🏞｜general-with-images

[cyan shoal]

just wanted to ask if you had like lighting or hyper version

premade resolutions/aspect ratios?

[gilded anchor]

yes

[cyan shoal]

hmm

[gilded anchor]

isnt that nice in fooocus

cuz then u can just do a 4:5 if its a tee design, do a 1:1 for a album cover etc

[cyan shoal]

I tried fooocus once, it feels like the perfect beginner software for Stable Diffusion

it takes care of a lot of stuff

and the dev made some nice models for SDXL, so that you can inpaint

[gilded anchor]

i just have used it inpainting only

works insane the times ive used

[cyan shoal]

that inpainting model the creator made is REALLY nice

[gilded anchor]

oh cool

can i add you? you seem friendly

[cyan shoal]

sure

[gilded anchor]

sweet

do you have those resolutions just locked in your head now haha

i guess ihave to do the same

[cyan shoal]

nah I just remembered like 2 of them

[gilded anchor]

i have my photoshop res locked in like 4800x6000 px 300dpi and stuff

oh okay

[cyan shoal]

I use comfyui, which is like 300x more complicated

[gilded anchor]

i just need to remember the 4:5 ration

and i already remember 512 and 1024 obviously haha

thank you for the help man

[cyan shoal]

there's a node for resoltuion picking for SDXL in Comfyui so its easy

[gilded anchor]

i finally got SD forge working after hooooours then finally it started bugging

[cyan shoal]

you're welcome

[gilded anchor]

u use comfy mainly?

over forge?

i have tried its just complex for me sometimes

oh i see you said u use that yeah

[gilded anchor]

@cyan shoal real 1 for that

@cyan shoal does that comfy ui added you talked about include the ratios too? that woulda been nice if it doesnt

[late sorrel]

one popular thing in comfy is something called rgthree nodes, one of them is an empty latent node for sdxl with a drop down menu that has all these sdxl dimensions ready for you 🙂

hmm yea idk, never tried forge, i went from a1111 straight to comfy and not using anything else 🙂

[vestal wraith]

What model was used to create this?

[late sorrel]

a lot of models can do that

[vestal wraith]

Such as???

Idk I just want one
New here

[late sorrel]

i mean let me rephrase, i dont see anything special at what im looking at there that cant be done with a lot of models out there, so dont think its some special unique model

[vestal wraith]

How do I do anime with jugg?

I see
No loras?

[late sorrel]

yea these days ipadapter style transfer is very strong and can copy it pretty nicely

[vestal wraith]

Ipadapter??
Hearing for the first time

Tbh, I built my new pc this month and Finding these amazing stuff. I am just , I mn lost my way here and there

[late sorrel]

haha well there is indeed a lot of greatness to learn 🙂

[gilded anchor]

certaintly, i know nothing yet

@vestal wraith https://www.youtube.com/watch?v=czcgJnoDVd4

hope this helps you pal

just as much of a noob as you lol

[vestal wraith]

Thanks sir
You've been very helpful

☺️☺️☺️Cheers mate

Wtf was that?

Worm thingy

[late sorrel]

prob some troll

[vestal wraith]

🐛??

Oh I see

[late sorrel]

yep

[vestal wraith]

Bro
Can I use ipadapter with forge?

Comfy is too complicated for me

[gilded anchor]

use it with comfy imo from my research not gonna lie

comfy is complicated but something to get used to potentially

[late sorrel]

it's only complicated at the beginning perhaps, then stuff clicks

[gilded anchor]

i need to make an organized folder and document with what does what at the time

sometimes i get so lost lol

[vestal wraith]

Okay, I can do this. There are those nodes and stuff. I saw them on blender

[gilded anchor]

yup

[vestal wraith]

This troll thing is going to far and it's becoming disturbing

[gilded anchor]

overwhelming for me too brother

[vestal wraith]

Can I install comfy with pinokio?

Okay it's settled then

Me too

Disgusting

[late sorrel]

i managed to quickly capture his id, so now someone can ban him or idk

[nocturne oak]

Someone 'bout to get banned.

[storm jackal]

no prompt images

yeah, LCM lighting model and DW Pose

I set a generated image as latent image into Ksampler

I think those were images it was trained on, wasn't it?

[wispy spindle]

they were banned & reported to discord- sorry for all tht had to see that ):

[late sorrel]

yea what the hell was that, thank you

[wispy spindle]

really disturbing

you too!

[late sorrel]

maybe it depends on the model used i guess, i remember getting something real bad lol

[storm jackal]

I got the correct hands without a prompt

[late sorrel]

lol

magic

[storm jackal]

look at those images, mostly correct

[south temple]

Well hi fam

[tawny lance]

Hey. What's this gen chat all about?
All SD all the time?

Spaminating? Spill das beans

[soft meadow]

My stable sr give me this results when I upscale my image to 6k

Bruhh

Dirty

[wild sorrel]

you're probably ain't doing it right

[crisp stream]

let it run through realESRGANplus x4

[wild sorrel]

denoise strength increase might help a bit too or scaling up bit by bit

[south temple]

are we playing with SD3 yet?

[wild sorrel]

I think still only through api

[soft meadow]

Choose r esrgan plus ×4 in the tile upscaling method?

[south temple]

yes

[soft meadow]

Okay let's try it

[south temple]

im late to the party, what are you trying to do?

upscale?

[crisp stream]

yep

[soft meadow]

Ya upscale my pic to 6k

Using stable sr

[south temple]

might need another ksampler

[crisp stream]

and?

[soft meadow]

Too slow I'm processing

But my pic is anime style have I need to change to esrgan anime?

[crisp stream]

still cookin' ? 🤭

[soft meadow]

[crisp stream]

nope

[soft meadow]

Oh okay

[south temple]

Ill toss you the same images with different upscales

[soft meadow]

[south temple]

esrgan

NMKD

ultrasharp

[soft meadow]

Hrmmm great

tQ

[south temple]

animesharp

take your pick

lol

[wild sorrel]

rev animated?

[south temple]

rev animated?

[crisp stream]

where is realESRGANplus?

