🏞｜general-with-images

[celest sigil]

CogVideoX from Flux image

[wispy nest]

if you want some comedy try asking CogVideoX to make R2D2 move forward

[sterile kiln]

This kind of image is openpose already preprocessed. Preprocess is for normal images with chars.
Just disable preprocess if you submit stick chara shapes like this

[random yarrow]

After a bit of a lull in open AI-video workflows, I love the new unsampling methods bringing new life to animatediff.
It allows for really expressive styles compared to commercial models - everything with SDXL:
https://www.instagram.com/reel/DCjPwlLtoh1/?igsh=Njd3NTVweTA0d2d4

juhoristoaukusti

Long time no (AI)-post! I’ve been building new expressive animation workflows for a couple of upcoming projects, so I decided to quickly edit something from the terabytes of footage on my drive. After a bit of a lull in hybrid animation—caused by the industry’s focus shifting more toward commercial video models—there’s a lot of new life in AI-an...

Instagram

[clever oar]

👋

[clear whale]

Please check out my new ai channels on Facebook insta and YouTube. A.i.AnimationHallucination- thanks. Steve. Here’s a praying mantis dancing!
https://www.instagram.com/reel/DCkblatNEdf/?igsh=MTlleGxoa3ozNjdmcw==

a.i.animationhallucination

#ai #animation #aianimation #prayingmantis #alien #dancing #psychedelic #trippy #comfyui

Instagram

[fleet goblet]

rabit

image to image moment (below is original image)

horse side-eye

[clever oar]

stop scare me i need sleep

[clever oar]

shark my ip adapter not work now -out of memory

sad

what did they update, what is now such a memory leak

[nimble mason]

comfy?

[clever oar]

a1111

last time i play with install cudnn and update automatic

it works one time

second generation not work

🙁

[nimble mason]

ouch

a1111 blows

[fleet goblet]

https://tenor.com/view/vsauce-gif-233952886679047004

[viral frost]

Don't just stand there! I'm going to drown in love hearts!

[nimble mason]

exactly

[fleet goblet]

Um.

Hmm.

The initial prompt I typed, which was fed into Llama 3.2 3b 8_0, was "The face that has seen what has not been."

The AI responded with "Write a poem from the perspective of the Mona Lisa's husband, Girolamo dei Rossi, about his wife leaving him and hiding the truth behind enigmatic smiles for centuries." which was the input prompt for pixelwave flux dev

And the output, was the mona lisa.

The Mona Lisa. The Face that has seen what has not been.

[rough shell]

iphone 14 pro

生成一个图片

um

[proper orbit]

Alfaria from wistoria sword and wand X Hutao from all well known genshin impact. Made with Comfyui

[heady lodge]

Does Flux look kind of uncanny to anyone else

It's halfway between completely realistic and plasticky fakeness

[scenic yew]

Yeah it’s trained on some synthetic images like mid journey which have the same exact skin. You can usually fix this by using loras.

[candid pollen]

Yes. I find I just don’t like the aesthetic at all. Sure, it may have better prompt adherencr but what good is that if the result always looks ugly?

[heady lodge]

Do realistic loras mesh well with loras of a specific person

[scenic yew]

Probably, but I can’t say for sure since I didn’t try it.

[scenic yew]

You can use pixel wave v3(flux finetune), I actually like the styles you can make with it. Realism is decent too.

[candid pollen]

The quality vs (ridiculously long) generation time vs cost tradeoff just isn’t worth it

[scenic yew]

Just use an 8step Lora or something, should take like 10-15 seconds for an image. Also, aesthetic can be easily changed with prompting.

And again, flux dev/pixelwave3 is one of the best models compared to closed source and open source.

[candid pollen]

Anything that interferes with other loras or controlnets automatically makes a model no-go for me. I don’t care in the slightest if a model is ”one of the best” if it’s not suitable for my purposes. I’m not interested in generic pretty pictures but specific controlled results that are fast enough for iterative generation.

[scenic yew]

The 8step Lora can still work with controlnets and other loras(I only tested other loras though). You can use something like svd quant to further boost speed, and use 6gb vram only without sacrificing considerable quality(similar to 8bit).

It’s understandable you want fast images, I want fast imgs too but with good prompt following and text and humans.

[sterile kiln]

because you're using small models, and settings that could be better.
For example :

[wispy nest]

SDXL/SD1.5 aren't that much faster than 4 step Flux tbh

[wanton dragon]

Yeah that image does to me but it still looks amazing quality in some ways?

[wanton dragon]

Random?

[wanton dragon]

Seems legit? XD

[clever oar]

what you mean?🙂

[wanton dragon]

It’s a joke, I was saying it looked real sarcastically? It’s cool that it mixed them together though.

[clever oar]

i have not ideal english not undertand all joke

[wanton dragon]

Fair enough

[clever oar]

its ip adapter and not work like before(

one generation and memory bug

[wanton dragon]

Strange, I get weird issues where it refuses to generate sometimes too

[clever oar]

it eats up a huge amount of memory

[wanton dragon]

Sometimes if I make the prompt shorter it works which is weird because I’ve seen prompts a lot longer then any I do before.

[wanton dragon]

What resolution are you trying to render it in?

[clever oar]

what I like most doesn't work🙁

[clever oar]

512

[wanton dragon]

Oh ok I’ve had issues with bigger resolutions, what about batch sizes and loras etc i think that can effect it too?

[clever oar]

no lora , all options default

[wanton dragon]

Is it an old gpu?

[clever oar]

1050 ti

but it works before

[wanton dragon]

I’m not sure about that one but it seems odd if it worked before?

[wanton dragon]

Did you change anything like install any extensions etc? That breaks mine sometimes

[clever oar]

yep i install some extrension but now i ininstall it i think

[wanton dragon]

It says your using too much ram

[clever oar]

i try delete venv folder

[wanton dragon]

Yeah I’ve heard that can help sometimes

[clever oar]

last what i change i install cudnn and cuda

[wanton dragon]

Could you still use it right after that ?

[clever oar]

it work good before

[wanton dragon]

No I mean literally after you installed those did it break then or work for a bit after?

[clever oar]

I can’t remember exactly, it seemed to work

[wanton dragon]

I’m not sure what to say then as I’m quite new to this, it does say it’s a ram issue though so something is using too much ram

[wanton dragon]

This “OutOfMemoryError: CUDA out of memory”

[clever oar]

armored cat😎

[wanton dragon]

Which model are you using ?

[clever oar]

epicphotogasm

1.5 base

[wanton dragon]

Ah yeah I’ve heard of that one

[wanton dragon]

What’s the prompt ?

[clever oar]

cat)

[wanton dragon]

Just cat?

[clever oar]

yep

[wanton dragon]

That’s weird then

[clever oar]

why

[wanton dragon]

It put a shell on it too.

[wanton dragon]

Ok that makes more sense

[clever oar]

because ip adapter on

[wanton dragon]

You got it working then?

[clever oar]

it works but only one generation

after that out of memory

[wanton dragon]

You mean one at a time or literally just one?

[clever oar]

yep

start second now

[wanton dragon]

Oh

[wanton dragon]

What about standard txt to image?

[clever oar]

it takes memory and dont back

[clever oar]

no problem

[wanton dragon]

How many can you do in a batch?

[clever oar]

dont know but i think enough

[wanton dragon]

Ok

[clever oar]

second

fail

[wanton dragon]

It definitely looks like some kind of memory issue due to the error messages but I don’t know why it would do it now if it didn’t before and you didn’t change anything since.

[clever oar]

I wonder why it takes memory and does not give it back if not used

3.8 of 4 gb now using without generation

[royal monolith]

how much space is open on your system drive

[clever oar]

8 gb

[wanton dragon]

Ah yeah didn’t think of that, it can course issues too.

