🏞｜general-with-images

[grizzled sage]

i want 24 but none were in stock anywhere. i settled for a 16 4080. i wonder where a 4070 would perform? still better than a 3090?

[dense tapir]

See, I was saving up for the leak of what the 4060 was supposed to be which was 7k cuda, 379MSRP and 16GB. None of that came true

Still saving though

If the 4090 were 1200-1300 I would have purchased it.

[static tusk]

4070ti is better than 3090, not sure what the non ti will have spec wise

[grizzled sage]

i started saving up when unreal engine 5 started showing off. it took me that long to see prices be reasonable. 4080 was actually selling for MSRP, surprisingly

[dense tapir]

See, I don't play the copium game so inflation adjusted and tech released the 4080 should be a 799 card

4090 1299

maybe 1399

[grizzled sage]

oh. you're insulting my buying decision because of youtube memes. intelligent conversaion

[dense tapir]

I am just telling you how my money works

[static tusk]

4070 is cutdown in specs from the ti version

maybe on par with the 3090

[dense tapir]

4070 with 16gb would be a good card if priced right. Sadly, I don't think they are capable of doing it

You know I seriously think that part of the problem is that Nvidia just has far too many models. Go back to what they had with the GTX1k series before all the creep, and madness started.

x060 ti still gets me to giggle and smh

[static tusk]

wait till they release the "super" variants lol

[dense tapir]

back when the 80 was the big dog

yeah, those too

[grizzled sage]

that'll be my next upgrade i think

[dense tapir]

although I always purchased the SSC models from evga because they were made to natively o/c so made a tad better and I ran them sub (or basically at the norm ranges).

my biggest issue is that there are currently no cards made for nvidia that I like the companies that make them. EVGA is gone now after 22 years so wtf? If you need 3 fans it is because you cut corners on the heatsink

the heatsink on the 080 and 090 is the most expense for Nvidia I read.

very much over engineered for the gpus we have yet to get

yet is around the corner with the 4090ti as I bet the last 16% of the die will be used for it.

It will rock when I can finally kiss colab goodbye and train locally.

Look at these pics as I see like tentacles in them and I never called them.

Almost nightmarish like

[static tusk]

definitely sad that evga pulled out of the GPU market. they were a solid company to trust

with excellent warranty

[dense tapir]

Yes, and I have no one else that I do so, for me, it would have to be Nvidia direct

everything I own, or have owned, is evga

[static tusk]

Ive used Asus for a little while now. no problems

[dense tapir]

Well, 3 fans makes me say no to them

also I was part of those who had a mobo from ASUS and their chinese knockoff caps blew as it did on millions.

[dense haven]

Doesn't the | symbol supposed to separate the generation?

[static tusk]

add [] around it

[dense tapir]

luckily mine blew after I ceased to use it and one day I tried to turn it on and no go. Looked inside and pop, pop, pop they all had just blown sitting there not even plugged in.

[static tusk]

wow. I havent had that luckily

[dense tapir]

I feel sorrow for the board partners with Nvidia and I understand exactly why evga left

[dense haven]

still not working

[dense tapir]

5-10USD per gpu sold by them so they have to cut corners like 3 fans instead of a beefy heatsink

[amber pumice]

Hello Everyone, I was wondering if I could get some advise on to why am I getting this results when upscaling, it takes 20 - 30 minutes to destroy my image

[static tusk]

I havent actually tried with 3 options. I normally only use 2 to blend different cars. not simply colours

[dense tapir]

3 shouldn't work

[grizzled sage]

i dont really feel sorry for nvidia board partners. i'm pretty sure its breaching new markets AGAIN. ML investments have paid off big time

and will continue to

[dense tapir]

I do

[dense tapir]

I do because when they make so little all they end up doing is fucking us

[dense haven]

still not working

[grizzled sage]

board partners have a bright future ahead of them and are probably making killer profits. they raked it in over the mining years. i'm still wondering what the real reasons evga threw the towel in were. we all heard the pr drivel. oh well. as usual, i don't buy it

[dense tapir]

Yes it is

[dense haven]

It's blending instead of doing red,green,red,green

[dense tapir]

that is what | is

[grizzled sage]

thats the results of it. Each step, it does red, then the next step replaces that with green

[dense tapir]

1 fram X then next fram Y then 1 fram X

[dense haven]

Oh eacg STEP

[grizzled sage]

[red:green:0.5] would do half the step count red, then the rest green

[dense tapir]

[man|woman] for 20 frames should become a trans

[dense haven]

Might as well prompt without the separator

[static tusk]

I use it with bmw|ferrari, for example

[dense tapir]

yeah, pure

[grizzled sage]

i use it with "[terminator|hulk]

[dense tapir]

LOL

lemme see

[grizzled sage]

fuck that's digging. reminds me i've been testing galleries and breadboard is good, but kind of sucks

bugs out ALOT

[dense tapir]

breadboard is that image thing of everything you have so you can sort them?

[static tusk]

yeah image tag and search

[dense tapir]

yeah, I fled that

Luminara I think it was called had that and I hated it

[static tusk]

I just name folders with the styles I use and subfolders

[dense tapir]

same principle

what kills it for me (back to gpus) is that I see partners charging 100 more than msrp. Look at their product and shrug, then I ask myself why?

thanks to mining everyone is still super greedy wanting those "I don't care what it costs I will buy the entire pallet" days back.

Clyde just blocked me

[static tusk]

with Ai taking off it may be a similar situation again as mining

[dense tapir]

Yes. In a way I hope it all collapses

[grizzled sage]

the hulkinator isn't in any of my saerchable prompt folders. it's a deep dig and i'm wonderiing if maybe it didn't survive the data nukening

judgement day

[dense tapir]

that is funny

[grizzled sage]

as i'm digging deep too, i'm thinking i need to throw another atomic salvo at some of this data

[dense tapir]

no effing way is the 4060 worth 600usd and street price 700.

4060 is only 8gb ram and 192 buss I believe. really bad

honestly look at the 4070ti and that is what the original 4060 was supposed to be.

AMD is sucking hard now too

I told people what they were doing back in 2020/2021 and I got jumped on then they just admitted to it

[grizzled sage]

they were a paper launch in canada. rdna3 may be a last hurrah for amd dgpus imo

[dense tapir]

they admitted to holding back stock to make the prices high.

[static tusk]

the Radeon team put their head in the sand for sure

[grizzled sage]

greasy

[dense tapir]

yes

[static tusk]

I think both amd and nvidia said they under shipped

[dense tapir]

As much as I despise Intel I was rooting for them and ARC. Price is great but damn for the rest. Not horrible but damn

yes

I will say if my 1060 blew I would buy an arc

[grizzled sage]

shrekinator

[dense tapir]

eyes on the chin, LOL

[grizzled sage]

i'm imagining it's his zany plan to sneak into central skynet

[static tusk]

check the mid riff

[dense tapir]

OMG

He was so intently staring at the eyes chin I just snuck in

I can say I am very pleased with this lora I trained. It took 2 days and a ton of work but I learned a few things. Apparently for me I must train for 35m-1h to get what a 3090 will in 10m

[dense tapir]

wth is that?

monkeynator?

[grizzled sage]

predator termiinator

[dense tapir]

lol, shit

See, RDNA4 is supposed to come with a ML chip, and a new CPU board has one. AMD is jumping into this now. Nvidia knows this but I firmly feel they are working together ELSE it would be like the old days where one see an opening to grab marketshare.

Heck, AMD could do it but refuses

[grizzled sage]

nice. i'd imagine it'd be hard to cybernetically enhance hulk gamma tissue though. it'd just crush any non gamma organic skeleton or machinery

would the metal endoskeleton have to be uranium or some?