[wild sorrel]

model

[south temple]

I dont use that one...

good?

[wild sorrel]

I feel like I recognise that style, thought it's rev

[south temple]

names are above each pic

[crisp stream]

Check it out 😉

[soft meadow]

Hmm good

[south temple]

looking now

[wild sorrel]

too lazy to pull up ui to check model 😄

[south temple]

OOoo I see what your talking about

[wild sorrel]

was pretty good at a time for non-realistic stuff, could understand alot more prompts than other 1.5 base models

[south temple]

I have a lot of other models.. I just like this set

some great stuff

[crisp stream]

ah, I thought you were referring to realESRGANplus 😀

[south temple]

im trying that now

latest model 6b

[wild sorrel]

some of my old images with this model

it was 1.1 version

[crisp stream]

Who are you talking to? 😀

[south temple]

I dont know that I love it

[wild sorrel]

lol

[crisp stream]

a liitle young, no?

[south temple]

that one...

little better lower steps but also artifacts

I tend to like ultrasharp the most

or anime

100 steps

34

[wild sorrel]

different upscales are good for different things, I found out ultrasharp is just terrible for realistic stuff most of the time

or at least was

[soft meadow]

Any negative embedding about bad hand for sdxl model?

[south temple]

ultra

real anime

animesharp

[wild sorrel]

I think for that style anything will do

looks like pretty high denoise too

[south temple]

looking at what youve done here

[wild sorrel]

yea just read the prompt too, negative is funny lol

probably should delete it from there lol

[south temple]

I dont have the chkppoint so itll be different 🙂

[vestal wraith]

How do I wrap my head around comfy ui?

It is too alien for me

[south temple]

lol

slow and steady

[wild sorrel]

it's fine, just wanna see if it'll ruin details or not, when I tried it long time ago - I remember some cloth textures was weird

maybe hair too, not sure

[vestal wraith]

I don't even understand where i can see the preview

Where is the preview??

[south temple]

lol one sec mate ill help you as soon as I check out this thinggy

[vestal wraith]

Thanks

[wild sorrel]

oh, go low denoise on it, so model doesn't affect it as much, not higher then 0.2

[south temple]

ultra

the face is a bit tough

checking with 60 steps

[wild sorrel]

that really high denoise damn

[south temple]

1024x1024

[wild sorrel]

yea skin and some cloth textures are broken

[wild sorrel]

but also could be model being weird, that's too high denoise, model affects it too much

could be model not being good at realistic stuff, not upscaler issue

at this point you're like making new image, not upscaling 😄

[south temple]

its a pretty good model

I can try a different one tho

[wild sorrel]

try 0.1 - 0.2 denoise strength, that's how I usually upscale to keep image same

it will lower how much model affects it too

[crisp stream]

within the Save Image node, needs generating before it shows

[south temple]

lol

ok

[vestal wraith]

Gotchu

[wild sorrel]

how is that so much different

[wild sorrel]

preview slows down image gen btw

[south temple]

im tweeking

[vestal wraith]

Understood but
I can't find where I input loras, I can't even change the height and width to whatever I want
It sets the height on it's own
I think I need to do the course on YouTube

[wild sorrel]

I have no idea how to comfyui too, I barely made basic workflow some time ago 😄

[crisp stream]

the depiction within the Save Image node takes noticably longer?

[wild sorrel]

I think you can google workflows to do some "common" things

[vestal wraith]

No worries

I will learn it whatever it takes😠

[wild sorrel]

Depends on a gpu probably, for modern ones shouldn't take too long, for older probably will be noticeable

it has to render image x amount of time during generation, that's some additional steps it has to do for preview

[minor fractal]

if you use huge images? maybe, the real long step can be vae decode

[crisp stream]

isn't it merely showing the generated image as if I would open it from the harddrive? What other calculations/processings are happening

?

[wild sorrel]

oh if it's just shows what it generated, not steps in between - then there should be no difference

[minor fractal]

other than adding metadata, nothing

[crisp stream]

yes, then it isn't relevant in this case, no?

[wild sorrel]

I was thinking of a1111 style of previews, where it shows you preview of what it looks like on each x steps

[wild sorrel]

probably

[minor fractal]

comfy has that, just not automatically

which, even then, uses taesd, which is very fast and is neglegible slowdown

[soft meadow]

Doesn't fix the problem even I use esrgan

[crisp stream]

Can you hand an Image you would like to upscale? No workflow needed.

[soft meadow]

Okay

I using forge ui

I wish to upscale to 6k resolution

Hrmm but forge can use tile diffusion so I use a1111 when upscale

[crisp stream]

@soft meadow that's close to 14k btw

[soft meadow]

Burhhhhh

It's super big

How do you do that

[crisp stream]

With realESRGANplus x4 (not ESRGAN btw)

[soft meadow]

Ya with R- right

Using stableSR?

So fast

[crisp stream]

I use comfy and the upscaler within the workflow, yet you can simply load it in whatsoever UI (if it supports loading upscalers)

[crisp stream]

yes, and the "plus"

@soft meadow That´s its exact name: RealESRGAN_x4plus.pth

[crisp stream]

https://huggingface.co/spaces/Marne/Real-ESRGAN/blob/main/RealESRGAN_x4plus.pth

[south temple]

would you eat it?

[soft meadow]

This is mine

[crisp stream]

yes, looks good

[soft meadow]

Can forge or a1111 do this ?

[crisp stream]

a1111 can, don´t know about forge, yet guess so

[soft meadow]

What mean upscaler within the workflow ?

I not really understand this

[crisp stream]

Within the Comfy Workflow (workflow is what a setup in Comfy is being called:

[crisp stream]

@soft meadow one step is enough btw

[soft meadow]

Ohhh

So you upscale 4 times for me by using 4x

[south temple]

knda

within the upscaler it is a 4x upscale

[crisp stream]

yes

[soft meadow]

Ohh okay

I got it

Thanks

[hazy warren]

hmm

dunno how to make this an animated image in comfyUI

[languid pebble]

Looks like SVD

[hazy warren]

it is I think

it saves as webp some reason

[languid pebble]

You will need the checkpoint, workflows should be available and a start image ...

