#🏞|general-with-images
1 messages · Page 124 of 1
someone that actually knows what theyre doing could have blended them in much better and messed arround with the lighting shadows and colors around the eyes from the glow. can also be done with inpainting but probably takes a lot longer than just manual editing with a little knowledge
Initial (left) vs Final (right)
am very very stoked for the new avatar series from the original creators. now that they've got avatar studios rolling
new story during aang's life
I thought they would be exploring a third Avatar 🤔
is that one of the na'vi?
a series of movies i guess is coming. a whole new series is coming for the earth bending avatar that comes after korra
........:D
i love james cameron but that overlap makes prompting for the cartoon so hard! URHH
ha i can imagine
probably have to be specific in prompts, or just put something like "james cameron" into negative
blue skin^^
alien, pandora
i have to do that with "IT" and "stephen King" when doing ronald mcdonald generations, and the IT still bleeds through usually
same thing with doom guy. always wants to make him into dr doom
best "non It" one ive managed. pretty janky, but the excessive number of loras probably didnt help
actually, i dont think i used any negative prompts on that one..
kinda botched the mouth on that one eh?
eh, prolly need a smidgen extra denoising on the teeth or to neg prompt the open mouth
to preserve the face your prompt isn't shifting the composition
cuz you can't be doing full denoise for obv reasons
that's an easy fix though
what's cool is it's holding the position of the eyes, the reflections and shadows are preserved, nose angle is the same, etc
imagine this, but on Pizza
wow nice
im just going to reiterate that dicordos posts deserve to be auto pinned. at the very least there should be a separate channel for these masterworks
i hope moders no here i quckly show you something
oh boy this oughtta be good
face sitting)
i think, your work is abstract enough that you dont need to be overly concerned about content violations lmao
im still absolutely baffled how your generations are so fucked up looking 😄
im not convinced you arent doing it on purpose
😁
i dont know i just only change angle((
one of the best parts with comfyui with this stuff too... this is fully automated
so these images can just roll off an assembly line unattended all day all night
i need heat in the winter anyway so it basically doesn't cost me anything... not that i'd care if it did
ha you should do some in the style of a van gogh painting
early 1900s monochrome photography
etc
bad girl
what do you mean? same thing as the "generate forever" function in a1111? get your settings where ya want and it just generates until you tell it to stop?
Only actor she costarred with that was good was Ryan gosling
yeah, and you can do shit like instead of generating the original image in the workflow, you can have a node open up a folder
and start feeding those images in to have the faces swapped one by one
shes single
you can queue up a huge list of prompts
ah, can do similar things in a1111 with stuff like batches, xyz plots, and there are several extensions for pulling prompts from text files or lists, or making random prompts, or mixing and matching from lists using different parameters. thats all baked into comfy by default, though?
always fun to use the random booru extension, img 2 img whatever picture, set to gen forever, and check back on it after a while to see what it came up with 🙂
picked up a few extra teeth, oops
yeah, more or less, there's some extras for that
the bigger thing though is the workflow
you can automate stuff like i'm doing here that you'd have to do manually in a1111 and/or with gimp etc (if you can do it at all without some of the special nodes)
so instead of having to go through and clean up images manually etc you can just map it out and let er rip
saves an incrdebile amount of time in the end
My time has come
愤怒的人
Beep boop. Error detected; response does not contained a malformed version of the requested image. Beepin fn boop!
Allow me.
Here is an image of an angry person.
imagine her hair as pineapples
And spiders crawling out of her eye sockets
Vampire earwigs boring holes out of her cheeks
And actually, she's Donald Trump.
Lion
Here is the image you requested.
Insane women from 3 body problem
Here is the image you requested.
a crying girl
Here is the image you detected.
Error detected.
Anyone else feeling melancholy
can someone explain to me how in the heal the heal tool works in photo editors? i cant wrap my mind around it, as far as im concerned its some kind of black magic fuckery.