[dense tapir]

It would take some new form of self healing faster than it can be damaged type nanite metal

[grizzled sage]

gammites

[dense tapir]

possibly

[grizzled sage]

with quantum tunnelling

[grizzled sage]

it would def have to be like an abomination type skeleton that extended out as an exoform, just like that

[dense tapir]

yep

[grizzled sage]

to help contain the power

[dense tapir]

gama radiation is powerful so yeah

*gamma

[grizzled sage]

https://tenor.com/view/hulk-endgame-snap-gif-16000400

[grizzled sage]

wimpy banner hulk

[grizzled sage]

the wulk?

[static tusk]

vampire hulk

lol

[grizzled sage]

ah yeah the eyeliner. i see it now

[static tusk]

hulk and arnie blended

[green plover]

My first successful character TI. Which started out as a prompt edit. This person has a name and doesn't even exist 🙂

[static tusk]

hulk - wonder woman. Im scared

[gleaming flame]

Any tips how i can make myself stand next to cristiano ronaldo on stable diffusion?

[static tusk]

controlnet makes it so much easier to tweak when the subject isnt moving about

[gleaming flame]

I always get ronaldo face x2

[green plover]

There's a couple of ways to do it. You can inpaint them in on an existing image or you can train yourself into the model and generate a new picture

[static tusk]

much easier to inpaint

[pallid bear]

humanitarian documentary portrait photography in Kenya, natural lighting, happy family

[smoky oak]

you could do some hacking

Generate yourself, then him, then the background, then composite them together

or you could utilize open pose, pose two guys (get his face on both), and then in paint one of the faces to be yours

[green plover]

"It's all over and I'm standing pretty In this (dust) that was a (city), If I could find a souvenir Just to prove the world was here. And here is a red balloon I think of you, and let it go.."

[river patrol]

A dog in the air

[wispy nest]

@native granite

Best I imagine is - images below

But idk how realistic my wish is relatively to SD models

[frozen raven]

What model did you use and what prompt?

[native granite]

I think these are screenshots from Final Fantasy 15

[native granite]

I think it can be done. There might already be some fine-tuned version on this style, even. If not, it could help to finetune on it

[wispy nest]

It's not SD, I'm looking for a model that capable of doing magic , but not anime...
That was an example of kinda what I'm I'm thinking of.

[wispy nest]

I have no idea how to do it and doubt my 1050ti capable of training \ finetuning.
i'm using kaggle or google collab usually to generate , which are too limited to do training too.

[native granite]

Well there might be models aready fine-tuned to Funal Fantasy 15, and maybe even some models that know of Final Fantasy 15 or similar styles

[rancid juniper]

Hey automatic 1111 users, I deleted my entire google drive.

im trying to use the Automatic 1111 webui to use SD

I have this one on my PC. Uploaded to the drive where it was before.

how to get this working again?

should i download anything else?

[stoic pivot]

Some time SD 5.1 scare me, I often need to write advance prompt with negative prompts to get something to work, other time it do this at the first try for the day with a simple prompt. -"mystical ghost girl play in a fine detailed palace room, smiling, Sweden".

[rancid juniper]

Can anyone help me?

[dense haven]

Can I add details in this image? It got blurry after img2img. Ultimate SD isn't helping. Using Anything v3 with baked in VAE

[wispy nest]

I'd suggest using img2img with low noise , but ...if it got blurry after img2img idk , how it even got so blurry?

Another option would be to upscale

or upscale - resize down - upscale again for a bit better result in a more steps

[dense haven]

Img2img blurs details in general. As if it's scared to be wrong haha

I'll try that

[wispy nest]

It doesn't for me

only on the edges where I inpaint a bit

[dense haven]

Well not dramatically

[wispy nest]

This looks kinda dramatically tho , there's something wrong

[dense haven]

The scale down and then up trick didn't work

[wispy nest]

it's opposite.
You upscale to add mroe details - then resize down - it keeps some details , basically making better version of an image you had before upscaling

not alot better, bit still

and from that point you can upscale again

[dense haven]

Oh I pasted wrong image

[grizzled sage]

the art style there reminds me of the marvel trading caards i used to collect as a kid

[grizzled sage]

Control net

[dense haven]

Scaled down and Ultimate SD'd back up

[dense haven]

Hm ok

[wispy nest]

other way around...

upscale first, resize down after

or don't downscale at all

[delicate wharf]

the open pose stopped working for me suddenly.. don't know why

or I am doing something wrong

[steel yarrow]

hair dye box photos are here

[frozen nebula]

Deliberate can be a picky model but when it slaps it really does slap... got this straight from a prompt no fiddling around or upscaling, just raw

[wispy nest]

interesting prompt idea

[lilac pilot]

Sweet

Vulcan salute not working

Might need to separate the fingers a little more

[smoky oak]

Oh my god hes hot lmao-

Currently using SD to refine a friends art

she asked for me to help refine it a bit, so I have been, and its been teaching me how to better img to img

[coral remnant]

👀

[dry crow]

Can you scroll down at the bottom is the gradio Version

[boreal widget]

there isn't one to show

this happened last year, around oct,

[dry crow]

I think the easiest way is to make a fresh install

[boreal widget]

I did

what you do is delete venv, repositories, and config.

but if a complete delete is in order, I suppose taht was coming

[dry crow]

No i mean a complete new install

[boreal widget]

just download a manual download, which I trust more anyway

[dry crow]

Make a new folder, call it SDAI, then go into it and make a right click, choose "git bash here"
Then copy this:

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

Paste it with a right click and hit enter

Then you can run the webui-user.bat and it should work without problems

[boreal widget]

so just do a git clone in a new dir, right

here goes

[onyx abyss]

thoughts?

[smoky oak]

goofy braids, but the rest is good

[onyx abyss]

what hairstyle should i try? i wanted to pick something so the photos would be consistent

[smoky oak]

the karen bob cut lol

[dry crow]

Looks good

[onyx abyss]

lmfao, i'll try it.

something about my prompt is sjust making it look like the chick from pulp fiction

btw what does sampling steps do?

[dense haven]

Hi. I've been struggling with lost details on img2img and also controlnet. Is there any way to add back or have SD reimagine details to my image? Ultimate upscale not helping either

[dense tapir]

what is your cfg at?

[dense haven]

7. Why?

[stoic pivot]

prompt is "vibrant, colorful, forest landscape", what shall I add (or remove) now? I want to see what this goes. (fixed seed: 320386984)

[dense tapir]

img2img I normally use cfg 13 and how much change you are after determines the denoise strength.

13 at .36 I like a lot or zap up to .75 if I want a total change.

[dry crow]

Falling feathers 😄

[stoic pivot]

vibrant, colorful, forest landscape (Falling feathers)

[dry crow]

Remove forest and add intricate

[stoic pivot]

vibrant, colorful, intricate landscape (Falling feathers)

That one turned out a bit odd. shall I remove fixed seed?

[dry crow]

Yea maybe ^^

[stoic pivot]

Prompt: "vibrant, colorful, intricate landscape falling feathers". With (Euler a)

[dry crow]

Maybe landscape:1.3

[stoic pivot]

vibrant, colorful, intricate landscape falling feathers

vibrant, colorful, intricate (landscape:1.3) falling feathers

[dry crow]

Okay the feathers are to intense haha

[dense haven]

Well I don’t mind change. It’s details I’m after. Just wondering if there are features that allows creation of details

I wonder if the fault lies in my input image having just enough details for sd to not add more details

[stoic pivot]

Feather landscape

[dense haven]

I mean I see people generating great images from pencil sketches

[dense tapir]

How is your prompt? Many factors in this.

[stoic pivot]

No Luke, I'm your fether.

[dense haven]

Well it’s a shopping mall with intricate details and 8k and all that

[dense tapir]

Well, SD will only add so much especially 1.5 or less

you have to be specific with it and bring a sledge hammer

[stoic pivot]

Updated prompt to: a vibrant and colorful intricate (landscape:1.3) with falling feathers

[dry crow]

Ouh nice

The landscape is there

You can try set the resolution to 768x512

More landscape mode

[stoic pivot]

768x512, a vibrant and colorful intricate (landscape:1.3) with falling feathers

[edgy patio]

wallpaper

[stoic pivot]

I have no idea why it skipped the landscape

[dense haven]

Ah. 2.1 then.
with controlnet, is there anyway I can set the dimensions to greater than 1024?

