeup

[slender umbra]

btw anyone knows why it says i'm using 4.4/12gb of gpu memory?stable diffusion running, how can I use more memory?

[static tusk]

with txt2img, increase the batch size. it will generate that number at once, using more vram

[slender umbra]

but i've got a 1050ti

and it already says i'm out of memory when trying to use fix and upscale

[static tusk]

you wanted to use more? but your out of memory I dont understand

[slender umbra]

why does it say I have 12gb of gpu memory when that is not the case?

[static tusk]

the dedicated gpu memory is the real amount

4gb

[slender umbra]

oh ok and the other 8? what are they for?

[static tusk]

its shared with DDR mem

[slender umbra]

Ooooh ok

and is there any way of starting to generate the image

but in a way it doesn't need that much memory?

Like, i think I could download the image and then upscale it, surely there must be a non manual wya of doing it right?

[static tusk]

you can add --medvram to the the lauuch arguments and it will use less, but will be slower

[turbid merlin]

just zoomed it in further

[slender umbra]

I already added it and xformers

i'll try with --lowvram

[turbid merlin]

how do i get it in centre :((

[static tusk]

you can use the extension - ultimate SD upscaler, it generates tiles then stitches them together. you can make a large image without needing more vram

[slender umbra]

omg yes! do you have a linknk or tutorial on how to install?

[static tusk]

https://github.com/Coyote-A/ultimate-upscale-for-automatic1111

[slender umbra]

/givelove pure_fire

[waxen raven]

Trained my face + a style in same model (different tokens) with captions

[slender umbra]

Thank youuuu

[frigid vigil]

Send in the Bacon squad!

[rapid current]

I can now basically turn anything into bone

[grizzled sage]

bones playing soccer

[strange cliff]

help

[grizzled sage]

turn it off then on again

[strange cliff]

you mean PC?

cuz I restart the bat and still didn't work

I have that file in there

I'm looking right at it

[grizzled sage]

delete the entire venv folder and let it rebuild

[strange cliff]

alr

[grizzled sage]

gl

a pc restart might help tho

who knows

delete the venv is a good way to "restart"

i like to get a nice venv built, then save a backup

thye've made so many coo as beans updates to control net webui extension in the last day

the normal map control is unreal. so good

[dense haven]

Any way I can add details to this? Tried img2img but the level of detail stays the same. Also tried upscale like ultimate sd upscale but all they do is refine the lines

[strange cliff]

thanks that worked like a charm

[rapid current]

idk but now it's bone

[grizzled sage]

thats fire hot. i was honestly just hoping. it's sort of a "catch all" fix

[grizzled sage]

control net? so kickass

[novel ravine]

yess

its awesome

[rapid current]

well bart's bone now too

[grizzled sage]

the outline engine.. MMm .. i gotta do that

okay so theres a little known feature in google image search which is PERFECT for this. under tools, type, select line art.

good luck

stupid ai can't even color the sun right /bitter about it kicking my ass at coloring books

[tiny prawn]

#1072220168534642768

[hasty nova]

I've made yet anoth embedding!

"Hero Diffusion"

[karmic tapir]

how to create images

[stone crescent]

"Summoning marilyn manson"

[celest mountain]

so yeah looks like not only SD can hallucinate duplicated heads https://www.youtube.com/watch?v=O62D_tU9xAQ&t=110s

[wispy nest]

https://civitai.com/models/9207/czeros-austin-powers-elizabeth-hurley-as-vanessa-kensington-1997 new Lora just dropped

[wispy nest]

https://civitai.com/models/9207/czeros-austin-powers-elizabeth-hurley-as-vanessa-kensington-1997

[spark spoke]

paco el flaco

/imagen

[wispy nest]

damn 50 d/l already ❤️ ty guys https://civitai.com/models/9207/czeros-austin-powers-elizabeth-hurley-as-vanessa-kensington-1997

[spring steeple]

a cat wearing a suit and tie with green eyes, a stock photo by Hanns Katz, pexels, furry art, stockphoto, creative commons attribution, quantum wavetracing

[wispy nest]

kek

[grizzled sage]

It's fun to stay at the Con Trol Net Yeeah

[dry crow]

@dense tapir pls stop cross posting spam your images

[dense tapir]

NO

[dry crow]

Why...

[dense tapir]

Because people are in each that don't sit in each

[dry crow]

Thats not the point, you would have to spam in every image channel than

[dense tapir]

Should I? Cool, I will.