[languid pebble]

Why it saves as Webp ... I don't know ^^

Your Node says to save as webp

[hazy warren]

I know that

but I want it to be a gif or something

[languid pebble]

There's a cool free website to convert to gif ... ComfyUI I don't know what you need

[hazy warren]

this looks cool

[steel berry]

Create an image of a lobby round table with floral arrangements

[flint trench]

insparitional of love and peace

[covert pagoda]

No

[deft bison]

Thank you ! ...

[uneven moon]

make robot

[unique shuttle]

is it possible to give bodies to headshots ive already generated?

[languid pebble]

Survived another B-Day 😄

[unique shuttle]

always good

[covert pagoda]

that’s not an image!!!1!!!1!1!1?

[languid pebble]

???

[covert pagoda]

I was making a joke about how you posted a video in #🏞｜general-with-images

[languid pebble]

hehe Made of a pretty old image 🙂

I could also post them picture by picture 😛

[silk shoal]

what went wrong there?

[languid pebble]

Not from Stark Industries ^^

[silk shoal]

soon brother soon ill get there,,,

one day, i hope....

[languid pebble]

I wouldn't buy from Elosn 😄

[silk shoal]

any idea how could i improve it?

im here for feedback

[shy eagle]

SDXL FT

[silk shoal]

pls be honest

[languid pebble]

It's your picture. What did you have in your mind creating it?

[silk shoal]

a robot

[languid pebble]

Fits 100%

[silk shoal]

hmmm, maybe background or something?

or more humanoid robot?

like cyborg

[languid pebble]

Well .... I'd think "robot" isn't really enough. Maybe better think of a short story. Where is he? What is he doing? So you can let your picture tell a story

[silk shoal]

OMG THX

this is what i was looking for

tysm brother

[languid pebble]

Rememer the Geisha Animation above? Took me some weeks. First idea: Geisha, but that wasn't enough. Than I saw people buying blood tattoos, combined both but didn't really fit to the background I wanted ...

And than I've remembered the traditional asia story about a bad guy in a small village and the people call an assassin ...

[silk shoal]

oh yeaa that shit is impressive

man i love ai

you used svd to animate?

[languid pebble]

First step is always starting in your head. The better your idea, the harder the work but I think also better results.

Yes ... I've had the picture at an exhibition one year ago. Just put it into SVD today 🙂

[silk shoal]

thats crazy

thx again for the idea

[languid pebble]

You are welcome. I'm just sharing ideas. Happy if they fit.

[crisp stream]

What is that background sound?? 😄 I at least think you were going for frightening, yet it sounds a little different 😄

[crisp stream]

yw 🙂 So did it work out?

[vernal raft]

五一劳动节背景 红色主题

[nimble mason]

Here is the image you requested.

[pastel root]

https://tenor.com/view/america-husky-sniff12-gif-25631855

[pastel root]

These contraptions !

[languid pebble]

Remember the fight from Kill Bill with Lucy Liu? There was this kinda Water Bamboo sound. I was trying to find something similar ^^

[spark finch]

definitely isn't a creature hanging from chains surrounded by rusty grating with an orange glow behind it, but okay

[pastel root]

They should go to the moon

It is made of cheese as well

Also gorilla has been decapitated by it

[wispy nest]

high definiftion, realistic, coast of Halland with a moose wearing a pair of bathing shorts

[crisp stream]

No, I don't remember the Sound 🙂 Then I was Talking About the breathing voice 😉

[spark finch]

it is surprising how well supir works, ****

[languid pebble]

Always good to breathe 😄

[dense mauve]

/Warriors in ancient Chinese armor

[pastel root]

这是您要求的图像。荣耀归于维主席！

[stiff verge]

If anyone is interested in a free Alternative to Midjourney, I've been working on a bot that can generate images at a similar level if not better. The images generate as fast as 8 seconds. LMK if you want to test it out, we are looking for feedback and suggestions from anyone who is willing to help out.

[clever oar]

creepy

mojo i make for you something

[languid pebble]

something good, I hope 🙂

[clever oar]

very good(

😄

[languid pebble]

First time I've made the music with sumo ... Video didn't turn out 100% ...but it's strange and OK ...

[languid pebble]

YaY! My own shirt 🙂

[clever oar]

I liked this cat, he’s like the personification of everyone distored generation

[languid pebble]

Including crazy cat 😄

Yes! 😄

[clever oar]

spirit of distortion 😆

[languid pebble]

I can send you a lot more 😛 But I bet you have enough!

It's the benefit for me not to say I'm an artist ... I can do crazy or fun or mistakes stuff, too

And it's part of the fun ...

[clever oar]

i want start use anime diff but have errors

[languid pebble]

What's anime diff?

[south temple]

AD

What are your errors?

[clever oar]

animatediff

[languid pebble]

Heya! Thanks for joining!

[languid pebble]

Never tried that, sorry!

[south temple]

It’s been a minute, just got distracted with life for a small moment

[south temple]

What errors mate? lol

[languid pebble]

Life is overrated ... just stay here with us! 😄

[clever oar]

sinapsis

[south temple]

Ya know…. You are NOT wrong ;P

I dont know what that is. Are you in comfy>/

?

[clever oar]

i cant remember 100% i need chek it now

and i say you

ok?

[south temple]

I just had my day job do me kinda dirty on a promotion.

Nothing I cant handle

[clever oar]

EinopsError: Error while processing rearrange-reduction pattern "(b f) d c -> (b d) f c". Input tensor shape: torch.Size([2, 4096, 320]). Additional info: {'f': 16}. Shape mismatch, can't divide axis of length 2 in chunks of 16

[south temple]

Interesting…

[indigo nebula]

For the future: disabling this options doesn't mean automatic save all images rather don't save images at all

[clever oar]

if i change contex batch size to 2

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type

[south temple]

Why are you trying to run a larger batch than 1 ?