[dry crow]

Try adding sun glare, that should bring us back to a lamdscape

[dense tapir]

I used to could max even with 6gb but not since the big Jan update can I.

[stoic pivot]

a vibrant and colorful intricate (landscape:1.3) with falling feathers sun glare

Nice tips

[dry crow]

Looks like a painting

[stoic pivot]

Small upscale

[dense haven]

I have 24 GB vram but my sliders still max at 1024

[dense tapir]

my sliders go to 2048

[dry crow]

Nice, try with Mountain view

[dense tapir]

I can't do that now as I OOM

[dry crow]

You can edit a config file to change the sliders i think its config_ui

[dense tapir]

I never had to for that. Weird

[stoic pivot]

a vibrant and colorful intricate (landscape:1.3) with falling feathers sun glare mountain view

I manage to upload wrong image first time.

[dry crow]

Yea for me its 2048 too

[stoic pivot]

I stop my uploads to not interrupt more important chat.

[stoic pivot]

2048x512 Nvidia 3090

[dry crow]

Wow nice

[fierce parrot]

#🏞｜general-with-images An astronaut Teddy bears A bowl of soup

[dry crow]

@fierce parrot in this dc are no bots

[stoic pivot]

I'm a bot now.

[dry crow]

Im trying that later today

[dry crow]

Upscaled could be a nice wallpaper

[stoic pivot]

I made this as a wallpaper for me,

The prompt was "A dark forest in Candyland with multi-colored (glowing) [gummy]-mushrooms deeply rooted in the (mossy) forest floor, (sparkling) glitter, 8k, master 3D render by PIXAR, ((close to the ground)) (close up), 12 mm, PIXAR fantast look, neon fairy dust glitter, render, tiny flowers around the stem of the mushrooms, inspired by the movie TROLLS, fantasy surrealism, bloom, bokeh, lens flare, night shot"

typos and all.

[dry crow]

Wow looks very cool with the bokeh in the back too

Can you try a miniatur tilt shift train landscape?

With a Locomotive

[stoic pivot]

One more coming soon, then I have to run off

[dry crow]

Very cool thanks!

[stoic pivot]

a miniature tilt shift train landscape with multi-colored (glowing) old locomotive pulling carts through a (colorful) forest floor, (sparkling) glitter, 8k, master 3D render by PIXAR, ((close to the ground)) (close up), 12 mm, PIXAR fantast look, neon fairy dust glitter, render, tiny flowers around the stem of the mushrooms, inspired by the movie TROLLS, fantasy surrealism, bloom, bokeh, lens flare, night shot

[dry crow]

Awesome 😄

[stoic pivot]

I keep the "fantast", if I remove it I get a look I do not like, it is a typo that do good for me.

[dry crow]

Okay did it change so much?

[stoic pivot]

Kind of, it often miss things and make doubles, when I changed it to fantasy I lost the locomotive.

[dry crow]

Ah okay xD

[stoic pivot]

It like my SD like "fantast" 😄

Added "fantast" again and got what I wanted 😄

Okay, gotta run, have fun now.

[dry crow]

Haha nice cya ^^

[dense haven]

Can't find it. even tried the controlnet yaml files

[glass grove]

#1072220168534642768 https://cdn.discordapp.com/attachments/1004159122335354970/1077521572937605170/IMG_20230221_155516.jpg enhance in 4k hd portrait in African dress

[onyx abyss]

You guys are so creative at this and I’m over here just trying to make hot chicks lol

[delicate scarab]

This is super. What other setting you are using?

[floral stump]

Can relate

[strange cliff]

so what make YOU happy?

[floral stump]

Sunsets for example

[strange cliff]

I see you find solace in the dark you little twilight creature you

[green plover]

I managed to coax Isis out of SD

[celest geyser]

I meant to try Atomic Heart tonight, damn you SD

[wispy nest]

#🏞｜general-with-images

[strange cliff]

👀

wut

y pizza got deleted

[green plover]

before regeneration

after regeneration

[peak aurora]

i can see the difference

[green plover]

my denoising strength is between 0.25 and 0.35 for that

[peak aurora]

even the facial features became better after regeration

[green plover]

Yes. and it makes a huge difference even if the face is blured from a bad generation

Because it can grab the shape and use it as guidance

[peak aurora]

got you.

[cyan narwhal]

So i've run a xy plot with different samplers on Protogen2.2 model. And got really weird results (also, it's 2023, why it such a pain to load ~80mb picture anywhere?)

like here, what is wrong with 15steps euler? it's competently different image and then it's the same

or this. like >100 steps more similar to 15-20 steps than everything in between

[wispy nest]

It's different samplers , they are supposed to be different

5 steps is just too low

[wispy nest]

nothing wrong , different amount steps can produce different result.

[sharp nexus]

#🏞｜general-with-images privet

[stoic pivot]

The horror of take a random furry prompt for a furry model and run it through Deliberate v1.1.

[steel yarrow]

better head off QowdWest

[pallid ruin]

It looks so good 🤗 😊

[grizzled sage]

some computing limits remain hard limits for a long while. Bandwidth is a big one we're all still dealing with. The main bus can only fit so much.

[hollow grotto]

all thanks to people moving to cellphones, and isp monopolies

[grizzled sage]

i dont think its' that. it's physical architecture limitations that there's not many ways to improve upon. its sort of like how an axel can only carry so much weight no matter how much you engineer and develop those thigns. steel can only be so strong. so big rigs have needed 12 axels for heavy loads all this time

adoptions of cellphones have probably accelerated miniaturization r&d if anything

and isps deal with internet bandwidht. i'm talking about mainbus system bandwidth

[fiery zinc]

@strange jungle

[wispy nest]

Where i can make pictures like this?

[hollow grotto]

#1019361238234443776 ?

[fiery zinc]

@fiery carbon batch = 10, img per batch = 1, CFG = 7, realisticVisionV13, controlnet = canny with default settings, negative prompt (nude, naked, nsfw: 1.3). No apparent nudity drift

[sterile kiln]

https://civitai.com/ is down?

[frigid gorge]

Works for me

[sterile kiln]

infinite loading

[wispy nest]

it has bad times for about a week now...

[sterile kiln]

Hmm fixed by emptying cookies... On my 2 webbrowser, weird...

[fiery carbon]

Ok, i am trying to keep the clothing consistent.

[stoic pivot]

I made this with "AnythingFurry_v1" and at one point it did go kinky even "NSFW" in negative prompt, so it there ang good "safe" furry model out there? Just now I have hard time getting into https://civitai.com/ to search.

[left crane]

Stable diffusion text-to-video is finally hosted on a web app:

https://app.kaiber.ai/invite/375d44e6-86ba-43f3-bdce-c49e0c45555b

Example follows: https://www.instagram.com/reel/Co8F4n2AD1_/?igshid=OTJlNzQ0NWM=

[summer berry]

Zoro kid

Messege zoro kid

[fiery carbon]

I am waiting for the masterpiece

[summer berry]

How I do text to image

[frozen nebula]

Just some goth / fetish style images I managed to generate recently

[fiery carbon]

I recommend installing sd-webui if your computer is capable.
Otherwise, use a chatbot.

[charred siren]

Hey everyone, I've created an implementation of ControlNet. It also takes your fine-tuned mode as an url (.ckpt urls from HuggingFace works) It's still WIP but I wanted to share since it might be useful for someone.
You can run it here: https://replicate.com/okaris/controlnet
Sample:
base_model: "https://huggingface.co/andite/anything-v4.0/resolve/main/anything-v4.5-pruned.ckpt"
controlnet_model: "Canny"
prompt: "soul reaper with a flaming sword"
Also make sure you checkout my side project https://avtrs.ai where you can train a dreambooth with your photos and generate avatars for free!