Of course I jest but still

The problem is this damn community has died. People wandering all over the discords then telling me they can't find help, or find people who even engage with anything anymore. I guess if it isn't 1.5 and titties on fake waifus the community up and died on people. If you know where the community ran off to there are a lot of people who would like to know.

I sat here myself for a week and watched hours and hours of nothing. Hell, I still see a lot of the day nothing. I agree with them this community is POOF.

At best I would tell emad that his damn community is broken at best and pretty dead at worst.

Seems Midjourney has won this battle and that sucks

[dry crow]

You mean discord wise? The people just want free bots it seems

[dense tapir]

I mean community wise. Sad because when 2.0 came out it fractured the community and many wandered off. It is pretty dead art wise or engagement wise

too

A lot of the people I hung out with told me fuck this discord, or this BS so they just stick to themselves now. With no engagement or art discussions they simply left to maybe never come back. I wondered what happened to everyone since the disaster 2.0 hit.

I prefer 2.0 but most don't

[grizzled sage]

i think you're seeing what you want to see. what probably happened was january, holidays for a lot of people, and general commerce is down all around. coming into february is the lowest point of all commerce, and now we're trending up

[dense tapir]

No, I think I know what I have been told and what I see, sure. Please, stop with the asskissing and throw your coloured glasses away.

[grizzled sage]

there was literally JUST the huge stelfie stream that had a massive audience

so much hostility for so much wrong conjecture

[dense tapir]

This has nothing to do with commerce this is free and MJ is not free and all I see is vibrant MJ communities and SD discords are pretty dead.

[grizzled sage]

commerce is commerce. the movement of people

the commons

[dense tapir]

Well, I know what I see, feel, and have been told and nothing you say can change that or will change that. Unless you are the pied piper to bring them all back that is.

Unless the titties come back in 2.2 people will stay gone. Of course I believe in SAI's move on that but still there is a reason 1.5 has the people. I go on discord and I get accosted by people telling me 2.x sucks and is trash/rubbish.

[grizzled sage]

i guess recognizing that this community is booming and growing is just ass kissing oh well

waifu diffusion 1.5 has titties and is on 2x. all sorts of 2x titties are on their way. how are you still on this

the base doesn't need the titties

[dense tapir]

I go on youtube and get accosted tewlling me how much 2.x sucks and is vastly inferior so don't sit there and tell me a goddamn thing to the contrary. It has nothing to do with I am seeing what I wish to see. Fuck that.

Little miss sunshine BS aside.

[grizzled sage]

you want to put hustler mags in public schools and libaries? because thats where base models wanna be

[dense tapir]

Don't strawman me, boy.

[grizzled sage]

well thats the goal of a base model. those locations

how can you avoid that point?

[dense tapir]

Did you not see above where I told you I was with SAI's decision about that? See, you don't pay attention then wish to tell me shit.

[grizzled sage]

it's a simple recognition of facts. no one would expect hustler xxx in public schools or libraries, so a base model that would want to be in those places shouldn't have hustler xxx capabilities

yeah. you make the base sfw and people refine the titties in

crowd sourcing the titties is probably even better

2.2 won't have the hardcore smut in it still. the nsfw filter will have thhe fixes it needed though. that means that 70% of what it blocked, the stuff that was entirely fine, is back in. the goal of the filter, to reduce the smut, will still be dialed in

well they're teaching kids that humanitarian ideas are a good thing, they're not showing them literal xxx pornography in schools. stop qanoning

this is why it's so difficult to explain to people why literal hustler material in an ai model isn't a good idea. the conspiracy about jwish people taking over the world always comes out to play. literal insanity

[grizzled sage]

treating people like they're people is no joke

it's literally the easiest thing you could do in your life. not hard at all

[crimson harbor]

super car

[green plover]

I think what it really boils down to is whether or not the person whos breasts are exposed consented to it

There is the line that shouldn't be crossed

I will be the first person to speak in favor of AI created adult content. If fact, it's my job, but I will also be the first to speak out against nonconsentual porn

[grizzled sage]

it's not "titties" they're taking out. What the goal to remove is literal XXX hustler smut. Conspiracy qanon'ers keep calling it all titties. it's honestly dumb

the model still has girls in swimsuits. i've seen 100s across my batch jobs

[dense tapir]

No, your smut is what exactly?