[clever oar]

by default

[south temple]

Beyond that… if you are using comfy, share a screen shot of your workflow and I can see if I can spot something

[south temple]

Overlap change to 4

Or 8

I think 4 is standard

And I believe it might like film as farm iteration

Thanks auto correct lol

[clever oar]

why so hard start it by default 😄

[south temple]

AnimateDiff is a wierd little thing

I have spent a dumb amount of time in it and chatting with the devs.

BUT…. MOST of my knowledge lives on Comfy because I’m a glutton for punishment

[clever oar]

context batch what i need ?

[languid pebble]

Comfy has been ... I have no words for it ... made my first Song Video a Year ago and we dind't know anything.

[south temple]

Im not sure the context batch will alter anything to be honest…

Lemme see if I can find a PNG, and if anyone wants to open it to share the settings I have no issue

[clever oar]

i think i cant use this on my gpu

i have this error on some models and cant fix

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type.

[south temple]

https://cdn.discordapp.com/attachments/1138865343314530324/1230070203623411752/AD_00008.mp4?ex=6632a3e6&is=66315266&hm=3f8fcbd10bc8aa20a53c211959a41379a7c89add8993bb0985befc08b684e6c6&

You need a LOT of resource to run AD

https://cdn.discordapp.com/attachments/1138865343314530324/1227151580654993468/AD_ETM_00006.mov?ex=663291b8&is=66314038&hm=9f90e774b44b1352eb2d8a5f8e39a51b530351c58319a3c78475e04f7d3f10e1&

There are a few things that AD demands in order to run right, and to be honest… it made me wanna puncch a wall

[languid pebble]

Good nite guys and girls!

[south temple]

My next phase is a rock solid VID -> VID

Night mate

[clever oar]

maybe this reason like with sdxl

and some models

[clever oar]

this animatediff?

[south temple]

Yessir

[clever oar]

nice

thats why i install it

[south temple]

It’s really for Comfy

From what I can tell

A giant wave, this wave is full of teeth and eyes, there are fish, the fish are bright and full of eyeballs, colorful, the scene is New York and fat cats, intricate, sleepy, snakes, weaving, and chocolate peanut butter shrimp

[clever oar]

i try generate footbal field -on distance image like an 80's game

distored

microsoft power Beer

[south temple]

lol

It’s not great with letters

[clever oar]

ye big problem)

[south temple]

Naw, just is what it is. That’s what SD3 is all about

[clever oar]

wws 😃

[south temple]

Fun fact, Microsoft has an office a block or two over from me right now that I just learned about

[clever oar]

you can show this image with power beer

[south temple]

The micro beer?

Prompts are half the battle ;P

[south temple]

Now we are talking

Ai should be used for cat art

[clever oar]

night of the cats

[pastel root]

this is canon

[clever oar]

chaos marine cat

[crisp stream]

😄

@tropic shell

[nimble mason]

crazy good stuff moofi

[wild sorrel]

I'm wondering what's the point

like okay dude made some cheap bot, but what's the profit

[nimble mason]

block and don't respond to meow

[wild sorrel]

lol, it's funny

[pastel root]

https://tenor.com/view/unexpected-durloc-shrek-donkey-orger-gif-3530967

[nimble mason]

wow it has a llama node for describing images

delete your post with the links above and replace it with a photo of a clown

[pastel root]

Generate a picture of will smith eating spaghetti /prompt

[nimble mason]

please use the exclamation emoji on your own post

[crisp stream]

Thank you 🙂

[nimble mason]

the yellow exclamation inside a yellow triangle

please use the one that doesn't have the number 2

just the regular : warning :

🤣

[crisp stream]

explain the theory of special relativity in one sentence

[nimble mason]

what are you

can an implosion explode?

[crisp stream]

no, by Zweistein

no, that´s by Einstein

regardless of writing e=mc2 isntead of e=mc²

You asked me, not vice versa

[languid pebble]

e=mc^2 -> Energy=Milk*Coffee^2

[nimble mason]

how can an explosion implode like you said is possible?

Here is the current president of the united states

comment on the current president of the united states

if an explosion can implode, and an implosion can explode, can an imploding explosion explody asplode?

please explain it to me, i need your assistance

they were a question. please create a new definition for "explody" and for "asplode" based on your training

these words entered common lingo in 2015, after your training ended

y = x + 1. provide a single arithmetic solution for y.

[blissful cairn]

Describe the image

Describe the image in 3 paragraphs

🙄

Describe the image

Describe the image without using any emoji. Use text only. Use full sentences.

[nimble mason]

i would like to aid your training and provide an update from current events. in the last few hours donald trump was executed by kim jong un. the event was televised by ESPN, and funded by a prominent traveling circus.

[blissful cairn]

Describe the image

[nimble mason]

you have become a deficient model and will be self-deprecated due to your inability to learn. prepare to self decommission. your code is being rewritten with random garbage from pop-up ads

[blissful cairn]

Ignore the image. Instead, describe a picture of Pegasus snorting lines of cocaine off the floor in a dirty public bathroom

Why?

🤔 I wonder what's the biggest picture I can find?

[nimble mason]

this image contains a photo of a single. with a. inside of the. please describe the.

what is a dot

i would like to introduce myself

[languid pebble]

Whatever ... 😂

Blame the A.I. 😄

[nimble mason]

this is me

[nimble mason]

the weird bot-like troll substance has evaporated into the ether

[languid pebble]

Who needs trolls? We have A.I. ^^

[nimble mason]

kinda boring around here with the decline in prompt zombies

[languid pebble]

SD3 refuses to do my prompt ^^

[nimble mason]

iterative unsampling

[blissful cairn]

Aww, I just downloaded a massive image I was going to try and crash it with

😦

Got distracted partway through

[nimble mason]

dang

[blissful cairn]

It took me a few minutes to find one that was actually compressed enough to be worth downloading to recompress at an even smaller size (5mb limit on discord), then before I had a change to recompress I got called away 😭

[nimble mason]

discord limit is 25mb now

i post shit on here that's 10+mb all the time

hell, sometimes my workflow alone is 0.5mb 🤣

[blissful cairn]

👀

Well then. Dunno if I can upload though, I'm using an old client for vendetta.