[cursive lance]

y lo siento para los que no hablen español jaja

[grizzled sage]

expect everything on civit to have smut in it. the smutty model authors didn't want to be "censored" and made a ton of stink about it, so if you turn off nsfw in your preferences, it only hides nsfw image generation examples, not the models that are capable of them. it won't tell you those models are nsfw capable either. this way, it does no harm to authors? ugh

[hollow grotto]

yeah i like the idea of civit but it seems like it's 80% porn models

[grizzled sage]

the only way to know if a model is nsfw is if you leave the option on and look otu for nsfw generation examples

[fiery lichen]

yea, put nsfw into negative is fine, the models are still good with other content

[grizzled sage]

"it''s fine if you just..." isn't the point. some people want to avoid it all together

if 'm putting a model into a demo space, i don't want it to throw buttholes out when it gets confused

[fiery lichen]

i did not read the earlier posts, I don't mind stumbling upon a generated nsfw image, i just go nsfw into negative when i see one\

i realise the case maybe different for others

[grizzled sage]

not only just the smutty merges, but that place is just flooded with a lot of content that's not appealing to me. signal to noise ratio is really poor and hard to filter there

[fiery lichen]

i only know to look for models on civit and huggingface. sometimes huggingface model makers dont show any image/prompt so i dont know what they do. civit is best for showcasing

i look for models on this discord too

[grizzled sage]

yeah there's absolutely a need that's not filled. people who legitimately want pg models without needing to wade through the elysium fields of smut

[fiery lichen]

thats usually how this type of tech gets pushed in the early stages...look at VR (recent and in the 90s and 2000s, smut were at the forefront with most content)

[grizzled sage]

i was there. i dont think smut drove vr much. there was a fad there wiith cardboard for a bit but that was gimmicky more than innovative. videogames actaully get all the credit there imo. john carmack and valve engineers

i name john cause i know his name

[torn venture]

now I'm getting scared xD

[fiery lichen]

I mixed up vr and 3d... late 90s-early 2000s, a friend developed 3d glasses for TV with wires to a box, most content were smut movies, regular movie folks were not too eager to jump into 3d at that point

[minor kiln]

breagink bad!

[grizzled sage]

nice with the toon/photo aesthetic!!

a little furrish but... you know.. i'm cool with furry art if it don't get too weird on me

[minor kiln]

yeah that was the point lol

[strange jungle]

Note the extension is not a continuation of what's above

[wispy nest]

what extention is it?

[strange jungle]

the box I drew

the new box I clicked with open outpaint

I clicked a new box that extends one square below the old

[wispy nest]

just try invoke ai, it does okay-ish, doesn't take long to install

or https://www.painthua.com/

or sd outpainting in a1111 , it's not bad , just needs few attempts

[static tusk]

From a colouring book pic to digital art. Who would have ever thought this would be so easy

[onyx abyss]

thoughts?

[onyx abyss]

Finally was able to get some texture on the face

[wispy nest]

there's something about the left side of the nose that doesn't feel right, but great work on the textures for sure there

[indigo rover]

ControlNet is pretty cool

[onyx abyss]

Ah, I see that now. Maybe I need to add “symmetrical nose” to the prompt.

[raven kiln]

Wow

[cursive lance]

Wow tris is beautifull... Is magnific and perfect

Is a master piece

[static tusk]

that was made using controlnet. thank you

[swift skiff]

Realistic raiden using controlnet

[dense haven]

Why do I lose details like this?

Oh, my prompt was bad

[grizzled sage]

nice!

[wispy nest]

out of the recently hot videogames that follows, any of those would deserve a LORA or a full finetune you think ?
Hogwarts Legacy
Atomic Heart
Wild Hearts

I feel like all 3 have some kind of graphical patern that could be nice, but also that Ijust should maybe go back to monster hunter world to make a monster model. I'm not sure, and trying to find a project to motivate me to get back to the training scene

[cyan axle]

This has just been incredible being able to generate such realistic images

Even within a fantasy theme.

[wispy nest]

is that zelda ?

[cyan axle]

Nah

used things like "High elf" as prompts

I play an old af MMO called EverQuest, so was trying to create Firiona Vie, an NPC character from that game.

[wispy nest]

felt like it sorry, I spent too much time with wavymuler that loves her

I loved everquest

[cyan axle]

I still love it. 🙂

[wispy nest]

I tried making a full DND finetune, using quite some everquest graphics

I failed though

[cyan axle]

They have time locked progression servers now, where it starts in the classic era, then every like 3 months, it adds an expansion chronologically to the server.
Makes for a great expedited trip down memory lane.

I wish I could force the heterochromia it did in this image, but it doesn't seem to understand the prompt. Just happened on this one.

[wispy nest]

I had to promise myself to stay away from MMOs at one point of my life tbh. I went on the "no life" path for too much time, I know I can't control myself on that side, I only play single player game now ^^ but it was one of my loves of that time

[cyan axle]

100%, totally get that heh. I had to do the same for a time when I had kids, but now the kiddos are a lot older, and I'm actually getting to play EQ with my daughter. It's been a ton of fun. 🙂

[wispy nest]

I'm still nostalgic of it but I know I can't delve on it too much either 😉 happy for you though. would it have been at another point in my life, I'm not sure I would have reacted the same to it

[cyan axle]

What are the rules on posting a bikini clad (fully clothed, but it's still a bikini) image here?

[grizzled sage]

RIP whatever that new everquest was supposed to be. landmark

Everybody loved to hate on smedley, but that studio went to hell as they put a tighter chain on him and then kicked him out. smedley is the reason old SOE kicked so much ass

[cyan axle]

Yah, EverQuest Next was a huge letdown to a ton of people. 😦

Landmark was just basically testing their tools for creating content, and putting those tools in the hands of the players. Which for what it was, was actually pretty damn cool

[grizzled sage]

Yeah landmark was the sandbox. it was nifty

[cyan axle]

Yah, it was def cool for what it was. It just wasn't what people "thought" that it was.

[grizzled sage]

And then i guess the goal was the wide everquest world would be built from featured landmark content

[wispy nest]

sorry I got pinged somewhere

[cyan axle]

People "thought" they were getting the EQ 3 pre-alpha content, but they were just getting dev tools.

[wispy nest]

rules on NSFW are quite easily sumurised through

[grizzled sage]

It was doing stuff with terrain editing nobody else has done since

[cyan axle]

Exactly!

That voxel based terrain editor was amazing

[grizzled sage]

maybe its been done since but not on that sort of brand scale

[wispy nest]

if you think your boss would be like "is he working or looking at porn" when looking over your shoulder, you have gone too far.
Just keep in mind there is no age limit on this discord, any 8 year old could be currently looking at what you share

I can go more into detail if needed be, but tbh it's mostly common sence

[cyan axle]

That's fair. I just already really like this discord and don't wanna annoy anybody lol
Just delete if it's too much I guess? lol

[wispy nest]

if not sure, you can ping one of us mods too, or use the "blur" feature too

[cyan axle]

spoiler: woman in fairly revealing bikini

[wispy nest]

like, this is OK to me

[cyan axle]

This is the image I had been referring to earlier in the other channel regarding prominent features heh

Ah ok, just wanted to make sure because of those "prominent features" haha

[wispy nest]

we are not going for enticing pose, we are not going for sexy, we are not trying to get anyone hard or playing on the rules

[cyan axle]

👍

[wispy nest]

just be in good intelligence on this and I'll be the last one to disturb the party, but I also get people that play on the Loli line easily so, I'll just say that the ban hammer isn't far either if needed be 😉

[cyan axle]

Oof, totally get that heh

Got 4 kiddos myself. :p

[wispy nest]

yeah, I'm endorsing this role for them specificaly. I'm not a father but I'm a family man for sure, and I love AI. this is an open space, and I like it open for everyone

[cyan axle]

Precisely!