[royal monolith]

that isnt' enough for your system to even run. clear that up and make sure you have at least 100G free

[clever oar]

gpu now

without generations

[royal monolith]

clear your diskspace up.

i'm surprised you can even boot that box right now

[wanton dragon]

100gb? I hope I don’t have an issue then my whole c drive is only 512gb ssd

[royal monolith]

he needs at least 50G to run just the system. and that gives a bit of padding

the thing to do is put the models all on a different drive and just point to where they are

[wanton dragon]

I didn’t realise it was quite that high damn

[clever oar]

I’m not sure that there is a connection, because during generation my 8 gigabytes do not decrease

[royal monolith]

will you please go clear up that drive

[wanton dragon]

I do that, I have about a 1tb of ai stuff in the ai folders

[clever oar]

its maximum what i can

[royal monolith]

we've had this discussion with you before. you keep having the same errors, you are in the same 'no space' situation, and you refuse to clear up the drive

[clever oar]

before i have 800 mb

i clear 7 gb

its maximum what i can do

[royal monolith]

how large is your system drive in total?

[wanton dragon]

800mb is definitely no go

[clever oar]

100 gb

[wanton dragon]

Sounds like that’s your issue then?

[royal monolith]

what sort of machine are you using?

[clever oar]

stable duffision on different drive

[wanton dragon]

How big is that drive?

[royal monolith]

your SYSTEM drive needs enough space for the swap to run, and other things that must use the sytem drive to run, whether you ahve stable somewhere else or not

[clever oar]

2 tb 100 gb free

[royal monolith]

when you get an out of memory error, it almost always points at not enough swap space

[wanton dragon]

Shouldn’t you not tell it to use part of the hard drive as ram if it’s an ssd? Unless I’m confusing it with something else?

[wanton dragon]

Damn is that all Ai stuff?

[royal monolith]

you can. not how things are setup by default though. swap's on the system drive by default. and other things will use the system drive whether you like it or not

[clever oar]

but the disk capacity does not decrease at startup, why do you think this is a problem?

[wanton dragon]

Oh I thought it would kill the ssd much faster ?

[clever oar]

no))

games

[wanton dragon]

Oh ok

[royal monolith]

what makes you think i'ts going to decrese just becaue you started the program?

you run the program, it starts trying to run, it finds that it doesn't hae enough space to run, it gives you an error

[wanton dragon]

Yeah also depends on extensions and loras etc right ?

[clever oar]

but it’s not the hard disk space that’s wasted, it’s the GPU memory

[royal monolith]

sigh. you might want to search google for your error and read through places like stack overflow

[clever oar]

i was on tech support channel and ask it

[royal monolith]

and i gave you the same answer there

[wanton dragon]

The error says “OutOfMemoryError: CUDA out of memory”

[clever oar]

he said it's not that important

[royal monolith]

here https://stackoverflow.com/questions/54374935/how-can-i-fix-this-strange-error-runtimeerror-cuda-error-out-of-memory go read through the posts there

[wanton dragon]

Wonder why they called it a strange error?

[clever oar]

i was -maximum what i can find -bat file options

[royal monolith]

strange to them maybe

[wanton dragon]

I guess.

[royal monolith]

your system does not have the capacity you are asking it for.

[wanton dragon]

That’s what I guessed but they said it worked before ?

[clever oar]

active gpu- 1 %
memory using -99%

strange

[wanton dragon]

No not strange memory is full

[royal monolith]

he's been in this discord for days with the same error

[wanton dragon]

Oh?

[clever oar]

can i clean it after generation?

[wanton dragon]

What do you mean ?

[clever oar]

memory gpu

[wanton dragon]

I’m not sure but it seems like you definitely don’t have enough memory to start with

[royal monolith]

yes. same error, same arguements, same won't fix the basic things, same stuff not working

[wanton dragon]

Damn

[wanton dragon]

Am I correct though? I’m quite new to using ai so some of what I’m saying is guesses based on my own experiences

[royal monolith]

"i have this pickup and i need to add this fridge to the back in the bed but i don't have enough room" 'take some stuff out" "but i know it'll fit, jsut tell me how to make it fit" "take some stuff out and make room" "no, i don't need to take some stuff out, just tell me why it won't fit when i have plenty of space in the front seat"

[clever oar]

i clean space maximum , cant do more , i change bat file i change resolution, dont tell him that i not do basic thing

[wanton dragon]

So yes then?

[clever oar]

i uninstall extension

[royal monolith]

you can do more, you just don't want to. you've got to clear up enough space for your sytem to run IF you want to generate images on it

[clever oar]

many thing i do after

[wanton dragon]

Sometimes you have to do more then you want to for stuff to work and even then sometimes your lucky if it does at all.

[clever oar]

cose i ask in tech support channel what help me

[royal monolith]

you asked, did you do?

[clever oar]

I did all the steps they told me to do there.

[royal monolith]

cuda doesn't have enough memory space and you don't have enough disk space for the amount of swap it would need. buy a bigger gpu

[clever oar]

before it works with same space

[royal monolith]

i hightly doubt that

[wanton dragon]

I thought it was the gpu even my 4050 6gb has issues sometimes and that’s much newer

[wanton dragon]

That did confuse me

[royal monolith]

an out of memory error almost always means you don't have enough swap. there are 2 solutions - get more memory or get more swap

you don't want swap for a gpu though, you want everything in vram

[clever oar]

incorrect comparison

[wanton dragon]

I got a better gpu 4070ti super ai 16gb

[royal monolith]

i've got a 4060 with 16g - never have cuda errors

(unless i kick off a massive batch file of 20)

[wanton dragon]

Nice, haven’t gotten to try my gpu yet the cable I need should arrive soon tho

[wanton dragon]

Not totally?

[wanton dragon]

I nearly got a 4060ti 16gb originally

[royal monolith]

that's what i have. it was on sale on newegg, which was nice

only problem i have is that one side is glass and you can't turn off the LEDs on the fans - which is annoying

[clever oar]

compare guesses and proven things

[wanton dragon]

Ah their £414 here and the 4070ti super ai 16gb is £889, got mine for £219 off tho as it had some slight damage cosmetically when I got it so didn’t look totally new, hope it works and I don’t have to send it back

[royal monolith]

here's the black and white - your system doesn't have the capacity for what you want it to do. you won't do anything people are saying to do, you have nothing but arguments about how you can't do things like free up space. so unfortunately you're sort of stuck

[royal monolith]

i'll take some cosmetic bumps for a 50% off price any day

[wanton dragon]

Na 25% still decent imo tho if it works ok

[royal monolith]

i'm happy with 25% too. i buy a computer to use, not decorate with

[wanton dragon]

Yeah it working correctly is what matters most so hopefully it does.

Plus it will be in the case anyway not on display so doesn’t matter visually

[royal monolith]

mine

cause of the LEDs

[wanton dragon]

Carnt really see anything tho

[clever oar]

8 gb its enought

[royal monolith]

exactly.

[royal monolith]

it is NOT ENOUGH

however, i'm tired of arguing with you about it

[wanton dragon]

Why take a photo then tho XD

[royal monolith]

to show you :) the box could have deep dents and cracked glass and it wouldn't matter - but it was in perfect shape so i didn't get a discount

[wanton dragon]

That does seem low for that yeah

[clever oar]

why before it works?

[royal monolith]

1. it didn't or 2. you upgraded some small piece of software or cuda or python or anything. it's not working now, we're telling you why, you won't listen. we're done

[clever oar]

i need only find software problem

[wanton dragon]

I asked about that

Well that conversation died suddenly.