[grizzled sage]

The brazen open hate towards LGBT is also, very honestly, VERY dumb

[dense tapir]

I swear the same shit coming from you is the same shit from the 1960s and early 1970s towards Playboy that is now considered so damn tame you could almost put it on primetime TV

[grizzled sage]

XXX hustler. this isn't hard to figure out. i've been clear and somehow you two took it towards qanon hate conspiracies

it's as dumb as pizzagate was

bro you brought them up out of no where. you care a lot

the hate is just honestly, so dumb

[dense tapir]

John Podesta. Don't bother to respond to me again as you are seriously one sick fucker.

[grizzled sage]

if not being able to openly hate people that way is what makes this server dumb for you, then yeah, please, go somewhere else

[green plover]

and I am perfectly fine with them taking adult content out of the base model as long as I can still refine against my own dataset

[grizzled sage]

you can refine to the moon

[dense tapir]

Exactly, and legally they really had to.

[green plover]

I know and I'm not the one who was complaining

[dense tapir]

One thing that caught me off guard is when I went to civitai and the first pics were of HARD core sucking an erect full on penis and p shots etc... OMG, I am so glad I wasn't at work. It defaults to NSFW.

[grizzled sage]

i stopped using civit. their users negatively vote brigade all 2x models so they're never found

[static tusk]

one of the reasons I dont create an account there. So I dont have to see it

[grizzled sage]

the entire review culture theri is tainted at the core

i dont see them ever pulling out from the momentum they've got now as the "porno" host

[dense tapir]

Yes, which is why I ceased to release anything

[grizzled sage]

honeslty, i think they're going to be shut down hard by authorities with how many celebrity lora's they're allowing to be hosted

[dense tapir]

Reddit anti ai crowd keeps hitting all my posts to where some have gone neg karma and civitai I know that crowd would hit me. I know someone who does post on civit and yea, bad mojo. He doesn't mind but I sure would.

Seems there is just no place to just post for the masses to use so screw it. I am sure others are feeling the same and I refuse to do full on XXX porn just for the endorphin hits from all the positive reviews. Just not my thing.

Funny that hard porn in the 1960s/1970s Hustler/Playboy is now VERY soft porn if considered porn at all. For me if I see a vagina spread or an erect penis it is porn.

I don't care if others want it just not for me but a site that is so riddled with it and defaults to NSFW with that on the first page? Yeah, I gotta take an issue with that.

SFW should be the default

[grizzled sage]

hustler has always been the magaziine that does full on penetrative xxx pornography. the smuttiest of the smut. do you not know larry flynt?

xxx has very hardcore meaning, always has

[dense tapir]

I only knew of him after he ended up in a wheelchair

[grizzled sage]

i think what you're talking about is how kink play is a trend these days. fad really. 80s coming back again

hedonism is always hedonism

[green plover]

.;l

[dense tapir]

I have some of the 1960s/1970s Hustler mags, well I did until a fire. Naw I see worse now on streaming, or main stream films. Now from about 1978+ it was damn hardcore as you can see when the mags changed.

[grizzled sage]

as a philosophical movement i mean. it's just always the same trends. all the time. thorugh the ages. moms. sisters. strangers. domination. submission. etc.. it's all been done

yeh larry flynt showed up mid 70s lol

started hustler magazine

changed it all

[dense tapir]

Oh, wait my bad I had Penthouse and Hustler mixed. Yeah Penthouse, and Playboy was tame Hustler was always a dirty rag as Larry did that to push the envelope.

[grizzled sage]

i'm saaayin

[wispy nest]

new Lora, going to release it tommorow, want to make sure its varied enough first, input was only woman, loraname:1 , no negative

[grizzled sage]

thank you. finally a bridge crossed.

[dense tapir]

Larry got off on all the puritans bitch, and moaning, and the lawsuits. I am not joking as he even said that in one of his interviews. Dude lived for that.

Thing is one of those "righteous with God type puritans" is who shot him and made him in a wheelchair for the rest of his life.