[wispy nest]

have you tried making something for/with them with SD? what age are they ?

[cyan axle]

I have actually!

15 y/o daughter prompted these two with some help.

[wispy nest]

those look quite nice ! base 2.1 model ?

[cyan axle]

She wanted to see "magical girls", and after I helped her with the first one, she did the second one almost entirely by herself, adding in things like the kimono and the background.

[wispy nest]

one suggestion maybe, POW, #⭐｜pow-info

[cyan axle]

what's pow? :p

[wispy nest]

maybe she would enjoy competing on a weakly basis. We make up a theme each week and people have 24h to submit a picture

[cyan axle]

Oh wow, neat. 🙂

[wispy nest]

there are small prices to earn but mostly motivation, people voting for you, you learning more and more about the tools, ...

it's a great avenue for progress in my opinion

[cyan axle]

Interesting

[wispy nest]

look at #1023999442338201721 for last weeks' submission as an example
theme was :
Bring the magic of the underwater worlds to life :cutewhale:
How would you capture the wonder and fluidity of the ocean depths?
What kind of unique underwater creatures or plants would Lyra have seen?
If you could explore one specific area of Lyra's world, what would it be?

I may start again next week, not sure, but in the past I made some models to help people with that week's theme

I need to get back into the game...

[cyan axle]

genning through a ton of images, but yah, i think I saw someone else also said that it's tough to get the mermaid tail to work. :p

[wispy nest]

well, SFW mermaid isn't easy for sure

[cyan axle]

That too!! lol

[wispy nest]

but going more description it helps

[cyan axle]

Like, this is one of the "better" sfw mermaids so far haha

[wispy nest]

like "naked girl with fish tail instead of legs"

mermaid is hard yeah

shell bra, I remember using that

lol

[cyan axle]

some of the nsfw ones are absolutely gorgeous though, if I could only get that kind of consistency with sfw lol

[wispy nest]

NSFW is quite good too. depends on the model you use for sure

some can't do a lot as soon as there is too much skin showing

[cyan axle]

should have a competition that allows for nsfw too haha

[hasty nova]

I have an embedding that helps with mermaids in 1.5

It's not perfect but it works a little better

[wispy nest]

yeah, but not on this server. we can"t host any NSFW channels since we are a community server. Discord's rules for being on the public servers list I think in part, but I also don't think Stabilty AI has ever endorsed NSFW image generation yet

[cyan axle]

Right right

[cyan axle]

What does embedding mean?

[wispy nest]

embeddings are a type of files. in this context, it means small training that someone has done for you, that you can add to your model to teach it more specific tokens. think of it like mods in skyrim if you like

[hasty nova]

Like a .pt file you put in your embeddings folder and use the filename in the prompt to enable it

[wispy nest]

that yeah

[cyan axle]

Ah, neat.

And I don't really mod games heh

Always been a vanilla purist. >.<

Which is probably why this has been such a pita hah

[hasty nova]

https://discord.com/channels/1002292111942635562/1070486045889663086 here it is if you are interested!

Anyway I must go, ping me if it helps at all as I would like to see what you make

[cyan axle]

Is there an easy way to force the image you're prompting to a particular depth in the image? like, having the character you prompt to be further back in the image so you can see more of teh tetails around that character?

[wispy nest]

it's a completion to the model, it's an aleteration of lots of weights in the model when you call on specific tokens.
In other words, the embeding trained on "Mr XXXX" will let you use the tokens "Mr XXXX" in your prompt, and that will access the training that was done and has been stored in that .pt file.
It's one of the training methods that exist out there. it's especially good for "style" training from what I got, but is on the lowest end of the quality scale, for the training methods out there. It has the big benefit to be very small to download, only a few MB

[cyan axle]

Amazing description!

[wispy nest]

lots of training methods exists, depending on your goal and resources

and about the depth, it's hard to say, I'm not entirely surz

I would tackle this right now by doing like this ;

1/ I would cut out my main subject from the first image, remove the background
2/ I would reduce the scale of the subject
3/ I would draw a mask, a black picture covering only the parts that aren't my main character after downscaling in step 2
4/ I would use some basic tool like paint to draw a very basic, stickman level, drawing of the background
5/ I would use that mask, and that base image with just the character and my handrawn background in inpaint, using the mask so that the main character doesn't get altered, and a low denoising so that the background gets nice and detailed

[cyan axle]

👀

Completely lost lol

[wispy nest]

lots of manual steps

sorry, does it feel like too much anyway or do you want me to clarify a step ?

maybe I didn't understand the question thoughj

reading it again right now, I feel like control net could help you a lot there

[dry crow]

you can just use close up shot, or portrait or cowboyshot for persons, for objects close up or portrait works too

you can also trick the ai by saying, woman, blue eyes, highheels, then it will try to get both into the image = fully body

[wispy nest]

true. every token in the prompt is considered. if you have a lot of tokens describing the background, it can tip the scale, push the AI to make more of it

[cyan axle]

Def too much for me for right now lol. I'll get there eventually though.

I suck at creating original art, even in paint lol. but I can describe tf outta an image! lol.

[cyan axle]

Ooh, see I like that. Describing both the eyes and the shoes would indeed trick the ai into making sure both fit.

[wispy nest]

well, even a stickman can become a real person with SD

[cyan axle]

Like, I can draw on paper kinda okish, but not so much with a mouse, and I don't have a drawing tablet yet.

[wispy nest]

like, i2i power https://www.reddit.com/gallery/yjnak3

[cyan axle]

Daughter keeps begging for a drawing tablet, but she only wants it to play OSU! on the thing lol.

So can't get a nice expensive one if she's just gonna be mashing the pen across the thing to play a rhythm game on it. :p

[wispy nest]

you can just install an app on her phone to use as a tablet for the computer though

I don't remember the name, I used one once

it's not that fast to respond, but for a child's use it stays coherent

[cyan axle]

Well, she's a teenager, so not really a child. Also, she doesn't have a phone. :p

[wispy nest]

good 😉 well not having a tablet isn't the end of the world. if you really need one of her drawing on the coputer you can always photograph it

[cyan axle]

True. 🙂

woah, trying to see what kinds of cthulean prompts I can get to work, cause I'm adhd af like that and a friend mentioned eldritch horrors. :p
Already looking pretty neat. 🙂

[cyan axle]

Ooh, got a really good one!

Cthulean horror themed characters. Nailed it.

Have any of you done any cthulean inspired gens?
Found any solid prompts for eldritch horror features?

[stoic pivot]

"a miniature tilt shift train landscape with a silly candy locomotive pulling carts through a (colorful) forest floor ond a winding candy track, (sparkling) glitter, 8k, master 3D render by PIXAR, ((close to the ground)) (close up), 12 mm, PIXAR fantast look, neon fairy dust glitter, render, tiny flowers around the stem of mushrooms, inspired by the movie TROLLS, fantasy surrealism, bloom, bokeh, lens flare, night shot, motion blur,"

I will try to refine it today

[stoic pivot]

"a miniature tilt shift train landscape with a (detailed steam candy locomotive) pulling carts through a colorful forest on a winding track made out of candy a small bridge passing over a small river foggy environment, (sparkling) glitter, master 3D movie render by PIXAR, ((close to the ground)) (close up), 12 mm, PIXAR fantast look, neon fairy dust (glitter), tiny flowers around the stem of mushrooms, inspired by the movie TROLLS, fantasy surrealism, dust particles, bloom, bokeh, lens flare, sunny day, sun rays", Model Deliberate v1.1

[stoic pivot]

Last for today, I hope they can be good for inspiration, it is just variations of the first prompt.

[grizzled sage]

hahahaha this has been my favorite personal meme remix yet

it uses his chin as the adams apple it's so good

[lofty tundra]

My first ever Stable diffusion image! I tried to make a medieval king eat a laptop.