I flipped my range rover over yesterday I'm so depressed
That's really pretty man
I hate if now man
i just dropped something you posted earlier into png info out of curiosity.... yeah.. i can see why you want to automate as much of it as possible. that workflow is longer than a cvs receipt
that sucks man
mmmm new ai go brrr
wtf 3k steps
fun fact: sd3 has 1 sampler
/imagie
its an audio generator
Idk what you think you were gonna achieve with this LOL. Unless it's bait 🤔
a future showroom of environment & desert
👀
It shows a cityscape scene with a magenta airplane, with the name "GAIMIN" written above it. The scene also includes other vehicles like cars, motor vehicles, and a frog. The sky is daytime, and there are city buildings and roads. There are no humans in the scene, only vehicles andwith the name "GAIMIN a frog outdoors.
bots dead #1047610792226340935
You're gonna start getting demands for images of dead robots
oh no
Generate an image where Hasbulla is depicted dressed as the Pope, facing the viewer while making the blessing gesture with his hand. In the background, include St. Peter's Square in Rome with St. Peter's Basilica clearly visible
#1047610792226340935 BOTS OFFLINE
table
#1047610792226340935 bots offline
He-Man The most pwerful man in the universe 😄
heman
offline bot
lol
can anybody tell me how to use this thing?
install it locally or use online services such as dreamstudio
thanks
Here is the image you requested.
Here is the image you requested.
Here is the image you requested.
Here is the image you requested.
Here is the image you requested.
I value your perspective and thank you for sharing as it led me to discovering your work: https://www.instagram.com/evan_painting0808/. It's quite inspiring - I'm even training an AI model on it now! I don't usually bother for art styles
omg poor kid 😆
hellow everyone please help me here. i am trying to train a model that can make all styles. all in one model. comment on this. which one looks more like anime ?
患者隐私保护
Here is the image you requested.
‘Cinematic Redwood’ SDXL finetune + Clarity Upscaler
Hey ! Do you guys have any tips on how i can reproduce an art like this? (windows base 11 wallpaper)
i'm sorry i'm too much of a noob to understand that, tried to google it but i'm not sure i get it. Is it something like img2img?
take it clownshark
stable cascade clipvision is actually really good at that if you want a 0shot approach and have the vram
lol
I am a lazy man and find myself gravitating to Fooocus’ img2img implementations
well i'll look into it then, thanks guys !
if you don't have cascade set up, dark's suggestion is the best imo
IPadapter is really good for this kinda stuff
beauty
bros making me hungry while fasting
maybe its because of the model, which one were you using?
Or its just the prompt
dreamshaper
food
I think the women I showed last night only ate this 😆
burger with colgate lol
Agate stone actually
That is a very expensive burguer 😁
low part not very expensive)
What would cause an image to do this?
Im learning about animatediff, so pardon the noobness. Yesterday was the rabbit hole of SVD
Isn´t the cfg scale too high? 👀
Im not too sure to be honest
Used Topaz Video A.I. to tweak the SD Video ...
Im using a real basic workflow to understand what’s going on under the hood
I took a real deep dive into SVD yesterday, but I questioned if animatediff was a better solution
Pardon the horrible screen shot. Not at the house, had to walk away from a screen
Does the checkpoint / vae need to be a specific set?
I'm not good in ComfyUI
the checkpoint in the left should be a stable diffusion model so that one is correct, but im not sure if this one is animatediff model
artstation model
cool
I was just kinda going down the instructional thing.
YouTubers make me want to punch them sometimes with a workflow tutorial…. Then you try and follow along only realizing they did a jump cut and have a complete different flow up
Or they only sell the workflow at patreon 😄
I’ll pay someone a few bucks for a workflow… this stuff takes time.
However, just sell the dang thing or help
lol
The former is annoying (jump cutting to a different flow) but the latter is justified, YouTubers have to make a living after all 😅
Right
I agree with a paywall, just be honest about it
Don’t make someone think they have a broken skill set or a problem… when it was built into a tutorial
Never start 2 A.I.s at the same time lol
That doesn't speak for the whole MJ community, 98% are probably over 18.
@south temple i found this UI for animate diff on civit btw. was going to try it out, but since i seen you experimenting lately i thought i'd mention it to you. i think a specialized interface coudl be good. https://civitai.com/models/192131?modelVersionId=216965
Added symbolic link feature so you no longer have to duplicate your models from your web UI! DiffEx is an easy to use Desktop UI for AnimateDiff CL...