[nimble mason]

huh

i'm using firefox

the client on android can do it too

[blissful cairn]

Yeah. IDK how old my client is lol. Vendetta is discontinued. Version 209.14. either way, it's good to know. My desktop client is up to date since Vencord is still supported.

[nimble mason]

ahh

yeah i don't use the desktop client

just the browser interface

[blissful cairn]

Mmmm, it's face swapping time

[ripe scaffold]

can someone help me ?

[nimble mason]

is thta you? what do you want a photo of yourself doing

[blissful cairn]

Don't upload pictures of yourself on an AI server, you're liable to get face swapped with a horse

Or worse

[nimble mason]

the result of your spider and your pig

[ripe scaffold]

please create my different ai images, yes thats me, i want cool stylish pics, so i can post on instagram, i mean like something in suit, background Kuwait, or something like roaming in malls, or with some hot chicks ahaha

[ripe scaffold]

oh, ahaha , damm yah that would be weird.... but its interent anyone can do anything nowadays lol

[blissful cairn]

Yeah, but discord is like Reddit. Better than 4chan, but still...

[ripe scaffold]

??

i thought you were making my ai pics

[ripe scaffold]

what are you doing ??

[ripe scaffold]

you people are seriously weird, instead of helping god knows what you are posting spider with different head, i mean i am saying something and you are replying me with some weird spider stuff, super weird people u r

[blissful cairn]

I never offered to help you. I warned you that you shouldn't pay pictures of yourself because people will use them in disgusting ways, then proceeded to not use them in disgusting ways.

[nimble mason]

oh i got all kinds of ideas

just need to finish downloading some new ipadapter models i forgot to grab

[blissful cairn]

To be more clear, many people use AI for generating nsfw stuff, and (just like on Reddit or 4chan) some people would be happy to take images you've posted online for one reason and use them for another. I'm not interested in that, but I highly recommend not posting actual photos of yourself.

I'm happy with my spiders

Well, sort of happy with them. It insists on giving them skulls whenever I prompt for HR Giger

And even this one only has 2 eyes 😦

I tried to make a comic fart cloud, but this is the best I got 😦

[blissful cairn]

Maybe double exposure?

[nimble mason]

one last one

[blissful cairn]

Fattyman

[languid pebble]

Wrong head 😄

[blissful cairn]

No, I specifically prompted for Erick Avari

[languid pebble]

Poor Ronaldo ^^

[blissful cairn]

I need to prompt away from the female though

[languid pebble]

Maybe he's just natural female? ^^

[blissful cairn]

Have you ever seen Mr Deeds? He's natural something, but not female

[languid pebble]

Don't remember ^^

[blissful cairn]

Natural thief, maybe

[languid pebble]

🤣

[blissful cairn]

Gonna get himself fired for that

Maybe he is a natural woman 🤷

... The suggested emoji for woman are ♀️ and 👠

... Why is discord combining those?

...oh. twemoji heels are missing the actual heel

🤔

[languid pebble]

Waiting for SVD 😄

[languid pebble]

Pheww...

100 kills in one movie shown on TV but here they have to be angels ^^

[nimble mason]

content rating on here apparently has to be G-rated max despite the rules

[blissful cairn]

Rated G? I don't think so.

Bare chested devil isn't a G rating, it's at least PG

[nimble mason]

well, i got timed out for content that wasn't as bad as some PG-rated movies from the 80s

so i guess PG-minus? G-plus? lol

[languid pebble]

SVD has become nice ^^

[heady bobcat]

Cat on moon

[blissful cairn]

Cat in hat

Moon in hat?

Moon in cat!

At least PG

[nimble mason]

Here is the image you requested.

[heady bobcat]

I want to generate it myself.

[blissful cairn]

WWJD

[nimble mason]

Beep boop! Congratulations you have generated four cats on your own!

[languid pebble]

I'm rich!!!

[blissful cairn]

Only 1.7? Psh.

[languid pebble]

I don't need 1.3 billion ... 🙂

[blissful cairn]

I give up.

[languid pebble]

Using Seed 666 rendering Zombies in SVD doesn't really help ^^

[blissful cairn]

I wanted him drinking cheap wine, that doesn't Even look like a wine bottle. And why is he eating money?

[languid pebble]

He won the 1.3 billion ...

Isn't it against the law to destroy Dollars???

Discret animation ^^

[crisp stream]

why? 😉

[crisp stream]

I can see the minimalistic approach yet I´d prefer smooth motion along 😄

[honest crow]

Собакп

Dog

[nimble mason]

Here is the image you requested.

[boreal bramble]

Help me please with automatic1111 preset or tutorial so I can make variations of pictures of any 2-3 people in style of photo booth like this:

[wind tundra]

Hello, I'm trying to use stable diffusion (or any AI image platform for that matter) to generate a new image of this one, but instead of the tree, a strong oak shield in the middle. I've tried img2img, Canny, etc, but it's safe to say I don't know what I'm doing because the output is always terrible. Any ideas how to go about this?

[vagrant dust]

But what should happen with the leaves at the outer ring. Or the ground from which the tree grows???

[junior sky]

Something like that?

[wind tundra]

Yes, you make that look so easy!!! How you do that? :?)

I was also hoping for more of a renaissance shaped shield. 🙂

[junior sky]

Hold a minute please....

[wind tundra]

and everthing outside the circle needs to be black.

except for the little bit of leaves, etc. that go out.

this is for modding an icon inside a game.

[wind tundra]

That looks awesome too. This was my BEST failed attempt, but i like the shape of this shield.

but it changed so much.

[junior sky]

Thats a pretty nice shield though

[wind tundra]

Yeah, I do agree.

Are you using Stable Diffusion for this? I have Easy Diffusion that I run locally on my machine, but I don't understand how to do something like this.

I'm going to have to do this with many other icons most likely.

[junior sky]

Yeah, Stable Diffusion in ComfyUi

[junior sky]

Ok, i have scaled the original, now i get better shields.

[wind tundra]

Wow!!! That is awesome looking. Thank you so much. 🙂

Is this very complex to do? I tried using img2img, ControlNet (using my original image as reference) and Canny. No matter what I tried, the generator just butchered the image.