[royal monolith]

sorry, phone rang and i had to answer it

[royal monolith]

that's the problem. you keep insisting is a software problem. it's not. it's your hardware. however i wish you success. maybe if you uninstall everything, clean up your system, move your models to a different drive, install fresh, and then try, you wont' get errors

[clever oar]

this is a complex of problems of both hardware and software, but if everything was fine before and the hardware did not change, you need to think in the direction of software

[royal monolith]

you are getting a CUDA out of memory error. Cuda is unable to get the amount of vram it wants. if you search google for the specific error you are gettin you will find a lot of hits. perhaps on some of those hits you will find a solution that'll work for you

[clever oar]

i read many forum about this

can you show me gpu memory usage when running stable diffusion without generation?

i think i have bug

🙂

[royal monolith]

you have a hardware issue

[royal monolith]

you can't 'run' stable diffusion without generating. all you can do is run an interface that will run stable diffusion

[clever oar]

yes its what i need

[royal monolith]

you've been given the reason for the error. you now need to do some google searches for the specific error and read through other forums

[clever oar]

can you tell me how much memory this takes from you without generation?

only interface

[royal monolith]

i have a 16 gig vram Nvidia gpu with well over 300gig of disk space free in case extra swap is needed - you have an 8 gig vram... something. any test i run will not do you any good

[clever oar]

just say it and that's it

[wispy nest]

sd3.5_medium-q3_k_m.gguf will take only 1.49GB of VRAM

and that size still works well

[royal monolith]

it

[wispy nest]

is this windows

there is an issue that affects windows but not linux

where loading a .safetensors file causes a big spike of ram usage

which is around 2-3x the size of the file

[wanton dragon]

They said and I quote “last what i change i install cudnn and cuda” just so you know. I did wonder if that was one of the issues ?

[wanton dragon]

No problem

[clever oar]

good joke wow

[wanton dragon]

Damn really?

[wispy nest]

yeah

[wanton dragon]

What’s a good joke???

[wanton dragon]

Damn and I don’t like Linux it’s way more confusing to use compared to windows and I’m pretty good at pc stuff normally plus not all software is on it.

[wispy nest]

Ubuntu is essentially easier than windows its just that people learnt windows first

[wanton dragon]

Really? Is it restricted in any way?

[wanton dragon]

Would it be worth dual booting my pc so I can use ai on Linux instead ?

[wispy nest]

has a level of restrictions that is actually zero

so its more than just being relatively unrestricted, the restrictions just don't exist in the first place

I'm talking about linux in general though for that, you would have to move off Ubuntu to get the purest form

I don't recommend dual-booting

[wanton dragon]

Why not ? I use windows for everything on pc

[wispy nest]

there's too many issues with it IMO

[wanton dragon]

Oh what like?

[clever oar]

maybe someday you will understand..

[wanton dragon]

What

[clever oar]

what is good joke

[royal monolith]

he said to 'say it' so i said 'it'

[wanton dragon]

I must have missed that as I didn’t see such a comment

[royal monolith]

:) it's there. just search discord for me and you'll see it

[clever oar]

Well, if you don’t catch the connection between the messages, then ok

[wanton dragon]

I didn’t

[royal monolith]

i'll try to explain this one last time. you do not have the hardware that is needed for what you want to do. and you DID say you upgraded stuff, so maybe before you did that you had just barely enough to squeak by. you do not now. and i have a whole lot more system resources than you do. me showing you anything that's going on, on my system, will not tell you what your problem is. which is that you don't have 1. enough GPUI 2. enough vram and 3. enough disk space.

[clever oar]

ok is simple
I asked how much memory consumption it has
and then wrote just say

[royal monolith]

i know. on my sytem it has 0 memory consumption as i load it all into vram with plenty of space to spare

[clever oar]

4 gb without reason its not normal .how many resources it doesn't matter

[wanton dragon]

Its starting to irritate me now too

[royal monolith]

and he's been doing this for days

[wanton dragon]

I wonder why?

[wanton dragon]

We tired to help

[clever oar]

Aren't you annoyed by the obsession with one piece of advice?

[royal monolith]

because he doesn't like the answer and is trying to get someone to give him the answer he wants, instead of the answer he needs

[wanton dragon]

No I mean we’ve being trying to help you and you don’t seem to understand or maybe not do the stuff we suggested might help?

[wanton dragon]

I dunno if that’s the case that doesn’t exist does it?

[royal monolith]

i had customers like this when i was at IBM - at a certain point you just escalate to the duty manager and get off the call

[clever oar]

Well, I explained earlier that I followed all the advice\

It’s strange that you were allowed to work with clients if you don’t want to hear people normally

[fleet goblet]

nomos8k is good

[heady lodge]

sorry but this does not look realistic to me

its close to realistic, like a really good video game

[wispy nest]

the nomos8k models are amazing yeah

[fleet goblet]

Leave.

Nobody wants these shit phishing schemes here.

[heady lodge]

Can Forge use nodes?

Or is that an exclusively comfyUi thing

[royal monolith]

it's a comfyUI thing

[wispy nest]

invoke is the other one, is on a different growth path but its node-based also

[sterile kiln]

Maybe try downloading it locally and scaling it 100% on your screen. Especially if you're on a smartphone, the image will be compressed.
The image is 3072x3072.
Then you also have this one. 3840x2160, focusing on skin details. Some small artifacts, but the realism is already quite good.

[scenic yew]

It does have an ai feeling, but the skin is pretty realistic. A woman could easily look like that. But I think when people want “realistic”, they want the low quality grainy amateur photo look while the image you posted is more “professional”.

[celest sigil]

There is too much shine and it has a painted look close-up.

[scenic yew]

Shine is normalish but yeah painted look close-up is true

[celest sigil]

She has scales above her eye.

[celest sigil]

@sterile kiln A slightly better version

A comparison between original and my version...

Even that looks plasticy 🤔

[sterile kiln]

what did you used to do this?

[celest sigil]

A bit better in this one.

[celest sigil]

This is my own upscaler workflow.

[sterile kiln]

HOnnestly i prefer original.

[celest sigil]

Really?!

Look at it full screen, at the lips and eye details.

[sterile kiln]

i did, its maybe more sharp, but not more realistic.

[celest sigil]

ok, we all see with different eyes 🙂

[scenic yew]

Crystal wizard did a test here lol, he took a real stock photo and people said it was ai generated because it’s too plasticky.

Many people could easily consider that as a real professional photo.

[celest sigil]

I think my image was too bright, but I've corrected that now. Looks good to me 🤷🏻‍♂️

[nimble mason]

yup that is the truth, that's why if you want a convincing fake, you wanna make it look kinda shitty

[celest sigil]

That's easy to do after, but getting some good details first is better.

For example... 😄

Then add some film grain...

[eager lintel]

share workflow man?

[celest sigil]

It's here...

[eager lintel]

is a picture man, link ?

[celest sigil]

The picture tells you where to get it 🤷🏻‍♂️

[south sonnet]

How would I go about adding a lora node into this workflow? Where would it go?

[scarlet sand]

"Photorealistic car driving on a scenic highway, multiple speed limit signs showing 30, 60, 70, 80, 90 km/h, responsible driving scene, clear road markings, sunny day, safe driving concept, high detail, ultra realistic --ar 16:9"
"Detailed dashboard view inside modern car, digital speed display showing 60 km/h, offline mode indicator with crossed out WiFi/4G/5G symbols, automotive interior, professional photography style --ar 16:9"
"Modern passenger van at Santiago International Airport (SCL), Chile, scenic Andes mountains background, multiple passengers with luggage, sunny day, ultra realistic, photographic style --ar 16:9"

[delicate wharf]

does someone know how to use new comfyUI?
the workflows aren't loading
I don't know why I updated
I just drag and drop And nothing happens now

[languid pebble]

Where did you get it?

[wispy nest]

could you try restarting

I restart Comfy regularly I recommend it

[void moth]

hi

[still rose]

yes flux with loras goes crazy

[viral frost]

Why is IT the only clown Flux knows?