@arctic laurel this is the selfie photography model from the link you posted, it looks nice
I appreciate you. Im more or less just trying to learn what the nodes actually do and mean instead of just tweaking random numbers lol. FUN! But also hell.
good luck!
This workflow uses SDXL Lightning generated images as reference for SD 1.5 AnimateDiff LCM video generations, using SparseCtrl + IPAdapter to guide...
This is a workflow that I want to try
This artist puts out some hella deep imagery.
Grumpy cat in fat 🙂
lol
SVD?
Yes 🙂
Automatic?
Forge ... but Automatic should work the same ...
I wonder if you can batch render SVD like animate
Rerender last frame and start new animation and put them together would be nice
That’s what I mean… maybe a slight overlap
But I think you will have to improve the last frame ... they are not getting better
That’s why I was looking into animate over SVD
Should be possible in Comfy ...
Comfy is supposed to be the everything system… which is why I wanted to learn the big ol beast
It’s just a rats nest
That’s cool
thx
big ol beast that makes rats nests? always beware the rodents of unusual size
Not in THIS Forrest son!
I need to crawl civit for a workflow. lol
I NEED A WORKABLE WORKFLOW that I can dissect
Man… this was not a thing. I thought I would be doing last week
no coding required!
\
Chasing that zombie stuff ?
yurp
Ugh, I need to go back to basics and read the git file stuff
So epic!
Yes but cheating 😄
I want to get to this kinda level… but then figure out an overlapping batch.
Good luck! I'm not sure whether the A.I. would continue the animation as expected. but I'd like to see what will happen
That’s kinda the best part of AI
Take that last 4 - 8 frames and run the same render
I am gonna make my GPU work lol
don't burn it ^^
you cant burn it
The gpu or the image?
lol
It handled some heavy bitcoin mining… it can handle what it was made for.
“Should”
Muhahahaha 😄
Stable Diffusion Forge using SVD
do i need automatic111
if i get this
It's kinda A1111 but with some more xtras ...
Should work in A1111, too if you aready have ...
Anyone know what might cause different results with the same identical prompt and checkpoint?
In Automatic1111
Different hardware for example ...
Different seed ....
I have not changed my hardware. seed is the same
this is quite strange
it is the realvisxlV40_v40LightningBakedvae
proteus is crazy lol
not sure whether they add random noise now ...
must be something like that
damn that sucks
is there a changelog somewhere?
Virtual Insanity
shit i can't bring that tune up without watching it. https://www.youtube.com/watch?v=4JkIs37a2JE
Jamiroquai – Virtual Insanity (Video)
Stream Jamiroquai’s Top Tracks here: https://Jamiroquai.lnk.to/TopTracks
Follow Jamiroquai:
Website: http://www.jamiroquai.co.uk/
Facebook: https://en-gb.facebook.com/Jamiroquai
Twitter: https://twitter.com/jamiroquaihq
Instagram: https://www.instagram.com/accounts/login/?next=/jamiroquaihq/
#Jamiroqu...
Insane in the mainframe ...
sawed off model bad hands in the prompt
Figured I would share a sample with the class.. this is my tunes I’m trying to put video to.
Fits to my videos 😄
lol 100%
good beat
Thanks friend. Comes out in April across everything.
No feelings hurt. I appreciate those guys.
I went after a 90’s acid tech sound with modernized bass
"legion of boom" was their peak imo
Im heavily influenced by Daft/plastikman/testpilot
drum an bass vibes in that small sample too imo
Hi guys, i´m new here, can i ask, how create a consistent model, like a influencer?
One of my friends is DJ Neelix ^^
I miss daft punk but i'm so grateful for their unreleased song that came out
🙂
really love their music video for it. so many people thought it was AI generated when it dropped
Always trying to sell hime some pictures ... but bad luck ...
(They have 2 albums unreleased)
Random Access Memories (10th Anniversary Edition) out now: https://bio.to/daftpunk
Infinity Repeating Android Tee: https://daftpunk.lnk.to/infinityrepeatingandroidtee
Watch more videos of Daft Punk: https://daftpunk.lnk.to/essentialsvideos
Listen to Daft Punk’s Essentials here: https://daftpunk.lnk.to/essentials
Join the mailing list: htt...
imo it's just a perfect homage to animation
I would upload the whole track and give you kids permission to use it with credits for all the help.