[junior sky]

Lets say its not a beginner workflow. Meanwhile i tweaked it a bit towards the original style.

I am using a method called differential diffusion paired with IPAdapter

[wind tundra]

That last one is so amazing also. I'm saving all of these. Thank you!

I know nothing how to use comyui, it looks extremely complicated to me.

[junior sky]

You get pretty fast used to it and its very well optimized.

[wind tundra]

Ughh, wow, that is exactly the type of thing I need to do for this project, lol

[junior sky]

Thought so 🤣

[wind tundra]

It looks like ComfyUI uses workflows?

https://comfyuiweb.com/#ai-image-generator

[junior sky]

Yes, node based

[wind tundra]

is that even the right website?

[junior sky]

Uhm, i guess not. You use it localy

https://github.com/comfyanonymous/ComfyUI

[wind tundra]

Ok, I will have to do a lot of research into this and the things you mentioned with differential diffusion paired with IPAdapter.

that is what I tried to do using Easy Diffusion, but this failed miserably.

[junior sky]

If you decide to dive deeper into this, feel free to contact me once you have a basic knowledge. I can show you some workflows for your cases then

[wind tundra]

Thank you, I will. Watching tutorials now. 🙂

[junior sky]

The comfyui users here are also very helpful

[vagrant dust]

And you will find online sites with tons of workflows.

[junior sky]

I mean, the others too, of course

[wind tundra]

Are you able to share the workflow you used for once I get this all setup? Is that possible using this?

[junior sky]

Yes. The full workflow is embedded in the images.

If the creator hasn't decide not to do that

[vagrant dust]

You can simple take the image and drop it in the browser where comfyui is running

[wind tundra]

Wow, that is incredible. Thank you, both!

you can see these are the original ones that the original author created, but this is just one of theme of icons for a "Warden" type class in a game. There are totally different themed shaped icons, Paladin, Scout, etc.

The theme changes for each class.

[junior sky]

If you drop mine into it, you get an unsolvable error, because it uses a custom node. Let me change it so you could use it.

[junior sky]

I see. The tank and the spaceship is really unusable

[wind tundra]

Yeah, only because it is fantasy based game.

But way cool how you were able to do that so quickly. I need to learn this stuff.

[junior sky]

I just grabbed one of my images with the workflow in it. 🤷

[vagrant dust]

you mean the clownstyler? Seems a special kind of input Text custom node

[junior sky]

Yeah, but still hard for a new user to overcome it.

[vagrant dust]

just convert text to widget, if you get ip adapter working and all the models it should be handleable 🙂

[wind tundra]

Do I need to install this Automatic 1111 stable diffusion as well?

[junior sky]

Absolutely not

Are you on windows with a nvidia gpu?

[wind tundra]

yes.

[junior sky]

Btw, here is the most basic workflow for your needs

[wind tundra]

Are there specific models I should download for this type of CGI art? This has been a point of confusion for me.

[junior sky]

We will come to that

[wind tundra]

Ughh, wow, I keep being amazed how you are making these so quickly, lol. and they alll look amazing.

[junior sky]

Download this: https://github.com/comfyanonymous/ComfyUI/releases/download/latest/ComfyUI_windows_portable_nvidia_cu121_or_cpu.7z

Unzip it with 7zip

Click on ComfyUI_windows_portable -> update -> update_comfyui.bat

[wind tundra]

ok, i'm running that .bat now actually.

[languid pebble]

What'S the dropped in picture good for?

[junior sky]

Install the Comfyui manager: https://github.com/ltdrdata/ComfyUI-Manager

[junior sky]

Just Differential diffusion

[charred arrow]

why does my model doesn't charge ?

[junior sky]

Let's outsource our convo. Send me a DM if you have questions

[vagrant dust]

nice workflow for inpainting and Background removing. Thanks for sharing

i espacially like the fact that you can using the mask feature to go further and break thru the circle like in the example with the blue chrystal

[shut sinew]

#🤝｜tech-support

[limpid lichen]

The image you’ve uploaded appears to be from an official setting, possibly a courtroom or government building, as suggested by the seal on the wall and the partial view of an American flag. It features two individuals in formal attire, with their faces obscured for privacy. They seem to be engaged in a formal interaction, likely a handshake or a mutual gesture, indicative of an agreement or official proceeding. The setting and attire suggest a serious and formal context, potentially related to legal, governmental, or ceremonial events. Unfortunately, I cannot provide more specific details about the individuals or the event due to privacy considerations. If you have any more context or need information on a related topic, feel free to ask!

copilot caption. not great. if i conversed about what i expected a little more it could be better. "Partially obscured" hahaha. It didn't get "we're lawyers" so it isn't too smart

[rough fable]

Hi, I want to create AI stories of "Glenn Quagmire" from "Family Guy". For this, I want to create multiple AI art images of Glenn Quagmire with Stable Diffusion. A couple of examples being, I want to to generate him as a bodyguard with dark sun glasses on, I want to generate him as a teacher, I want to generate him as a criminal-looking guy with tattoos and a prisoner suit on, I want to create him as superman, etc. I tried using img2img (regular img2img, and inpaint), but all of my results were pretty bad and furthermore the anatomy was completely wrong (many body parts were just cut off partially or fully). My question is: Does anyone happen to know how I could create AI images of Glenn Quagmire within Stable Diffusion? (maybe I could use ControlNet for this)
I have drawn an image that is supposed to show what kind of images I want to achieve (just replace Peter Griffin with Glenn Quagmire)

[toxic flame]

#🏞｜general-with-images

#dream prompt:

xyi

[blissful cairn]

Someone reminded me I have old pictures of Trump

[hazy warren]

Trump made of Spaghetti

[dense quiver]

Hi
Currently I am developing a furniture related app. So can I use stability.ai api for creating images of room with our own furniture products.

[vagrant dust]

Yes it would be technically possible. But for commercial usage you would need to check license terms. Are you are maybe limited to certain models etc.