1978 I believe

[grizzled sage]

All the single legos all the single legos all the single legos all the single legos

[grizzled sage]

generic dancing characters even look so fine with this control pose

these images have MOTION

2 is dead in the water for adoption until it gets a controlnet i'm sayin

[fiery zinc]

@reef root This is what's in my directory after installing the extension:

[reef root]

You're putting it in models or extensions

It just doesnt exist

[grizzled sage]

wrong folder

[reef root]

It's the only folder

[fiery zinc]

The models go in models under extension\sd-webui-controlnet\

[grizzled sage]

i dont know why you've got two installs for 2.1 and 1.5 models. they're the same install

[reef root]

Right but there is no folder named models under that file extension

[grizzled sage]

you installed it on the other install for 2. but not there

because for some reason you're using 2

[reef root]

I upload a virtual machine, when I do so, it asks which one I want to load : 2.1 or 1.5

[grizzled sage]

sounds like a gimmick, a con, a scam, a shimdimeroo

[reef root]

What?

It's been working for weeks

[fiery zinc]

ok, I think I understand

[grizzled sage]

i'm going to chalk this up to their gimmicky sales implementation theres no reason taht service should be charging twice for that

[fiery zinc]

RunPod uses docker images, if I remember, it comes preloaded with a model

so yes, in a way, he's getting charged twice

they dont' share the same code, they prepackage 1.5 and 2.1

so they're two separate instances

[reef root]

They aren't charging twice. The only time I've ever ran two models at the same time was to confirm my findings today

[reef root]

Yes

[fiery zinc]

yes, but that's why you don't see it in 1.5

because you only installed it in 2.1 image

you have to do the same process again, but now in the 1.5 image

[reef root]

No I re install it

[fiery zinc]

they're two separate docker instances

[reef root]

Every time I've tried it

I've re uploaded the file every time I've tried

But there is nowhere to place said file

I can install the extension no problem

I do that every time too

Thanks for yalls help

[regal belfry]

i see some artists using "delicate, masterpiece, beautiful detailed, colourful, finely detailed, intricate details, (1 girl)1.1, solo, (from front)1.3, looking at viewer, (a beautiful 19 age olds cute Korean girl)1.3, instagram, (kpop idol, korean mixed)+, (blonde hair + short hair)1.3, (50mm Sigma f/1.4 ZEISS lens, F1.4, 1/800s, ISO 100, photograpy)1.1, (medium breast)1, (photorealistic)0.8, (Arcueid Brunestud, tsukihime, red eyes, white turtleneck + navy mini skirt)1.3, (beaming smile)1.2, (moon, buckwheat field, thigh)1.2" as their prompts in the metadata
is there a way to tell what model or what ai were they using

[fiery zinc]

not from the prompt, no

[reef root]

So am I SOL?

[grizzled sage]

once you isntall the extension on 1.5 the folder is literally it's install location

[fiery zinc]

I'm sorry, it just seems broken

[grizzled sage]

if its not there its not intstalled. unless they're doing something sneaky

[regal belfry]

whats all these weights besides the prompts

[reef root]

I install it with the webui

There is no folder browser within the web ui

[regal belfry]

is this something SD does?

let me rephrase

[grizzled sage]

instructions right?

[fiery zinc]

different UI have different ways to add weights

[grizzled sage]

https://github.com/Mikubill/sd-webui-controlnet

[regal belfry]

does the above look like a stable diffusion prompt

[grizzled sage]

Open "Extensions" tab.
Open "Install from URL" tab in the tab.
Enter URL of this repo to "URL for extension's git repository".
Press "Install" button.
Reload/Restart Web UI.

Usage

Put the ControlNet models (.pt, .pth, .ckpt or .safetensors) inside the sd-webui-controlnet/models folder.
Open "txt2img" or "img2img" tab, write your prompts.
Press "Refresh models" and select the model you want to use. (If nothing appears, try reload/restart the webui)
Upload your image and select preprocessor, done.

[fiery zinc]

I dont' recognize that one, but assuming is similar to A1111, then those numbers are multipliers

so "promptA" is 1, (promptB)1.2 is 1.2, and so on

is my guess

[reef root]

I've done this exact pattern of steps, thank you very much for typing it out, but there simply is no folder titled "models" in that location

[grizzled sage]

when you install it, there is that location bro

[reef root]

I just restarted my entire computer and redid the entire thing

And here's a screenshot

[regal belfry]

so it is stable diffusion compatible is what u are saying

[fiery zinc]

probably, but maybe from a different UI

[regal belfry]

cool

[fiery zinc]

in Automatic1111, it would be something like (promptB:1.2)

different syntax

[grizzled sage]

wrong folder

you know that too. i think you're trolling

[fast quest]

#1072220168534642768 sohar castle