If your into that sound
i am but the technical stuff of animating against timelines and music, my headspace just not there lately
That’s what the video editing stuff is for
lol
Getting a frame rate of 30fps, and a timeline of at least 5-10 seconds of clip is the challenge
Tried video editing yesterday ... took the way to complicated software 🙂
DaVinci resolve is where you wanna live in todays world
but pretty hard for a first time user 🙂 Audio was way to long. no idea how to reduce the length of the pinal product again 🙂
the models are trained to render at 8fps. better to render the keyframes and interpolate
is there a free version of resolve?
you can use free version with small limitations ...
maybe worth learning
Resolve is free… but studio is for life and worth the money
Transformers was built in resolve
the bumblebee movie made me cry. at the end ||cena gives the air fist to bumbles an he gives it back..|| oh man. choked me right up
The workflow across the bottom is left to right and you can skip stuff you dont need in your workflow.
But bar none, there is no better color grade system on the market that beats resolve.
Dang! a few more pictures ^^
That … looks wicked!
Before I left the house for the day…. I was getting really … uhhh… broken images
Animations looked ok!
sometimes it just doesn't want to work ^^
It was bad
lol
Like the output wasn’t high enough, or a wrong model was used some place
yes ... strange ^^
Im trying to worry less about the what is being produced until I solve the how / output of the model lol
Still nightmare fuel is on point
This is my hell
Does anima require a specific checkpoint model? Or is the main driving force in the diff loader?
Or the VAE … lol
I think I need to read all the GitHub stuff
https://new.reddit.com/r/StableDiffusion/comments/1bitk3i/recently_made_this_video_for_a_job_all_still/ this video edit's song has been sutck in my head all week. neat house beat
protip: start+shift+s (win11 snip tool)
i know that feels. bout to do it right now.
You can do with stable diffusion ...
100%
Which is what brought me into this stuff lol
But it's a different part as you tried ...
Text2Video and Video2Video AI Animations in this AnimateDiff Tutorial for ComfyUI.
Install Local ComfyUI https://youtu.be/KTPLOqAMR0s
Use Cloud ComfyUI https://www.thinkdiffusion.com
Workflows https://civitai.com/articles/2379/guide-comfyui-animatediff-guideworkflows-including-prompt-scheduling-an-inner-reflections-guide
AnimateDiff Models http...
This is where I’m at… back to basics lol
Google for Deforum ....
I have been working with that about a year ago ...
Looking
they're using sparse control. neat stuff.
Bad luck they didn't want to use A.I. in their videos ... 3 months later all the DJs had it ^^
lol
linkin park released their new song with deforum driving it. that was cool
People that argue AI dont understand it
Download/Stream Lost : http://lprk.co/lost
Video Credits:
Directors: Maciej Kuciara, pplpleasr
Production Company: Shibuya - http://shibuya.xyz
AI Production: Kaiber, Jacky Lu - https://kaiber.ai/, Sagans, @sagansagansagans
Editor and Compositing Lead: Anthony Scott Burns
Illustration: jun._.ka
Animation: Alasdair Willson, Andrew Hawryluk,...
Yes ... nowadays it quite normal ... I have been one of the first working on music videos ... but took me a damn long time to learn ...
looks so dated by todays methods
Well, I can gladly help offer knowledge on the editing tip.
I did a lot of video editing for skydive stuff
Got the point why I think for videos with those lights effects might better be created with deforum?
It lost me the second I read “ (using slow motion incurs an additional 0.2 credits)”
I want to run local.
Is a1111 really better to run and learn vs comfy?
Bad luck I only have a broken early version on my new notebook ...
Easier ....
new horror)
thats actually pretty cool
thx
i saw something similar a few weeks ago, except it was a creature emerging from a puddle of blood. looked very smooth like the one you linked.. would love to know what extension or workflow they use
Old version ... hope it works here ....