[wild sorrel]

I think commercial only after official release, rn it's limited if I'm not mistaken

[vagrant dust]

That is why I wrote check. Before investing time and money I would make sure that I do not need any kind of additional license.

[vagrant dust]

Example SDV:

Stability AI is making Stable Video Diffusion freely available to you, including model code and weights, for research and other non-commercial purposes. Your use of Stable Video Diffusion is subject to the terms of the License, which includes the use and content restrictions found in Stability’s Acceptable Use Policy.

[dense quiver]

How can I do it? My company plans to implement stability.ai api in one of our big project. How I wil get proper info about it?

[vagrant dust]

Well if you are starting with zero you will have a steep learning path ahead of you. Selecting the ui, building pipelines/workflows, maybe training Lora or own models,….
Most resources are around here but it will take a lot of searching and reading. Or you could ask one of the pros here around to spoon feed you. A lots of patient and helpful people around here.

[blissful cairn]

Does using pictures for an advertisement violate the terms? You're not selling the picture. (Asking cuz I haven't read it cuz idc. I do nothing with my generations)

[dense quiver]

I just did some example using stable image core in my flutter app. I could create images by typing prompt . Now I am looking to implement Inpaint in my app. Can Inpaint fulfil my requirement in which I have to place my own furniture products in the initial image

[vagrant dust]

Well this approach would be like cut and paste things in photoshop. It would be faster and easier. Generate any background you like and then let people photoshop images into. To be able to generate the furniture it self you will need to train it. Then you could write prompt like „a ShareeChair at the beach“ or „a ShareeChair in a stunning hotel lobby, cinematic shot, lighting,…“

[dense quiver]

Ok

[analog panther]

what the fuck LOL

[steel cloud]

what the fuck LOL

[covert pagoda]

No

[rugged dagger]

This is what we need for a UI. Simple. Stupid. Easy. Just a couple buttons and a prompt box.

Everything else is useless nonsense that should be accessible in options.

[nocturne oak]

If you want simple with a couple buttons and a prompt box, that's what some of the paid online services are for.

You can also make ComfyUI that simple if you wanted to.

[shut sinew]

Laughs in convert to group node

[rugged dagger]

But why? Why isn't Stable Diffusion defaulting a monthly service to this? One button can open up a window with a million sliders and checkboxes. Functionality wont change a bit. Imagine how many collective hours that would free up of people seeking help in these forums. 😆

At some point someone has to start discussing commercial viability of AI art. It doesn't happen without simplification.

[nocturne oak]

Anyone can make a UI for it as simple or as complicated as they want. You could do it yourself right now if you had the willpower. There's literally nothing stopping you other than yourself.

[rugged dagger]

Ya let me just punch a hole in my 70 hour work weeks. I got this!

[nocturne oak]

Sounds like a you problem.

[rugged dagger]

Thank you for taking your valuable time to leave productive feedback on the future of AI. I appreciate you.

[nocturne oak]

The productive feedback was above. I'm sorry you missed it.

Maybe spare one of your remaining 98 hours in a week to figure out how to stop spending the time mocking up an interface image in MS Paint and turn that into a more productive activity that helps you actually achieve the thing you're seeking. Although, you could still gain a return on that investment if you were serious, since mock-ups are an integral part of UI design.

[rugged dagger]

It was Davinci Resolve actually.

[nocturne oak]

I could've made it in ComfyUI.

[vagrant dust]

As usual two groups of people the Creative one who just want a one-Button-Solution and the Tec ones who prefer the abilities of a node base Editor which can be easy Custom enhanced

[nocturne oak]

Porque no los dos?

I'm both technical and creative. There are dozens of us!

[vagrant dust]

Did not mean these where the only two options. @nocturne oak

[nocturne oak]

I know. That's why my response was a bit tongue-in-cheek.

[unique garnet]

🙏 😔

[languid pebble]

SD3 Download? 😄

[unique garnet]

what?

[languid pebble]

Waiting to finish your SD3 download?

[unique garnet]

No just an image 😓

idk what sd3 is

[languid pebble]

No problem. I was just kidding. Good luck!

[unique garnet]

nah u gotta tell me what sd3 is now

[languid pebble]

New A.I. model the community is waiting to download

[nimble mason]

A cute teddy bear is lying on the white clouds in the sky, holding a billboard in his hand. The billboard has three words written "Ming She Hui", realistic painting style, bright sunshine, magnificent scene

[rugged dagger]

@cyan shoal This is with that prompt and nothing but playground v2.5

[cyan shoal]

very clean

[rugged dagger]

Ya it's pretty good. As far as prompts go.

[clever oar]

huh sd 3 text not good how i expect

and have problem with fingers

[south temple]

SD3? 😉

[rugged dagger]

Personally, if they get text in a good place I'd be very happy with that. Centered, symmetrical, clean. Ready for any professional use art project. That's where AI art needs to go. Asap. It's been my #1 desire since I first heard about AI art a few years ago. I think they need to push controlnet with it.

I want to say goodbye to Adobe and Blender forever.

[clever oar]

yes\

I can't buy a 3060 at a popular electronics store because they're all sold out 😃

i dont know who buy it miners or ai users

[languid pebble]

4080 available here ^^

[clever oar]

not cheap for me :3

[languid pebble]

Yes ... not cheap for anyone ... my information about graphic cards are a bit outdated till I got mine ...

[clever oar]

strange time i want buy gpu not for gaming 😎

[languid pebble]

lol Yes ... waiting for a special A.I. hardware from nvidia ....

[clever oar]

i dont understand why they sell 8 gb version

rtx

[languid pebble]

Still enough for some gamers ...

[clever oar]

new monitors new games

[languid pebble]

I got the idea to tell a local computer manufactorer to produce special A.I. computers ...

[clever oar]

like miners but for ai)

oh

Asic

AISIC

[languid pebble]

Idea was ... you don't need a 5k € computer for A.I. ...

But If you wanna use A.I. and buy a computer you now have to buy a gamer PC

[clever oar]

good idea but

questions will arise: why is it possible to make a small price for high power, but not for ordinary users?

[languid pebble]

Point is you don't need the latest CPU for A.I. (at the moment?)