[static tusk]

starting to get the hang of multi controlnet

[marsh vortex]

I love the demo lol

[languid cradle]

#flamrngo

#imagine Superman

Gg

[wispy nest]

Oh no something bad happened.

Something came out of the computer.

Oh noooo aahhhhh...

[marsh vortex]

what have i done

[steel oyster]

anyone have some model recommendations for a fantasy-ish more concept art sort of style?
https://image.invaluable.com/housePhotos/malletauctionjapan/49/276849/H3876-L14896920.jpg
something like that
https://i.pinimg.com/474x/40/8b/2f/408b2fea320bbdcd864ecfb5357f58ad.jpg

https://i.pinimg.com/236x/0e/1b/93/0e1b939b65e9c4f670db1ceca6a8b22e.jpg

more like these two really, not even sure how to describe that style

[fiery zinc]

have you tried prompting for fantasy artists, like "in the style of Frank Frazetta" or "Boris Vallejo"?

I'd try it myself, but I'm currently in the middle of a controlnet experiment

[whole pivot]

isometric diorama prototype, 8-bit, Astronaut on an overflowing alien planet, particle effects, 2.2 gamma, sony a7r7, Tamron 10-24mm f/3.5-4.5, ISO 3200, extremely detailed, 8k texture

[steel oyster]

thanks, I'll give these a shot

[fiery zinc]

let me know if it works 🙂

[steel oyster]

great results with frazetta so far, trying boris now

[steel oyster]

these are perfect, how did you come up with the names?

[fiery zinc]

I happen to have their art books from back in the day 😄

[steel oyster]

ohh badass, well thanks again!

[fiery zinc]

np

[signal basalt]

Do you need a powerful computer to run stable diffusion? My request is really out there

[fiery zinc]

for rendering, you may be limited in size with small vram

and of course speed

[signal basalt]

Darnit

[fiery zinc]

for training, you really want 8GB vram or more

[signal basalt]

I got a potato laptop

[fiery zinc]

you can get away with textual inversion embedding training with 6GB, but less than that you won't be able to train, probably

what's your spec?

I'm doing most of my work on a 1660Ti laptop with 6GB VRAM, so it's not exactly a speed demon

I just have coffee and do other stuff while it's rendering

[steel oyster]

ohh, have you had any luck training your own styles?

[fiery zinc]

I haven't tried styles yet, only faces

[steel oyster]

I have 8g but every tool I've tried training with fails due to lack of vram

what do you use to train the embeddings?

base automatic1111?

[fiery zinc]

you can do textual inversion with 8GB, and probably LoRA too

A1111 for TI, kohya_ss for LoRA

https://www.youtube.com/watch?v=70H03cv57-o

https://www.youtube.com/watch?v=2ityl_dNRNw

[steel oyster]

thanks!

[fiery zinc]

np, good luck

[steel oyster]

unrelated question, when inpainting, should your prompt be about the whole image or just the area you want replaced?

[fiery zinc]

either, just make sure you have an inpainting model

if your preferred model doesn't have an inpainting version, make your own: https://www.reddit.com/r/StableDiffusion/comments/zyi24j/how_to_turn_any_model_into_an_inpainting_model/

I generally stick with the original prompt if I'm just replacing a lower resolution face and such

but if I want to add details, I'll specifically prompt for those

[signal basalt]

Ok I just looked it up, I'm not good with computer stuff, but I have a geobook Minecraft laptop, I don't see all the specs but I'll tell you the ones I do see,
Processor Model. Intel Celeron.
Processor Model Number. N3450.
Processor Speed (up to) 1.1 gigahertz.
Storage Type. eMMC.
Total Storage Capacity. 64 gigabytes.

[steel oyster]

thanks, that cleared up inpainting quite a bit, that reddit post was really key

[fiery zinc]

ah, yeah, not going to work. Doesn't have a dedicated GPU as far as I can tell.

your best bet is to use a google colab

[fiery zinc]

For example, I repeatedly inpainted to add the tank, the damaged road, the buildings in the back, the fire, and the smoke, each with their own specific prompts

[steel oyster]

good to know, thanks

I kept getting funky results but my model not coming with inpainting was probably a huge reason

[fiery zinc]

yeah, then you'll get misalignments or worse in the mask

not so bad if you just want to insert something, but bad if you want to mask-and-fill

[onyx abyss]

what's a good scene background for a realistic person i'm refining.

[strange jungle]

Tiny Obama

SD's version of the image

[strange jungle]

odd SD added the text bubble

[whole gale]

Does anyone have a bunch of good openpose images? I'm struggling to make poses that result in good-looking characters.

[onyx abyss]

thoughts on this one? i was going for a different look but this was pretty stunning and the hands dont look too bad

[onyx abyss]

https://www.reddit.com/r/StableDiffusion/comments/1180fls/sharing_my_openpose_template_for_character/

[smoky oak]

Currently training a LoRA on Na'vi

here is hoping it comes out good

[smoky oak]

squashed

[grizzled sage]

ohhhhhh wait they wanted entirely different suits on the guy. not just replacement colors. mb

i dont understand the point of this exercise. it can obviously be done

[static tusk]

hahaha.. bro you just used photoshop, right? /s

[idle oyster]

Hi

[analog goblet]

🙂 yes, that looks like photoshop

that's not at all the point of this exercise 🙂

[static tusk]

Like how I used PS to change the suit colour, which magically added a watch to the mans wrist.

[grizzled sage]

such an incredible amount of photoshop cheating

sure am pretty good wth the photoshops

[analog goblet]

those even all have the same lighter blob around the background of the head 🙂

[grizzled sage]

i wonder if that gives you a clue to how did it almost i wonder

oh why why why why why is a light blob up there, and i wonder why the blob is therrrerr

[analog goblet]

new task would be: make it in a way that it doesnt look fake 🙂

[grizzled sage]

its like, one of the ez methods i said, roughly done, which you've thrown back at me and said i never help you

spent days with you an you come back an say i'm not solutions focused. seirous troll vibes

[analog goblet]

yes, i m sorry about that, i might have been wrong here

right now i'm rather confused a bit

[wispy nest]

How can I generate a imagery

[analog goblet]

google automatic1111

[burnt gulch]

/dreambot

/dre

[crystal delta]

FAQ: What is the copyright on images created through DreamStudio?

Images created through DreamStudio are fully open source – specifically falling under the CC0 1.0 Universal Public Domain Dedication, available at https://creativecommons.org/publicdomain/zero/1.0

[burnt gulch]

/dreambot

[wispy nest]

no dreambot are up around currently sorry

[compact inlet]

/dreambot

[crystal delta]

FAQ: I'm new here, how do I generate images ? Where is the bot ?

Welcome ! There is no bot currently to generate your images on discord. You may want to start by taking a look at the #1014939219904450590 channel. You can access Stable diffusion in different ways : 1️⃣ the official website, https://beta.dreamstudio.ai/. The easiest and fastest way to access Stable diffusion with 200 free credits. For any question on it, you can find help in the #1025467151206854736 channel. 2️⃣ Installing Stable diffusion on your computer. There are numerous projects that let you do that, and you will find help in the #🤝｜tech-support channel. 3️⃣ Running Stable diffusion in the cloud, through rented GPU services, using notebooks. You can find lots of them shared and discussed over in the #1011228442399883294 channel.

[wispy nest]

this needs an update for the first link, but the rest is all valid

[cold berry]

/faq command is being reworked atm! In the meantime #1072229020520947753 should be up-to-date!