Dang ... not good in choosing the right formats 😄
wow, so detailed 😍
I love the illumination
Its hard to do very dark images with the models that i use....or im just bad at prompting 🤣
There are Loras that can help 🙂
Brb - rabbit hole of YouTube videos
sonic the watermelon. he's the fastest watermelon
how am i free when you hold me so tight
i want texture not hibrid lol
speaking of worms. the new beetlejuice movie. if they use cgi worms, i'm going to sue
watermelon in shape of sonic
i want result like you
Starfleet rule 1: don't forget your good morning coffee-
the fuck is this bullshit. not my picard. he drinks tea
i was wrong. that was a one off. the prompt worked the first time but no other times
i did get this though.
hehehe
haha!
lol nice
watermelon in the shape of sonic the hedgehog
these prompts remind me of an old video game magazine. april fools edition had "chronic the hemphog" as a cover
prompting potential 🙂
💀
with promt flowwolf
i'm also using a newer model i just snagged. https://civitai.com/models/241445?modelVersionId=272411
This SD 1.0XL-Checkpoint is capable to produce nearly everything.. It is very good creating extremly realistic pictures, anime and art. But beware!...
the magazine cover i am remembering must've been crossed memories. the actual april fools sonic joke was this guy. lame pro magazine. https://twitter.com/Sonic_Hedgeblog/status/1112506661518872576
Gamepro’s April 1993 issue featured a parody ‘Lamepro’ issue within, featuring Bubonic & Snails. #repost https://t.co/5oh4tDt2ZN
can i run it on 4 vram?
lol
its sdxl
OH NICE
cheeto 🤭
wow its a merge of sdxl-dpo and another model
so it's better at prompt adherence a little bit?
I just checked it out
can't say. i just picked it up cause someone mentioned it here
my model not listen promt like your
Oh man if only someone had an SDXL finetune based on all the GamePro house art
That would be intense
Doritos
#💬|general-chat message this conversation
josh promt
game magazines were amazing
xl has a higher base resolution and a lot more parameters
sd15 is very capable still. i often find i get high quality images out of sdxl with less effort though
i like that negative prompts seem to be way less important with sdxl
I always start without negative prompt ... using XL
im going to be honest. i start without negative in 1.5 too lol. if it looks like shit i throw in an embedding and that usually solves it. dont often have to prompt out specific things in 1.5, and almost never in sdxl
all my negative embeddings are for sd15 lol
sdxl i use a few negative words. "giraffe" because people often have ultra long necks
lmao and your neg is "giraffe" 😄
there's an animatediff motion module for xl but it's beta
doesn't do people very well
Now things are making sense
I was trying to use an XL based model through animate
Which might explain the ugly rendering
im not super familiar with animatediff, but doesnt it have its own models?
or does it just work alongside whatever checkpoint youre using?
It’s ugly… but explains what I was working on in basic B mode lol
The checkpoint model I think is XL
Which might explain this ugly snot
that's the wrong vae being used
Ah!!!
XL requires the XL VAE.
im still learning comfy so maybe im reading it wrong, but im pretty sure you have a 1.5 checkpoint
itt unnoises the noise haha
variational autoencoder
It's what takes an image and encodes it into latent space or vice-versa.
checkpoint is the data set of images - VAE keeps the image in check, and ksampler is what twists up the image ?
and yeah, again not sure how much of this applies to animatediff, but using the wrong vae for model withh get shit results in image generation too. also check your models and see if theres any recommended settings for clip skip cfg etc
Well, technically a checkpoint isn't actually images...but for layspeak, it is.
It's all the resultant data from training. That is what can then be inferred from.
theres some models (for image generation at least 🤷) that if clip isnt set to either 1 or 2 they will come out really bad. so, might be worth messing with that for animate
Im trying to make this as child speak as I can lol
Just troubleshooting and understanding the basics of what the flow needs to run.
"Any sufficiently advanced technology is indistinguishable from magic." - Arthur C. Clarke
So the checkpoint / VAE needs to have the same type to render an image correctly.
Don't forget to read the description of the models ....