[fleet goblet]

supir upscale

nomos to upscale it to 1 megapixel, then supir to restore it

I need to look into InstantIR

Let me sharpen the input image then feed it into instantIR

[shrewd sequoia]

带有“bit”字样的3d图标简单些

3d ICONS with the word "bit" are simpler

[heavy python]

1 hour to generate. I must be doin somethin wrong. here's my source. gotta use the hugginface access token first. hugging-cli login then use the token. first run default prompt, 2nd run... mike tyson as an ewok elder 🤞

hour generation for a RTX 3080? what am i doing wrong here

[viral frost]

You've overflowed your available VRAM and are swapping with system memory, which is very slow. Also, never use a bare .py to run SD. I assume you got this code straight from huggingface, right? This is just demonstrating the bare bones method for using SD with diffusers. It might be helpful if you're a code tinkerer, but you'd be better off using anything else -- Automatic1111, comfyui, InvokeAI, Forge, SwarmUI, etc.

[clever oar]

Myke?

[solid gate]

what ai tool is this?

[royal monolith]

ltx maybe? they just released today, and comfyUI also released support for it today

[solid gate]

ltx looks low quality

[solid gate]

the video won't play, or is it just my computer?

[royal monolith]

plays on mine. something in your discord setup. once in a while i can't play a video in the discord player, either. i have to download them

[wispy nest]

can't play it here, using brave browser

[royal monolith]

not if you're doing img2vid

[solid gate]

uhmm i go to check

[vagrant dust]

I would go with something like wav2lip on an plain video created with svd, animatediff mochi or whatever.

[clever stag]

@royal monolith Yes but that wouldn't work for any images with backgrounds that are not static:

[royal monolith]

works fine for videos

[fleet goblet]

upscaling michael stevens be like

I initially did this with supir + SDXL but that didn't give me what I wanted.

But this works.

Flux Redux + vae encode at 0.25 denoise

^ original image

upscaled with nomos8k jpg to 1 megapixel

[rotund ember]

Hi, Does anybody know why i dont have preprocessors?

Could anybody help me with that thing, i want to make exactly the same Face in different images, But i ended up with - it doesnt work

[fleet goblet]

used merge string node to add more positive tokens involving quality and noise

Bottom is original

[clever oar]

what change?

[pure surge]

Author: @eternal marlin
MotionPrompt: $minecraft_i_am_steve
Command: </random:1272485059353640963>
Background: from video
Model: v2-turbo
Explore more features at viggle ai

[clever oar]

if i put promt in flux-hybrid of cat and turrtle

in all the images the cat is simply hugging the turtle 🙂

[solid gate]

idk

[potent nest]

Im on a wizards in military dazzle camouflage kick.

[stable scroll]

Dart vader santa claud

[vagrant dust]

Here you go with your Dart Vader

And Santa Cloud

[vagrant dust]

Santa Claude

[clever oar]

Mojo

[languid pebble]

dicordos 🙂

[clever oar]

good news

i solved my problem with out of memory

[languid pebble]

YaY!

[clever oar]

in new nvidia driver i enable CUDA System Memory Fallback

it takes RAM if vram is not enough

[languid pebble]

Congrats!

[clever oar]

thx😎

[languid pebble]

So my next PC will have 128 GB Memory 😄

[clever oar]

By the way, this should ideally work by default, I used to think that this is how it works

[languid pebble]

Sounds a bit like the Apple systems ... but they ONLY have shared memory ...

[clever oar]

mac book?

[languid pebble]

M2, M3, M4 Systems, yes

[clever oar]

The interesting thing is that this does not take up a lot of RAM when the gpu memory is full

[languid pebble]

CPU and GPU share the same memory ... but they don't include enough for A.I.

[languid pebble]

I am happy as long as it works for you!

[clever oar]

driver when this function enabled- 546.01 WHQ 2023.10.31

[languid pebble]

I have the latest ....

[clever oar]

my driver was 530😃

[clever oar]

thx i too

[languid pebble]

Best wishes to you! Did you make Flux work?

[clever oar]

today I dedicated the day to experiments to solve this problem

[clever oar]

dream on this old card

[languid pebble]

At the moment you should be able to use Flux Pro in this LLM Chat: https://www.youtube.com/redirect?event=video_description&redir_token=QUFFLUhqbDJja3EzUmhLT3N0WFFaT3F2WkJsWGdWNDgwQXxBQ3Jtc0treTlhanQ1alAtZlBLZWdHLW11T2JEVTlVa3I3MjR6aVkxWlE2Y2xELWNNZ2hxLUNpS0thZDJzUnhWd1c3UlpTWDBzUzVsZklabzFOUXY1anlSSkh3aWplenFzSDVmNkNhZTZWUTZQaG1OdWJlQnVxcw&q=https%3A%2F%2Fmistral.ai%2Fnews%2Fmistral-chat&v=o3Bgl6Vjm6w

for free

[clever oar]

chat gpt analog?

[languid pebble]

It's a Beta of another LLM but it can use Flux Pro

[clever oar]

chat with ai

[languid pebble]

And create pictures with Flux Pro

Created with it

[clever oar]

I discovered that chat gpt sometimes he lies, he says what you want to hear.

but he lies very cunningly, so it’s very close to the truth

[languid pebble]

If it doesn't know the answer it hallucinates ...

Ask it something about yourself ....

[clever oar]

no sometimes he says I don't know the exact answer

haluoai better than kling with same picture and promt

more details

and live

[languid pebble]

It's pretty good but only for shorts

[clever oar]

what good for long?

[languid pebble]

I don't know a solution for that but 20 seconds would be a big step

[clever oar]

maybe after 30 years we will make videos for 5 minutes

[languid pebble]

For a music Video 20 Seconds would be enough

[clever oar]

play with ip adapter

[languid pebble]

Not made to walk on land 😄

[clever oar]

takes his first steps🙂

[fickle zenith]

un oso de 5 brazos

[royal monolith]

but he's @nimble mason not turtleshark

[clever oar]

its clownshark brother

[clever oar]

cristalwizard after party in the morning😆

[royal monolith]

ROFL!

[clever oar]

its real photo?)

[royal monolith]

nope.

[clever oar]

I thought this was a veterinary clinic

[hollow bridge]

#✍🏼｜rules-and-tos

[merry iron]

Thank you for using comcom analytics.
"comcom analytics" supports all community managers (moderators and server owners) by stats, visualization, and analytics.

If you have any questions, feel free to ask us!
Your dashboard
Help
Support server

Other languages
en: help
ja: help Japanese

[river oriole]

IMAGE OF A Orangutan

[royal monolith]

What are tou talking about?

[vagrant dust]

He spams all channels because he got some troubles with the website for stable-audio... He asked the support and did not got an answer in a few minutes, so he wrote these messagen in 3-4 channels here....

[royal monolith]

figures. and this discord doesn't have anything to do with stable audio at all.

[vagrant dust]

Yes, but maybe it still helps him to get off his steam 🙂

[clever oar]

Mojo help this man with stable audio😆

[languid pebble]

I don't use it ...

[clever oar]

becouse you know true...

[languid pebble]

Using Suno 🙂

[clever oar]

now he see answer what to do🙂

[languid pebble]

Happy now with your new driver?

[clever oar]

on 200%

but my model no have enougt quality

want buy new gpu

[languid pebble]

There are tons of models on civitai 🙂

[clever oar]

I promised myself I would do it after I got my teeth fixed.

[clever oar]

i cant use >1.5 base models

[languid pebble]

OK, that's not real fun any longer ...

[languid pebble]

Try: https://mistral.ai/

[clever oar]

i need no only online generator

i want tools like in sd

[languid pebble]

So try that one .... Flux for free

[clever oar]

now i use flux online

[clever oar]

have restrics?

credits or time

[languid pebble]

Not yet ...

[clever oar]

you know game of thrones?

[languid pebble]

Not my thing ...

[clever oar]

face id +

[languid pebble]

Good work

[clever oar]

thx

i first try it

like result , maybe its better than rope

[languid pebble]

So many tools to try 🙂

[clever oar]

i hope my processor suitable for new card

mojo

pika can do image to video?

[languid pebble]

Not 100% sure ... to many toys to know them all

Cheeky Bot 😄

[lapis quarry]

Do I need to download both separate models, or does the inpainting version still do what the normal version does but just has an inpainting thing?