[wispy nest]

Control net is quite the stuff for sure, variations on my avatar

[eager niche]

Ronaldo playing golf

[sand sonnet]

My first AI generation

[hollow grotto]

😂 #1019361238234443776

[sand sonnet]

😭

[sand sonnet]

Bet I’ll post it there instead

[fiery zinc]

I think that's many people's first fail 😆

[atomic nacelle]

un logo reebot, UNDER ARMOUR, nike

[kindred parrot]

It's me again! Please forgive my reminder ❤️

Don't forget we have a server-wide survey going on right now to collect your inputs and feedback on ways we can give you what you want from this server!

The survey is being closed <t:1677517200:D> so make sure you fill it out if you haven't already!
(If you would prefer to remain anonymous, simply hit "next" on the first question without putting your discord username in!)
https://docs.google.com/forms/d/e/1FAIpQLSdpaL4lcXvdexbLEnAnP-G3w5e56nXuq_0-CfWDFKpmpAlSww/viewform

[wispy nest]

hummmmm milk

[young mountain]

Ramen I want

Ramen I get

[wispy nest]

I just finished eating so I'm good but damn, I'm happy I just did or I would need to

[stoic pivot]

Todays random images.

[fiery zinc]

I don't think this is too bad a bowl of ramen

granted, will need to fix the chopsticks

maybe some other topping

but for first render, I don't think it's that bad

[stoic pivot]

It look good, try something fun and odd as topping, something folk do not expect, like a strawberry 😄

[fiery zinc]

no, @young mountain would complain 😆

[grizzled sage]

bdsm weird memes before they're remixed. this is the 50 shades of grey for memes.

[dense tapir]

LOL

[wispy nest]

xD how did that happen ?

lol

[dense tapir]

No idea but I LOL every single time I see it

[steel yarrow]

horny police, you have a complaint Sir ?

[grizzled sage]

nope. just pointing out the obvious weird sadism that bdsm promotes. i'm not about to partake in your hedonistic domination/submission fantasies

being horny is part of the human condition, but we are all reasonable people that are supposed to be able to contain it

[steel yarrow]

those are two trans humans, one feeding oil to another

because that's what their cogs need

[grizzled sage]

i dont know why people think obvious lies about their intentions will work. it just makes you look more disingenuous

[dense tapir]

I gotta agree with flowwolf on this one. First time I saw that pose with all the differences I immediately knew what was going on. No, I don't care for it but you do you.

[steel yarrow]

I really don't need anything of what I do that have a justification, much less have someone else's twisted context imposed on it, that's the dangerous part of being falsely virtuous. I don't need approval either. That's a pose. If it raises your BP, that's your problem. Doesn't make your morality superior.

have fun with that context

[grizzled sage]

there's no twisting it. that meme is bdsm 50 shades of grey culture. that's it's whole thing

nail meet hammer

[steel yarrow]

that's a repurposed pose, if all you have is a hammer, everything might look like a nail, but it simply isn't, it cannot be

[dense tapir]

Well, if that was an adult and the child below does that change anything?

[steel yarrow]

now that we're playing "what if", what if you looked at an image and didn't project all your dark into it

how bout that junior, that twist "to protect the children" kind of went kaboom there

so I'll leave it on your hands

[dense tapir]

If you can't see the bad in that, junior, you are too far gone. What's with the the above down? Where are her hands or is she bound? Why is his/its hand on her ponytail as if by force? That last one no innocent picture would be of.

Now you do you but people know, and will call it out for what it is regardless how pervs wish it not to be perceived as. As I said you do you.

[steel yarrow]

Oh that's so insightful and really revealing of all that goes in the dark corners of the mind. Also, the way you're losing your pose is quite telling. Enjoy the images anyway.

[grizzled sage]

Junior is a sweet movie

[dense tapir]

Yep, it was.

[cinder vine]

CONTROLNET

[grizzled sage]

heyy i just posted on thhat thhread. yeah seriously!

[static tusk]

removing one of the chat rooms and adding a controlnet channel would be cool

[grizzled sage]

put it at the top then, everything is under control

[wispy nest]

Not sure where it would fit best, but I think a space to talk about it would be cool for sure

I'll suggest it next mod meet, but let's copy that to the #1002293361526460608 channel too

[fiery zinc]

then it'll just get 99% of the traffic 😆

[wispy nest]

for now but it's like every craze, things smooth around

plus you can't imagine how much trafic some communities get here 😉

lots of people active everyday

[static tusk]

unless there are some major updates I think controlnet will be a mainstay

[wispy nest]

yeah, but like i2i will stay a staple, it is less talked now than it was at first. like today, I wouldn't go on a wall of text explaining how to control composition in i2i when control net is around

[grizzled sage]

lots of good source images for cnet using this search too. https://www.google.com/search?q=optical&20illusion&tbm=isch&tbs=itp:lineart

[wispy nest]

but even yesterday before CNet came into play, it wasn't talked a lot anymore because people mastered it

[grizzled sage]

"lineart" is one of those googel deep cuts thats suddenly useful now

[wispy nest]

once we get the hang of it, controlnet will be just in about everything and we won't talk that much about it, but about the next thing that changes in some other areas

you can even .... generate lineart :p

[grizzled sage]

i think it wont even be just cnet. i think this will be a processing stack that all cfg/attend and excite/self attentitive guidance/other extra networks and trained models for guidance, will just become a big stack that you can swap modules in and out of. CFG was the first, but it was just a quick dirty hack

cnet models will just fit in as one of those modules

anyone familiar with avs in winamp?

[static tusk]

too many years ago

[wispy nest]

for sure, preporcessing like this is here to stay

[static tusk]

quick google. ah yes I had a bunch of animated speakers on my install

[wispy nest]

just like Hypernetworks, embeddings or others complementary networks

I'm sure we will see more controlnets apear with different input types

imagine music input for example

the video making AIs that have already been teased here and there by google and other sure will hit us and have ripples in the image generation world. I personally can't wait

[static tusk]

for sure video will be the driving force

[wispy nest]

for now, #1011634831467221033 seems like the most fitting for controlnet though

all inputs are pictures, and it behaves comparable to img2img

[static tusk]

true

[wispy nest]

we talked about it a little, we'll talk more when more mods are around.
For now, we change the description of the img2img channel to include controlnet, and invite people to talk about it there
we may then rename it to something more generic than img2img, or create another one, or I don't know but I know we'll try to find the best, least confusing way 🙂

and on that, I wish you a good time of day that it is locally, and I personally go sleep. see ya !

(and happy to have met someone new and interesting !)

[left crane]

Quick reminder that stable diffusion animation is now available as a cloud service

https://app.kaiber.ai/invite/375d44e6-86ba-43f3-bdce-c49e0c45555b

[smoky oak]

is there info on how one would do that on their own PC?

I have been meaning to make an animation like that that goes between 10 different men across history

[smoky oak]

Vmles

Tk OAVE 💟

[smoky oak]

those are pretty sick :p

[rain gazelle]

I did a lot of gens today... lots of keepers

[smoky oak]

Very neat stuff :D

[smoky oak]

Those are all sick :p

[fiery zinc]

@honest atlas

[honest atlas]

where do i find that

this?

[fiery zinc]

you go to a project, then select files and versions

[fiery zinc]

yes

[honest atlas]

and which one of those 2

[fiery zinc]

not the -unpruned

2.13GB one

save you space and doesn't have junk in the file

[honest atlas]

and, is v1 better than v2 oj?

(as in should i not go over to v2)

[fiery zinc]

not sure what v2 is, honestly

maybe based on the v2 base?

[honest atlas]

idk

[fiery zinc]

I don't use openjourney, so I can't really help you there

[honest atlas]

v1 basically has 500k downloads, compared to v2 like 40k

in last week

[fiery zinc]

This repo is for testing the first Openjourney fine tuned model.

for v2

so maybe it's still in testing phase

[honest atlas]

okay

i should go for safetensors?, from what i know its more stable on SD ay?

[fiery zinc]

both are fine, safetensors is "safer" if you get it from an unknown source

but I believe hugging face scans for malicious codes in the ckpt files

[honest atlas]

so which should i pick

[strange jungle]

I dl'd openjourney-v2.ckpt

[fiery zinc]

doesn't really matter. If you're using automatic1111, then just go safetensor. It supports both.