That’s what caused me to delete and start from scratch again. I was running before learning to walk
checkpoints are a vast latent space
Some wont a special Sampler, more or less steps, higher/lower cfg
The data is encoded in specific ways for each model type. So yeah, you've got to have the right VAE to decode it properly. 🙂
Are there “better” VAE than others? Or does it not work like that?
better vaes for some models. novelai based 15 models benefit from the anime vae that novel ai made
AFAIK some checkpoints have VAE baked inside
@hazy warren
Well swarm looks like the bees knees
Rose $hawty
ok, i have it. proof that comfy users are a bunch of masochists. how tf is someone supposed to even comprehend this?
I've seen worse.
yeah in seriousness it looks pretty well put together
Imagine if they didn't have the groupings.
simplest comfyui workflow:
loll
lol what’s that workflow for?
you install a ton more python scripts from various unvetted sources, then watch a 5 part youtube series about it
The answer is: Everything.
no coding required!
the animatediff extension is superior for these reasons bro. auto111 1.8 with ad and cn. no fucking around with wires
No code… but knowledge
There's a difficult balance to maintain with a workflow...how much do you keep in one workflow and how much do you transfer between workflows? I'm about to the point with the one I made yesterday where I want to break off the SUPIR piece again, but I'm hanging on since I put these "chooser" nodes in that give me clean stop points along the way combined with bookmark nodes to easily jump around the workflow.
I think swarm is supposed to be auto
the rumors of memory efficiency are grossly exagerated. it was a bigger problem when the stable release of auto wasn't updated to pytorch 2.1 yet
no. but you can run the automatic 1111 in api mode and script that
That’s the part that has me not installing it
Well, that feature for me is the whole point
theres a lot of better solutions to your longer animation issue. deforum extension is robust af too
there is some way it blends context batches togheter though. freeint is integrated now and it has overlap between context batches
queue up 300 frames and it'll give them the overlap you define
Ok. I’ll keep reading 🙂
deforum agaiin 🙂
put a few girls in there, then ya got the perfect image
i dont think theres many workflow sheets that go as hard as deforum does
Getting tricky my promts just look like a thesirise and a dictionary had a child
Cleann
😮
me on mondays
doing a dancing robot with lcm and the v3 motion module. should take about 20min and look fairly consistent. i'm just winging this one but i've got hope
my vram through the whole generation will look flatline like this
What happens if you go over 120 frames?
Can you in theory stay 500 frames or will it break?
So in a way… it’s like a batch
yup!
FPS 80?
You know what's rarer than a diamond
the overlap is how many frames the batches overlap. i'm experimenting with 8 frame overlap and 3 freeint passes
fk yeah. it'll render at 8fps how it was trained, then interpolate 10x
high fps gifs ftw
As long as those passes are not anything beyond the stitch
i'm not sure of the math behind the scenes. if you set it to 0 overlap, it'll look like it's just starting a new animation
Need to sleep now! Have a great time!
It’s the frame blending as my understanding
The higher the pass, the smoother the blend between A / B video
yeah its how many frames blend across context batches. freeint is supposed to improve consistency
wow its done already. interpolating the frames. lcm lora ftw
Only 4 more hours before I can go play with you guys lol
Would love to see 240 frames at 60fps
lol. i just wung this one. more stylish than i was expecting
Yup
deadass looks like a lapdance
That’s my guy
Waaahahahahahahahahaha
john an garfield
Dating a amish girl
THATS A MINOR
Irrelevant to the joke
And that's a heart attack
They look good together
Happy amish
Guy cane out well
Does ai have age if it isn't living?
You wanna explain that to a judge? 😉
I am digging this video to animatediff stuff
I can make a video edit… and send that model through
That’s wicked.