[vagrant dust]

The inpainting models are trained for the inpainting task, so i would suggest to use the base one for text to image and the inpainting for the inpainting tasks.

[languid pebble]

For you 😂

[heady lodge]

Scale of 1-10 how much like a real photo does this image look

It's almost perfect to me but something seems uncanny about it

Maybe it's the skin detail

[clear pendant]

hello guys, i made a song with the brazilian poetry of Carlos Drummond de Andrade! hope you like it
https://www.youtube.com/watch?v=du9A9zH_vy0

[sterile kiln]

Flux Fill outpainting 🙂

[languid pebble]

You can see the borders of the original picture

[sterile kiln]

On this one yes, not on others

[languid pebble]

Thinking how it could be if you split the original in 2 parts, outpaing both pictures and put them together again

[shut sinew]

Got the GGUF Flux fill model working with SD-PPP in Photoshop
https://github.com/zombieyang/sd-ppp

Here is a messy workflow, for those interested

[fleet goblet]

Anyone know why LTX video instantly gets stuck on text encoding?

[shut sinew]

Do you run out of VRAM?

[fleet goblet]

Shouldn't the text encoder be running on CPU if --lowvram is enabled?

I have 16gb of VRAM, on intel arc. The issue isn't memory-related, or at least I doubt it.

The text encoder model loads fully, as well.

lemme try this

ope

I should be using the base comfycore ltx implement

[shut sinew]

Maybe the intel arc is the issue? Im not sure how well it works with this

[fleet goblet]

Actually, I looked at the github repo issues

A lot of people noticed these custom nodes aren't operating properly.

I'm going to see if the base nodes work.

[shut sinew]

[fleet goblet]

yep it works

2.52S/IT*

It's actually faster than Flux is.

That's funny.

LOL

it asks for long-form prompts, so I used an LLM alongside wd14 tagger to automatically make one based on the input image.

it worky

austraia

[sterile kiln]

I just asked him, maybe i'm just a Github rookie...
https://github.com/zombieyang/sd-ppp/issues/60

[shut sinew]

Not sure what that is about

[shut sinew]

Faster maybe, but the quality?

[fleet goblet]

\

[fleet goblet]

My prompting skills aren't the best.

he was there now he isnt

I wonder why you watermarked this.

Or if you did, at all.

You can't copyright AI-generated works or derivatives.

So good luck with that.

Then you posted a non-AI image into an AI discord

Was the base image generated with an image model

yeah no

you cant copyright this

So if I brought you to court, showed them this image, and they asked you the origins where would you go?

I doubt this 100%.

You didn't make this image.

[shut sinew]

If it's not AI why post it here

[fleet goblet]

Exactly.

[shut sinew]

Also youre describing working with AI and just trying to make it look cool

You could have said you painted it, but you obviously did not

What work? Im not saying it's not work to make this, im just saying youre trying to make it what it isnt

Go advertise somewhere else mate, this isnt the right place

[fleet goblet]

What you're describing isn't legally copyrightable

🤷‍♂️

So I can (or anyone can) steal and remake your images.

Just saying.

[celest sigil]

Also, why ruin an image with a blurred mess all over it. It's like painting a picture and scribbling all over it to try and stop anyone copying it. It doesn't work, and isn't enforceable, so completely wasted.

[wispy nest]

I am guessing this was a bizarre conversation.

Oh I am a newbie again! How quaint!

[fleet goblet]

A person trying to claim modified AI-generated works makes it copyright.

(Which it doesnt)

[wispy nest]

Put that copyright watermark right on your AI generated image!

Said all of Instagram

[vagrant dust]

Well there are all different kind of laws and verdicts in the world it is hard to keep track. Most western countries decide that a prompt and some postprocessing is not enough "creation" to be worth to be protected. On the other hand a decision in china has voted for the artist who created a ai image that was later used by another person on their "chinese instagram account". So as usual there is not one clear worldwide sets of rules....

[trim pulsar]

Anyone know how to ensure my AI creation doesn't show hands and face? even when i add prompts for it for some reason it still appears

[dark gazelle]

hello

[fierce jackal]

i asked for goku

what

is this

[fleet goblet]

what that is

is bad settings or config

[fierce jackal]

i just want to know what the problem is

[royal monolith]

it looks like you're using the wrong VAE

[fierce jackal]

you were correct

[royal monolith]

it can happen. everythign working now?

[fierce jackal]

yeah

gonna use lora now

[hearty patrol]

Hellooooo?

Anyone here?

If anyone goes to this message, someone please help me on this and how do I access the app?

[quiet current]

No idea what app you re talking about and you re showing screenshot of the Open AI entry. This is a stability diffusion related discord server... Open AI != Stability AI. Nonetheless It should work the same way I guess. Get your api key from open ai website somewhere and copy paste it in here.

[viral frost]

Would you write a description of an image based on what is not in the image? Image generators don’t understand negations. Focus on what you want to see and use photographic terms like “close shot” or “macro photograph”.

[opal flume]

Erstelle eine Kreide-Zeichnung eines kleinen Sees im Wald mit einer Windschiefen, alten Blockhütte und einem Anlegesteg an dem eine kleines Wichtel-Mädchen steht und verträumt in den Mond schaut

[wispy nest]

guys i’m new to this server where do i begin and why can’t i find a guide

[potent nest]

What are you wanting to do?

@wispy nest If you are wanting to generate images there are a ton of Stable Diffusion based image generators online. If you have decent GPU and dont mind learning a bit you can setup a self hosted framework like Automatic1111 or ComfyUI.

[wispy nest]

yes i want to generate , i am using a mac book pro m4 would that work ?

ill look for automatic 1111

thanks

[potent nest]

I dont know anything about Mac but that seems like a really powerful laptop.

[robust path]

In that case you better look for Draw Things app. It's especially created for apple devices. Models are optimized for it.

[potent nest]

Yes^

Hopefully he/she sees that before getting too deep into A1111

[viral frost]

I’m using an M3 Max 64 GB and run Flux dev with comfyui. I’ve used InvokeAI and A1111 in the past and prefer Invoke’s interface, but its Flux support is rudimentary and its development cycle is much longer than with comfy. DrawThings may be faster, but in my experience the converted models end up being too resolution-restricted to be useful to me. Flux isn’t fast in comfy for me, but I can generate a 4k+ image within about 40-50 minutes. I don’t know what amount of memory your computer has, which would be important if you wanted to use a large model like Flux or SD3.5 Large and/or wanted to go to large resolutions.

[potent nest]

Yea I wouldnt click that.

URL is obfuscated.

@dry crow Sorry to ping but you should delete that spam. Its likely 1 click and it will steal Discord UID.

[celest sigil]

It's on all channels, but mods here don't seem to exist anymore.

[dry crow]

Deleted, thx

[clever stag]

@thorny ridge under the hood it's just an array of strings, that's all wildcards are.

You can achieve wildcards in both python and in the prompt itself with the examples I gave.

easy.

[thorny ridge]

right i jusut wanted to let Chef understand it's an array of sentences not words as your explanation implied

also it's neat to learn the technical term 'wildcards' especially since this can all be used/setup without code using specifiic comfy nodes to do this without any any coding

[clever stag]

It doesn't matter a string is a string: the dog is [a matted red|a neon blue]

[thorny ridge]

i get it yeah but you understand that by setting up a poor example you could unintinationially be misleading right?

[clever stag]

What are you talking about... That's how you do it in the literal text prompt...

The dog is [red|blue] will iterate over the "array" of wildcards each prompt.

[thorny ridge]

Chef was asking for a way to generate random backgrounds, I offered wildcards as a solution bc as we agree they may be a series of sentences, by saying instead "wildcards are <a series of one words>" as your example you're misleading him to think "oh well i guess thats not what I want bc I want a series of sentences not just one word" and we both know that well but sometimes you gotta get into the head of the more common folk and realize how theyy may interpret what's natural for you and me

[clever stag]

You keep saying "one word" a string is a string. You can have a sentance with wildcards.