[honest atlas]

ait

[fiery zinc]

if you're using another UI, then check their support

[honest atlas]

a111

[fiery zinc]

some UI may not support the newer safetensor format

[strange jungle]

I dont see a safetensor there

[strange jungle]

https://huggingface.co/prompthero/openjourney-v2/tree/main

[fiery zinc]

I'm just looking at the screenshot he posted

[honest atlas]

do u know more about v1 or v2

[strange jungle]

on huggingface it shows v2

I know nothing about it...want to try it

[fiery zinc]

there's also https://huggingface.co/prompthero/openjourney

he has two separate projects

[honest atlas]

from what i see i dont have to use the prompt

[strange jungle]

v2 says no longer have to include the prompt

[strange jungle]

for V2

[honest atlas]

ye

do i need to be registered to see download buttons

nvm

[strange jungle]

for the v4 one, it says include the prompt

Is the one I linked newer, even though it shows V2?

[honest atlas]

idk xD

[strange jungle]

this is confusing

[fiery zinc]

the v2 mentions "This repo is for testing the first Openjourney fine tuned model", so I assume it's still in testing

[honest atlas]

idk what v4 are u on about

v4 refers to midjourney algorithm

prolly thats what confused u

[strange jungle]

ah

it says you have to place mdjrny-v4 style in prompt, but then I go to the sample prompts, and they don't have mdjrny-v4 style

doesnt matter

neither model will load

error: header too large

[fiery zinc]

maybe try the one from civitai then?

[strange jungle]

for both openjourney-v2 and mdjrny-v4

[fiery zinc]

https://civitai.com/models/86/openjourney

[strange jungle]

OK, Im trying it

that one works

I wonder why the others dont

par for the course tonight. BB had a 4070TI open box for 650. I added it to cart, but then before payment it showed unavailable. Someone else then added it to cart, bought it. He already has 2 4070TI's and 2 3090's he doesn't use, but he bought this one - and it let him buy it when it wouldn't let me buy it.

Hmmm...I didn't put mdjrny-v4 style in the prompt, but it still makes distinct images

[fiery zinc]

yeah, I believe for full checkpoints, the prompt is just a stronger nudge, but not mandatory like embeddings

[strange jungle]

exactly the same style with "mdjrny-v4 style" in the prompt

[strange jungle]

I see

[molten prairie]

zaha hadid

[hasty nova]

@amber smelt here it is

[amber smelt]

ty!

[smoky oak]

priddy :>

[amber smelt]

What is causing this dramatic color shift? It occurs with or without color correction enabled

(color correction slightly lessens it)

(right is with CC)

[dense tapir]

I never use that, or even knew about in automatic1111

[amber smelt]

Mind you, the huge shift still happens if color correction is off, just a little bit less

[dense tapir]

I take it this is img2img?

[amber smelt]

Yes

[dense tapir]

I haven't played in that since 2.0 dropped but that was my playground before then. Is this 1.5 or 2.1?

[amber smelt]

1.5, controlnet doesn't work with 2.0 and I've heard 2.0 isn't great

[dense tapir]

you aren't using any loras, right?

[amber smelt]

What's a lora? I am using dreamshaper but that's a 1.5 model

(No idea what a Lora is btw)

[dense tapir]

controlnet can break them, so no worries. I don't use controlnet so I can't help there until it comes to 2.1 as I don't touch 1.5.

[strange jungle]

that looks good

[smoky oak]

Man I love SD

[dense haven]

Does anyone have any idea how to generate xiqu characters? I can't find the right keyword for it

[quasi mist]

Sry im just trying to make car images irdk

Why is only one of the images closer to what I wanted? none of the others look as sexy as this, and also the finishes and materials and colors are actually unexpectedly good.

W I like this one even more even though it IS NOT what I had wanted it to output

[wispy nest]

@brisk holly

@clear parrot

nope, no bot aroud currently

[crystal delta]

@wispy nest

FAQ: I'm new here, how do I generate images ? Where is the bot ?

Welcome ! There is no bot currently to generate your images on discord. You may want to start by taking a look at the #1014939219904450590 channel. You can access Stable diffusion in different ways : 1️⃣ the official website, https://beta.dreamstudio.ai/. The easiest and fastest way to access Stable diffusion with 200 free credits. For any question on it, you can find help in the #1025467151206854736 channel. 2️⃣ Installing Stable diffusion on your computer. There are numerous projects that let you do that, and you will find help in the #🤝｜tech-support channel. 3️⃣ Running Stable diffusion in the cloud, through rented GPU services, using notebooks. You can find lots of them shared and discussed over in the #1011228442399883294 channel.

[smoky oak]

Wow, POS

@tropic shell Isn't that against server rules?

[wispy nest]

yep it is

warning sent, next is timeout

[smoky oak]

Thanks, not a fan of hate speech lmao

[wispy nest]

me neither, sorry I took some time to come back

[smoky oak]

No worries

just glad to see a server that actually takes that stuff seriously. Too many let hate speech slide lmao

[wispy nest]

in case of this, right click on the message, applications, warn staff

it sends us a ping

[smoky oak]

Oh, thats dope!

I never knew that. Will def use that in the future

[wispy nest]

we have other ways, like you can react to a message with ⚠️

[smoky oak]

Very cool, very cool

Will have to keep that all in mind

For now, I must go back to trying to figure out why my stable diffusion is suddenly not working-

oh, I was on the wrong model

working now, how weird

[dry crow]

Nice didnt know that

[smoky oak]

wouldn't let me even get a 512x512 image once without giving out of VRAM

I usually generate 8x 768x768 images at the same time with no problem lmao

[onyx abyss]

wow thats cool as hell.

[smoky oak]

Thanks! I decided to retry some of my wombo dream prompts in SD

[wispy nest]

hum.... anything changed?

system update ? graphics card ?

[smoky oak]

it ended up being the model it seems

[wispy nest]

ok

nice if you found out

[smoky oak]

its an inpaint model, all the others are working fine

very weird, never had that happen before

There is a little owl that has been hooting outside of my window for thelast 4 hours, and its my little guardian angellll

https://cdn.discordapp.com/attachments/514246592761430045/1078567615938711562/20230223_224137.mp4

(very quiet)

[dense tapir]

the neck that never stops, lol

[smoky oak]

N

E

C

K

[dense tapir]

that is a sign of the nonsense they did with 2.1 doing most of it 512x512 then finishing it 768x768 so you get stretched necks

[smoky oak]

rich

[smoky oak]

oh, I like that one :D

[dense tapir]

Yeah

[smoky oak]

fox sleeping with riches

[dense tapir]

Until the dragon comes

[wispy nest]

Peacock dragon

[sonic panther]

a girl with smile

[sonic panther]

young girl, bed, smartphone, surroundings, room, toys, colorful, playful, carefree, long brown hair, pillow, face, phone screen, fixated, activity, content, diverse collection, stuffed animals, action figures, board games, youthful energy, bright colors, playful decor, innocence, distraction, joy, amusement, digital world, physical toys, bygone era

[tropic shell]

I'm going to assume

There was

S p a m

[smoky oak]

hMmMmMm

[storm stump]

#1072229057707659404

[smoky oak]

@sonic pantherDecided to do a megabatch

your prompt lol

[tropic shell]

Hooray for Guiz. Though ❤️

[smoky oak]

Guiz?

[tropic shell]

Also. Those are SO WHOLESOME

Guizmus

[wispy nest]

someone not liking being linked the FAQ and going all slurs directly

I removed the offensive messages and gave a warning to the user

[smoky oak]

oh yes, thank you again for that

making me hungry 😩

[violet carbon]

Hi

How do i start?

I'm New here

[bitter nimbus]

A coat of arms with a goat

[wispy nest]

hey @violet carbon