ok well i finally figured out how im going to be learning comfy. So, i dislike "open"ai as much as the next person, but damn you've gotta admit their LLM's are just on an entirely different level from anything else publicly available. only hiccup i had on this was the inital png of the workflow couldnt be read by gpt4, so i had to crop it into 3 different images. it rebuilt them in a way it could read, then followed my instructions specifically. we live in the future 😄
theres also something kind of satisfying about using censored models to help learn opensource stuff haha
Imagine a heartwarming scene painted with warm, homely colors, depicting an intimate family dinner from a viewpoint overlooking a round table. At this table sits a family of five: a pair of parents in their thirties and their three children, engaged in the simple joy of sharing a meal. The room is adorned with family photographs neatly arranged on shelves in the background, among them a single picture from Italy, adding a touch of heritage and memory to the scene. A relatively small window adjacent to the table frames the strong yet tender hand of an elderly woman as she serves a dish of pasta with tomato sauce, her gesture a symbol of generational love and care. Beyond the window, the kitchen hints at ongoing culinary endeavors with large pots simmering on the stove. The ambiance is bathed in the warm, yellow glow of the room's lighting, focusing on the family while the edges of the room fade into softer darkness, creating a cozy and inviting atmosphere that radiates warmth and familial bond.
there is a wall and serving window you, capturing the essence of generational love and care as the elderly woman serves the family from the kitchen.
random story time. friend of mine from ontario said the menonite village near him would often send the girls into the woods to find boys for , you know, bolstering their genetic diversity
ruminate on that one
wheres this menonite village, so that i can uh make sure that i dont ever go near there
dispicable right
your image is ready
heh the prompt: something, but actually nothing
neg: a thing
how do you feel to be one of the greatest A.I image creators on this server?

So beautifuuuuul 😍
Can anyone paint me a small room? As if the screen size is the glass door and you are looking in…. If that makes sense… I just want to test a theory.
Or do you know if you can make a 3d constraint….
Well…
Interesting
Better
Ouiji board on a tombstone.
heheh
I started making ai images when I broke up with my girfiend two years ago in August I feel humbled I have grown a large audience on instagram i had help with my photo realistic by fibek a old friend of mine with 50k subs I was a protece of his kinda I wrote my own anime promt and have made over 100k images to this day in two year now im hitting a real block of promts but I have a 4 to 5 nonths of art that I haven't posted im sitting on I love the community from stability and night cafe as I was a former partner with a 5% pay out code from the founder of night cafe I have come a long way but I'm not done and other people's ideas and feedback is what makes my art the art people want to see
I owe it all to my bro for showing me
Thank you
🪦
While the meme is ok, SD3 is still on track for a release.
I would say one of the most complex things to do is take a pen sketch and put it thru a ai witha. Complex promt
"release"
Crispy focus
it aight doe
remember that scene in game of thrones where they’re like “the mountain has been brought back to life!” but he’s like…kinda weird
well
That’ll be SD3
It was never dead. The community drama doesn't represent what's actually happening behind the scenes at SAI.
so?
All’s I’m sayin is, let’s let the images speak the thousands of words they surely have.
blockchain partnerships and invites being halted and ceos quitting... yeh. all just community invented drama
It's always more complex with details that are not out in the open.
Not in the open on account of the embarrassing reality
There’s no need for closed discussions in an open source culture
thats always the grift with blockchains. pretend its more complex and not really a ponzi scheme
教教我怎么使用
No. Not out in the open because it's still a business with things that simply are not happening in the public eye.
That's simply the nature of business in general. The complexity I'm referring to isn't anything to do with blockchain anything.
Fair
you'll know what i mean sooner than later
only a matter of times when crypto shenanigans start up
So...just make empty statements and say things like "you'll know soon" and we should all just take your word that everything is burning to the ground? "Run for the hills! The crypto is coming!" If you've got some sort of accusations, just state them plainly so they can be discussed.
😮
Stwelery
photogasm
I just got a very interesting reply from chat gpt wich proves its not open sorce
Actually says that there are diffrent levels of access to chat gpt
sddxl juggernaught
Was it “Yo Momma?”
My type of woman
Ask it to build me a damn workflow and output a json file
but i mean, that’s their stance
I mean.. IT should know
proteus
The taste of the cig back wards was like the smell of my range rover almost exploding after it flipped over
Quit realistic image leaf has nice detail
It’s not, they sell it, that’s how it got so much funding.
Also, I tend not to keep the images I create, I tend to just create them, post them, then delete them.