[thorny ridge]

your example ciited one word

[clever stag]

....

[thorny ridge]

or an array of just words

[clever stag]

Mate, it's a string...

Do you know what a string is?

[thorny ridge]

II'm not sure why you keep sayinig "you can have a sentence with wilidcards" after everyythinig i just typed i thought ive been pretty clear were both on the same page on that

[clever stag]

the dog is [a matted red|a neon blue]

[thorny ridge]

are you being dishonest or do you just not read?

"I offered wildcards as a solution bc as we agree they may be a series of sentences,"

[clever stag]

You don't know what a string is?

[clever stag]

But that's not a wild card.

That's a pre made system.

[thorny ridge]

yyeah dude im a developer too i just dont know why yyou're asking me or why this is a topic after I said "I offered wildcards as a solution bc as we agree they may be a series of sentences,"

[clever stag]

A wildcard is the actual process of making the string array yourself.

[clever stag]

Again those are not wildcards.

Those are a premade array of strings.

[thorny ridge]

the whole point i was trying to make is that by showing one words in your example of wildcards it may not be instantly obviious to everyone, same way you're askinig me, you acknowledge and understand ther'es people that dont get that concept

[clever stag]

This is a wildcard: wildcards= [a.b.c] or in the text prompt my wildcards [a|b|c]

[clever stag]

It's more obviose than just linking a node.

Anyways, good talk.

[thorny ridge]

yeah i get how they work and how yaml works and how it's struuctured anad how its essentially string templating to some degree

i get all that

you're missin the point, by citing a poor example you could be misleading someone to think the system is limited to a string without spaces

[clever stag]

Yes, it's not a wild card, it's just a collection of strings.

[thorny ridge]

and by citing strins without spacecs your misleadng people to think it cant have spaces

you get it now?

[clever stag]

https://github.com/DigitalIO/ComfyUI-stable-wildcards Go tell these guys they are using a misleading, poor examples 😄

[thorny ridge]

well i mean iif you wanna go inito github and start reading README that already demonstrates abilitiy and understanding and dedication to wanting to unuderstand

the context is this passive discord message that kind of contradicts what i had said

[clever stag]

I'm the only one here that seems to understand what a wild card is.

[thorny ridge]

so i just wanted it to be clear its not limited to just one word in case that caused confusion

[clever stag]

If you don't know what a string is that's a skill issue, I'm sorry I don't know what to say, it's pretty self explanitory. Even the literal wildcard package uses it as an example: A painting of a {boat|fish} in the {sea|lake}

[thorny ridge]

well i mean you gotta be a little more sensitive not everyone is smart like us

if they're asking how to generate random backgrounds you have to take that into reference as far as their skill level

adjust and adapt and accomodate not expect them to not what a string is

[clever stag]

I don't think a single person would look at a wild card and only think you can put one word.

[thorny ridge]

i didn't know what wildcards was a few weeks ago, i have 16 years ofo programming experiene but specific niche stuff like that has to be spoonfed sometimes

[clever stag]

Development doesn't revolve around what you know or don't know.

I explained what a wildcard was and you started arguing.

[thorny ridge]

againi the person was asking for a simple request, that means basic skill, low understanding of this technical stuff, offerinig a poor example like you did could cause confusion, iit's all good tho no big deal 👍 we're all just trying to help right?

[clever stag]

Yes and that's why I wanted to correct what a wild card was, as it didn't seem like it was understood.

But all G.

[thorny ridge]

i think ive made clear a few times already we both understand what it is on a technical level, you seem to fallback on that as a wayy to avoid taking responsibility for sharing a poor example to the potential layman but it's cool eiither way no harm no foul 👍

[clever stag]

"poor example"

[thorny ridge]

like this whole chat started bc you cited a poor example, you keep trying to avoid the subject of that quick message coud be confusing and instead focus on the technical aspect, which is totally irrelevant

[clever stag]

No mate, you didn't give an actual example, you gave a tool to generate wildcards, you didn't explain what a wildcard was or understand what it was.

[clever stag]

You should create a PR and tell them they are using a poor example to explain their wildcards: https://github.com/DigitalIO/ComfyUI-stable-wildcards

It's not my responsibility if you don't know what a string is.

[thorny ridge]

I offered the term which is like an entry ppoint and ii offered an example of a set of wildcrds, specifically that offer a mix of sentences offering exactly what they're lookinig for, when you then chiime in with examples wthout spacs that could cause confusion, especially if the person isi trying to do basic stuff like that

[clever stag]

Again, that's not a wildcard, but sure.

[thorny ridge]

sure it is, it's a zip file, its in .yaml format and it's compatible with a comfyui inode for wildcards, you could say " a white cat in [random background]" and there we go itll insert random backgrounds for ya or if you just want the backgrounds you'd just use the tag for those random backgrounds

[thorny ridge]

https://civitai.com/models/137660/loads-of-halloween-backgrounds so yyou disagree that this set of wildcards categorized as wilidcard by civit in .yaml format isn't considred a wildcard?

[clever stag]

No that is not a wildcard, thats a system to generate pre made sentences.

[thorny ridge]

wildcard is a whole system, it includes the wildcard sets created by the users, and the corresponding templating code created for comfyui or a111 or whatever tool you use

[clever stag]

No a yaml file is not a wildcard.

[clever stag]

No it's not it's a array of strings. That is what a wildcard is.

[thorny ridge]

it's a specific term for a whole system of stuff

[clever stag]

Yes that stuff being an array of strings and being able to iterate / randomize a element in that array.

[thorny ridge]

so you disagree with CivitAI giving this file a type "WILDCARDS"?

[clever stag]

Again a wildcard is not just a single set of array of strings,

Do you know what a string is, first?

Because you're struggling with my example where I used one.

[thorny ridge]

simple question bro, do you agree or disagree with that labeling?

[clever stag]

Just because it's a labeling doesn't mean it's a wildcard, that labeling is saying it's a pre made array of strings.

To tell someone that's a wildcard and that they can't make their own strings is just wrong.

[thorny ridge]

i don't think i ever said that

[thorny ridge]

so you DO disagree you think the label doesn't make it a type "wildcards"

[clever stag]

You're literally saying wildcards are only found on citiv, you didn't explain what a wildcard was or frankly even understood what one was.

[thorny ridge]

i never said that either, i think providing the term and a ready-to-use example was the most helpful route yeah

[clever stag]

So when I gave an actual answer to what a wildcard actually was, you're arguing with me and only giving examples on cit,

You're even saying mine, and the readme is a poor example of what a wildcard was 😄

[thorny ridge]

you never gave me an actual answer, youo're just deflecting the whole point of this chat lol

[clever stag]

Are you trolling now?

[thorny ridge]

that's not what i'm sayinig either, again you seem to really lean on making up things other people say

[thorny ridge]

i didn't think that was you givin me an actual answer, that's a bad answer imo

[clever stag]

You literally said "not true"

[clever stag]

If a string confuses you that's not my fault.

[thorny ridge]

yeah not true bc the layman might thing wildcards are an array of strings w/o spaces

so its important to make iit clear it's not

[clever stag]

No, no one would think that.

[clever stag]

Yes, common sense would.

[thorny ridge]

if you cant have the ability to empathize and sympathize for others with less understanding and take a step back and realize how thinigs you say might be misconstrued by others that's not my fault

that's all thisi was about at the core really

[clever stag]

https://github.com/DigitalIO/ComfyUI-stable-wildcards

Lol, then make a PR and tell them that they are wrong because you don't understand what a string is. Line 20 in the Readme: A painting of a {boat|fish} in the {sea|lake}

[thorny ridge]

no dude you're missing the point bro

[clever stag]

No mate, you clearly are.