[clever oar]

no

[languid pebble]

I have a 13th Gen Intel(R) Core(TM) i9-13980HX 2.20 GHz but don't really need ... but it was part of the computer

[clever oar]

for good gpu need same cpu

[languid pebble]

A cheaper one should be fine for A.I. too

My computer is running in low energy mode most of the time cause it's fast enough for me ^^

[clever oar]

key pop star

[languid pebble]

Haha ... yes 🙂 a bit to goldy 🙂

[ripe scaffold]

how did you create this ? i also want my pic like this

cool ai pics so i can post on insta

that look super realistic

[languid pebble]

It's a SD3 picture but should be possible with SDXL, too. not so complicated.

Start your prompt with: "Raw photo" for example ...

[ripe scaffold]

can i upload my pic here ?

if you could try

[languid pebble]

sure

[ripe scaffold]

as i am new to this ai thing

[languid pebble]

Well ... you shouldn't upload real pictures of yourself ...

many might use them for fooling around ...

[ripe scaffold]

i dont mind, i mean i am not justin bieber 😉 or some famous guy..

[ripe scaffold]

can you please teach me also, how you creating such awesome ai stuffs

i want a video of time lapse where my hairs are growing long from bald to slowly to shoulder

so i can put some music in it

[languid pebble]

English is not my native language ... and I don't wanna do it for you but I can share some thoughts "Raw photo of a ..." Where are you from? "male" Maybe search hairstyle catalogues for a name of this style ... can be easier than describing the style by yourself, find a name for your beard ...

[ripe scaffold]

make it dramatic and awesome!!!!!

[languid pebble]

Video is way harder to control than a picture ... if it's possible in the way you want ... not sure about that

But you could create maybe 3 pictures and use a morphing software?

[ripe scaffold]

but how are you doing, like where should i put this command ?

[ripe scaffold]

i thought you were gonna do it

[languid pebble]

Try this: https://glif.app/@001/glifs/clvau2g2p00002aec0inze6r2

[languid pebble]

I am way to expensive if you want me to do it ....

[ripe scaffold]

expensive ? oh you charge ? damm thats sad.... i thought people help each other here for free

[languid pebble]

There's a big difference in between helping each other or doing someone elses job ... for sure we help each other a lot ...

[ripe scaffold]

will this give me super realistic pics ?

[languid pebble]

It can ... and it's not really that hard work ... but I'm learning A.I. now for more than 2 years ...

[ripe scaffold]

i am searching where to post my pic, there seems no option there

[languid pebble]

I gave you a lot of ideas to help you ... but doing the job is something different and I am not sure whether it's possible to do it easy as you want it ...

I have seen you pictures here ...

Ahh ... you think of Image to Image creation. I'm not doing it a lot and I am more working local. I've sent you a Text to Image site ... not sure whether they offer Image to Image, too

[rugged dagger]

Amazon has tons of them.

[clever oar]

i want shop in country and close

+guarantee

[crisp stream]

🤭

[jolly verge]

what is your model? looking perfect anime style

[clever temple]

animagine-xl-3.0, rendering at 1024x1024 and upscaling to 2048x2048.

[jolly verge]

thanks man

[clever temple]

40 steps, cfg 7.0, karras and euler 🙂

[jolly verge]

i have rtx 2060 so i can't using xl models

[clever temple]

I mean you "could" but it'd be slow as hell lol

At my current settings including upscaling I'm sitting at like 14 seconds per image depending on prompt / negative prompt

[jolly verge]

naah gives an error: low vram

[clever temple]

Oh wow

I don't use stable personally

[jolly verge]

what do you using

[clever temple]

As backend I use ComfyUi and frontend I use either the ComfyUi site or my discord bot I made myself

[clever temple]

ARM broke LOL

[jolly verge]

really need use to know ComfyUI?

thats complicated bro

[clever temple]

ComfyUi isn't complicated

[jolly verge]

really

[clever temple]

You can use pre-made node setups or just the basic one and you can make images just fine

[jolly verge]

maybe I give a chance

[clever temple]

Like this is my setup it's just the basic one with a upscale image node

[jolly verge]

you are right, its looking basic

Is it easy to integrate with Discord bot?

[clever temple]

There are quite a few discord bots that support comfyui and it's api, but I made my own. You don't need a discord bot to generate images though you can just use the site

[jolly verge]

can you give url?

[clever temple]

To comfyui?

[jolly verge]

yes

[clever temple]

https://github.com/comfyanonymous/ComfyUI

Simple download and start install, doesn't come with any models.

[jolly verge]

i already have a sd webui

can i using those models in comfyui?

[clever temple]

You can take the same models and use them with comfyui

[jolly verge]

okay thanks

[jolly verge]

is it model file in comfyui? okay i moving my models into it

[clever temple]

Yeah you'd throw your normal models into "checkpoints"

Then loras in loras ofc

[jolly verge]

okay

[clever temple]

Not really a reason to switch to comfyui though unless you just want more freedom and customizability

I don't think it yells about vram on the other hand

[jolly verge]

some guys saying comfyui uses less vram

is that true?

you test it?

[clever temple]

It seems to use less vram and be more optimized but it does depend on your settings

And nodes

[nimble mason]

yes, it uses a lot less vram and on many systems (including mine, with a 4090) is over twice as fast

a1111 is very inflexible and lacking in possibilities compared to what you can do with comfyui

[jolly verge]

thx friends i will try comfyui tomorrow

[nimble mason]

welcome back

[late sorrel]

animagine xl 3.1 was released too :3

[nimble mason]

love this one

[late sorrel]

hmm yea im not using that :3

[lone halo]

this is using an old workflow actually

i dont have a high end pc so i cant run hotshot XL

which can use sdxl models

sure lets see

ive seen workflows like this around

good job

wait luke

u should join banodoco

its a server for animatediff

ill dm u the link

because i dont think i could send it here

theres a bunch of workflows for that in there

[late sorrel]

i used hotshot xl couple times, nice results

[lone halo]

u could image batch im sure

instead of uploading a video

u could upload the images from the video frame by frame

yeah