[thorny ridge]

if you're already deep in github and you're reading through the technicial docs that already demonstrates a hiigher level of understanding and ability

if youre asking on discord you're obviously not gonna be on github bc you could otherwise find the answer yourself

[clever stag]

I gave you the liteal example of what a wildcard was and you said "not true" so I pulled the readme, that's not "deep in github" that's a simple google search for wildcards in comfy.

[thorny ridge]

again that comes down to your inability to sympathize for other's lesser understanding and skill level to capture it, my only quip was yyour example was misleading, that's all, you keep arguing at the technical level but you can't argue stupid

if someone is stupid and you show them an example using words without spaces they very well might believe its just a system for random words without spaces, its not common sense, we can agree to disagree on that

[clever stag]

Then create a PR and tell the devs of the literal package to change the example because you don't understand what a string is: https://github.com/DigitalIO/ComfyUI-stable-wildcards

[thorny ridge]

dude you can link me to that github however many times you want but you're totally missing the point

[clever stag]

If they are too stupid to test it with a space I don't know what to tell you.

[thorny ridge]

exactly there we go

now iti shows

you see how you reveal your lack of sympathy?

that was kinda my whole point im glad you understand it now

[clever stag]

Are you saying the devs of the literal wildcards are not sympathetic?

[thorny ridge]

no i'm saying the devs properly set their docuument bc iif you're in github you already passed the test for skill

its all about context man, within the context of giithub be as technical and as brief as you want, all the implications in the world are fine

[clever stag]

So you're saying I have to accomadate people with spaces because they may not understand? 😄

Mate, get a grip.

[thorny ridge]

exactly now you're getting it

[clever stag]

Lol, you're a joke.

[thorny ridge]

that's how some people are, i guess i ust have a different way of seein things

i know what confuses people

[clever stag]

It's not my responsibility for you to not know what a string is.

[thorny ridge]

ive dealth with a lot of confused people over simple things that are natural for us

anywayys i think we can consider this topic closed right?

we unuderstand each other, we've come to terms with our points of view, now im gonna get back to my programming biz

im messing with SD35L + canny to see if it's any good

this controlnet stuff feels really slow

it does have to load the full model + the 8gb control net model + text encoders so I could imagine how that would slow me down a lot but i'm consistently getting 13-15 minute times for these images

[clever oar]

I don't think it's a good idea to modify images in sd 3.5

heavy

[wispy nest]

the size of sd 3.5 and flux is kinda the new standard

next models will likely be bigger

[thorny ridge]

@clever stag @wispy nest I submit Evidence 2b for the record, while yes wildcard can also refer to lots of other things especially in programming, and even in card games, im sure we agree that no one word has exclusive usage to every domain, fun side note I think Impact makes the best wildcard nodes for that stuff

[clever stag]

Mate, you're still trying to argue that ONE meaning on a external website is the definition.

Wildcards are universally understood in programming, they exist in python and comfy USES that specific system to create them from a string array in python.

I really still don't think you know what an array or string is.

[thorny ridge]

lol bro move on like whatever man you obviously dont want to listen to reason, let's just agree to disagree, and I do know what a string is, it's that thing I use to floss my teeth with. right? 🤓

[royal monolith]

code, stop - don't feed the trolls

[clever stag]

You literally just pinged me...

Take some time to learn about development 👍

[thorny ridge]

okay sir I will, I'll make sure to take extra time to learn about development thank you and keep your teeth flossed too 😉

@royal monolith you know me well enough to tell by now right? lol

[wispy nest]

I don't think you are trolling its just typical internet semantic argument
not rly worth having

[royal monolith]

yes ;)

[royal monolith]

oh he very much is

[thin compass]

Girls, girls, girls. Come on, now. You're BOTH pretty.

[fierce jackal]

I just asked for goku and got a freaky ass image

[modest hawk]

I'm reinstalling Stable Diffusion again, I need a name for the 4th iteration. :<

Stable Diffusion: The Fourth Wall

[royal monolith]

SD4D ?

[frank latch]

My workflow. simple.
The program is almost completely different in order to change the internal algorithm.

[modest hawk]

When copying a prompt from Civit

What button do you press to push it into the WebUI? :<

[thorny ridge]

@plucky hamlet using Redux, I input an image of a frog, an image of a dust cloud and for the prompt I used:
dust cloud in the shape of a frog
i set the frog weight to 4 and the dust cloud weight to 3, I ran it twice and those are my results. I'm sure if I mess with the weights and the shift I could get a diifferent look too

[thorny ridge]

@celest sigil you wanna give it a go too? I was really trying to go for an outline of a frog without any features. Have you tried redu?

[celest sigil]

I have. Do you have a comfyUI workflow?

[thorny ridge]

sure ill post the one I used to make it easier to get going

[celest sigil]

I used the workflow in this for multiple images https://github.com/diStyApps/ComfyUI-disty-Flow

[wispy nest]

that looks fantastic

[thorny ridge]

@celest sigil there you go lmk if you need the liink foor the custom node

[thorny ridge]

lol we just had a doubter that doesn't believe in the power of reduux @plucky hamlet saidi:

You only mention that because it's new, it's just like any other ip adapter. I already tried it and it's not suitable for that task. yes I can change the look and texture of things, but I only want the shape. For example a dust cloud in the shape of a frog, not a dusty frog, which redux will make.
So I'm like let me just go into google images, type in dust storm, frog, paste it in and see what I get

[wispy nest]

happened to me in another server lol

[thorny ridge]

i think the problem is that everyone's initial impression is negative bc the node they gave us suucks w/o any way to control the weight so unless you're in the know youre just lke "yeah this sucks" and i dont blame them

this time im setting frog to 6 (low) and dust to 2 (high) and shift to 1.5/1.5

awww it crashed mid-render (damn new comfyui)

[wispy nest]

I don't think that's the only issue
I noticed its often the same people, who try a new thing then tell everyone its bad
like they have done it for multiple tools

[thorny ridge]

oh i see, you're notiicing patterns in people, and their concerted efforts to shit on good things, very insightful lol

[wispy nest]

it happened too often so I noticed the pattern lol

[thorny ridge]

still jus getting a dustyy frog

[thorny ridge]

flux or sd3?

[viral frost]

Flux

[thorny ridge]

nice atmospheric vibe its got going on, i'd imagine that's like a pier

[viral frost]

I was trying to reproduce an image from a weird dream I had a couple nights ago. Only partially successful, but I thought it had a nice look.

[celest sigil]

From Flow 😄

[thorny ridge]

no bueno

that's just frog + dust, he said he doesn't want a frog, he doesn't want a dusty frog, just the silohuette or outline of the animal with the target texture

here's my latest take using shift 4/5

[celest sigil]

I know that, I'm just showing you what Flow gave on 1st try 🤷🏻‍♂️

[thorny ridge]

so you think you can nail it with Flow? are you gonna tweak iit see if you can hit the goal?

[celest sigil]

No, it's not customisable enough.

[thorny ridge]

so with Flow you input both images and a prompt?

[celest sigil]

Yes, or no prompt if you like

[thorny ridge]

cool that UI looks slick too

[celest sigil]

It is, and has lots of flows and tutorials to try

[thorny ridge]

so that's a program not a website?

[celest sigil]

It's a node for comfyui 😄

[celest sigil]

How are you inputting 2 images into this workflow?!

[thorny ridge]

Flow is a program?

[thorny ridge]

here's the workflow that inputs 2x images into redux

[celest sigil]

Ah, I see it now sorry

[thorny ridge]

I think I got it this time, clever prompting witih the help of chatgpt

[celest sigil]

No, that's just a double exposure image

[thorny ridge]

you can call it what you want but the goal was " dust cloud in the shape of a frog"

i think thats as close as im gonna get wiithout throwing more time into it

[celest sigil]

Which it really isn't

[thorny ridge]

lol its the closest ive gotten, everything else was just dusty frogs

so latest prompt, input images

Picture a 2D abstract outline of a creature, composed entirely of a dusty texture. The creature's silhouette is clearly defined, with soft, irregular edges that blur and fade into the surrounding space, giving it a hazy and ephemeral quality. The texture inside the silhouette resembles fine particles, with varying densities that create a subtle interplay of light and shadow, hinting at depth without forming any specific details. The overall look is minimalist yet organic, with the dusty effect emphasizing an impressionistic, ethereal representation of the creature's shape rather than a concrete depiction. The focus remains entirely on the silhouette and the textured dust effect within it.

nope got a bird with dust on it

i can see why OP would say dalle3 could nail it, im sure if i gave it like 1 or 2 shots it would get the concept perfectly

[celest sigil]

1st try with ComfyUI workflow, needs some tweaking

[thorny ridge]

yeah that's the spirit of the text

idi say you got a lot closer than me, 2nd try ywith a turtle lol im done now

[celest sigil]

I just used a cloud shaped like a frog

🤣

[fleet goblet]

I know this is usually for images, but I believe it's still AI so it relates here as well

I love to mess with audio-based AI tech but SD hasn't done anything else other than SA

so ive been messing with RVC

https://tenor.com/view/dance-teamfortress-aukvary-gif-26200274

[robust path]

azazeal 🐸 ☁️

[thorny ridge]

Nice how you do it?

[robust path]

prompt: "dust cloud, shaped dust cloud, dust cloud in vague shape of a frog, it is a cloud of dust which is reminiscent of a frog, soft shaped dusty cloud"
three renders, created the gif

no redux or whatsover

whatsoever

[thorny ridge]

i think the key was getting the prompt down you managed to capture the essence well, i think the use case or benefit of redux would be like iif you had different colored smokes or different animals you could swap in and out with a generic prompt to essentially make variations easily

same prompt you used but with redux and inputting purple smoke

two more examples, left is input image, right is output using @robust path @plucky hamlet

[clever stag]

@dry crow bot made to steal users info, pretty rampant on this discord server.

[dry crow]

thx we will ban it

[plucky hamlet]

Yes those are still just a frog surrounded by dust clouds. You can get those without the help of redux. Also why do you set the weights so high? I like what GalaxyTimeMachine achieved. This is a good base image for i2i. Removing the eyes and adding more smoke and dust should be easy. @celest sigil what workflow did you use? Did you use Flux?

[celest sigil]

It was Flux. Isn't the workflow embedded?

[plucky hamlet]

The default Flux workflow or the default redux workflow?

[thorny ridge]

I think what @robust path did is closer to the original wording you used than what @celest sigil did, I did a bunch of those but I didnt consider those any good bc I felt it didnt adhere to the goal. I think if you have an idea and you want to change the texture and shape Redux is better suited than prompting bc you can get consistent images using the input image, similar to how I input various textures or smoke images and got a different colored frogs, I could replace the word frog for 'rabbit' in the prompt and get same texture, different shape.

it's all for fun so whatever works for you right @plucky hamlet ?

[plucky hamlet]

Which image by Eloyse do you mean? The animation? The animation is just a dusty looking frog surrounded by clouds. What I'm looking for is a material or a large number of small objects/particles forming a shape. That's why I liked the image by GalaxyTimeMachine because it really looks like a sandstorm at a distance that takes on the shape of a frog.

[thorny ridge]

i disagree i don't think iit looks like a "dusty lookng frog" but rather "a frog made out of clouds"

this just looks like a normal frog with clouds surround it

this looks like clouds shaped to look like a frog

your original words were: " cloud in the shape of a frog"

[plucky hamlet]

But the clouds in both images don't form a shape

[thorny ridge]

theh clouds in the second image do form a shape, i can clearly see a frog, you can't?

that doesn't look like a frog head made out of clouds? lol

thhis doesn't look like "cloud in the shape of a frog" at all

[plucky hamlet]

well I would call this a shape emerging from clouds

there's still to much random clouds around the frog that are not part of the shape

[thorny ridge]

you can call it whatever you want my guy i was just following your original words that's all

[thorny ridge]

im sure better prompting can solve that

using words like neat, clear, distinct figure might help

anyways its all just for fun, i love remixing so i thought it would be a fun challenge to try to get that cloudy frog with redux, jusut when i had given up Eloyse came around and inspired me to give it a second shot

[plucky hamlet]

Having a consistent shape is the tricky part. I already got images like that even before I went to discord. But that's not what I'm looking for. The chellenge is to get a single shape rather than a merge of to shapes as in Eloyse's image.

[thorny ridge]

I tried a pink smoke and changed the prompt slightly:

dust cloud, shaped dust cloud, dust cloud in vague shape of a frog, it is a cloud of dust which is reminiscent of a frog, neat shaped clear outlined cloud

[celest sigil]

I think you'll find he's talking about this image #🏞｜general-with-images message

[thorny ridge]

oh yeah thiis one was excellent it really delivers on the prompt, mind derp there

[spring mist]

successful fail

[thorny ridge]

dust cloud, shaped dust cloud, dust cloud in vague shape of a rabbit, it is a cloud of dust which is reminiscent of a rabiit, neat shaped clear outlined cloud
trying a rabbit and to make it neater clouds, I think I ihave to be more verbose in my prompt

@wispy nest @clever oar

[thorny ridge]

@wispy nest

[plucky hamlet]

Did you use redux?
I made this with t2i
"Houdini particle simulation, fluid simulation, cloud simulation, (clouds:1.5) shaped, sculpted like a (rabbit:1.4), cotton crafting"

[clever stag]

tl;dr on redux?

[plucky hamlet]

t2i with Flux

[thorny ridge]

lol niice i can kinda see the rabbit, lost in the clouds i'd say

[plucky hamlet]

Do you know how to change the color in redux? I have this image of a brown tornado simulation and I want it to be a black tornado, but I always get brown ones even if I increase the weight for black.

I also cropped the image in my workflow

[thorny ridge]

Doesn’t look very cloud-like just a white frog on top of a cloud cool pic tho

[royal monolith]

i know, but i thought it was fun :)

[thorny ridge]

I would set the weight to 3 and the prompt to black tornado. Do you have ReduxAdvanved or just the normal one?

[plucky hamlet]

I have the official one. flux1-redux-dev

[thorny ridge]

That’s the problem

Not the model but the node that lets you do the redux Apply node there’s an advanced one that lets you control the weight

[scenic yew]

Is this redux? If not, schnell can do it with 1step.

Prompt: Clouds shaped like rabbit, wonderful

[thorny ridge]

Yeah this is what I imagined when read the prompt, I guess it isn’t a matter of text 2 image but OP wanted a way to generate variations of that concept so that’s why I suggested redux. Part is nailing the text prompt and I think so far that’s the winner

[scenic yew]

lol this is what I got with flux ip adapter, might have to lower strength

[plucky hamlet]

I thought you meant the model. Yes I have the advanced node. But I use really low values. If I increase the value it looks more like the input image which also includes the color.

[thorny ridge]

Maybe run the image through the canny node and then feed that into the advanced node

Try 0.1 and 0.2 with Canny node so you get the shape but not the colors

[plucky hamlet]

This is nice. If I try the same prompt in dev it adds too much detail like fur and it doesn't look like clouds at all.

[scenic yew]

I actually like flux schnells style at 1step the most, after that it makes everything more plasticky like dev as well.

The only problem is that 1step is grainy sometimes.

[thorny ridge]

It’s frankly a little annoying how one man’s trash is another man’s treasure. The image attached I was like so close to deleting it bc it came out nothing like the original, I considered it a trash failed remix but I shared it anyways, now it’s like above average reactions on it having received 22 reactions since I share it yesterday which is pretty good overall. I guess it’s annoying bc I don’t see the appeal in the composition or aesthetics but whatever I guess I’m glad i made something others enjoy

[remote flower]

#🏞｜general-with-images A scenic photograph of Bolzano or a professional law office.