I'm a image horder I think it's a bit deeper man like hear me out have you heard of the seven keys of icann
lol
😮
Get them to have like 100 ciggies in their mouth
Make them smoke ice
This just my phone storage
EnJoY tHiS sAvOrY mEaT pOpSiClE
…..yeah me too
who knows what kind of products we'll get from cultured meat cells
indeed…maybe we’ll even figure out the carbon offset thing while we’re at it lol
we should just sequistor carbon. don't get me started on that. lock it up
soylent cicles
now make it steak
concept lora, it's made out of cake, but meat instead
sweet, finally got my SD powered music visualizer upgraded to SDXL turbo.
fingers are so cursed lol
what a time to be a weird creator of digital…well, art
700 watt lava lamp
pikamuchi :3
why they dont do yoga?)
"When loading the graph, the following node types were not found:
IPAdapterApply
Nodes that have failed to load will show as red on the graph."
hmmm any help?
Do u have comfy manager
Install missing custom nodes
Maybe you're loading a workflow with an old version
its possible
Can you find ipadapterapply when seachting for the node
They changed it after faceid came out I think
Hm check your console log when it boots
See if there's anything about it
I think I found the issue
incase anyone else runs int it
2024/03/23: Complete code rewrite!. This is a breaking update! Your previous workflows won't work and you'll need to recreate them. You've been warned! After the update, refresh your browser, delete the old IPAdapter nodes and create the new ones.
Oh shit lol good to know
hooray for that lol
any advice on how to re-wire this?
ip adapter advanced
found
Gotcha
well shoot... new error
I need to figure out the model loader now
lol
clearly there needs to be a thing
SO CLOSE!
lol
found

it better be equipped with a nuclear thruster! ... 
Reminds me of certain song by Tool
Cute 1
saw someone prompting this on midjourney. is this possible on SDXL, by any chance? been trying for half a day to no avail
Yes with regional prompter
#🏞|general-with-images flower vase primitive
#1047610792226340935 bots currently offline, you can run it locally or use online services like https://seaart.ai
i mean thats considered cheating since it's almost like reverse engineering the composition, as compared to a clean text to image output in the MJ output. if that's the case i would rather stick to controlnet since it's not hard to replicate using the MJ as a base
Midjourney is mid
Lol
Whatever gets you there. Sdxl literally can't handle it via text, but can easily do it if you're more explicit where things go. It's just using the tools we have available.
I like midjourney for some things, but with v6 it has become so insanely over stylized that I usually don't like the output.
Both MJ and SD are comparable except large cityscapes art/photography in my opinion, where MJ somehow.. 'gets it'
Looking at how sd3 is going with all the teased images, I think it will get better since the prompt adherence is pretty insane
Whatchu mean ?
Gimme the prompt
I didn't prompt the original image
This is the original image.saw on Instagram saying it's MJ with niji
Anyway, another controlnet attempt with some quick color correction on my phone
well done ! ...and what speed!!! ...
Sonic the Mario. He's the fastest Mario 🎶
Sonic the chocolateball?
yep
are there any really handsome really intelligent cool people here willing to help me with an error 🥺
You should post in #🤝|tech-support for assistance. But also, you should share the entire error and should share the configuration. That error can be caused by a lot of things, even down to exactly what it is telling you in plain words...that you don't have enough VRAM.
how though
i have 32gb
im generating a 20 step image
I doubt you have 32GB of VRAM. Maybe 32GB of RAM, but not VRAM.
ram ... yeah
ok friend in vc just corrected me
i have 20gb of vram
my bad on that one
would you hop in vc with us to try and figure out the problem 🥺
Again, you should be posting in #🤝|tech-support and you need to share the configuration.
instant-id, ip-adapter, and source
roop better
You should probably stick to Lightning or Turbo XL models...and even then, I'll only say probably ok. Generally, though, 4GB VRAM isn't going to get you anywhere on XL.
SDXL Turbo more easy than xl 1?
Turbo & Lightning have a lower computing requirement. "More easy" isn't really something I'd say about any of it.
roop is a piece of crap lol. unless you prefer that codeformer baked look. i could run codeformer on these images too, but i don't. guess why?
bee house
dont know i think is good
It's not.
dont know i get same style like original picture
myabe i not pro like you and dont see details
but for me is very ok
It's fine to like something that's not considered good. I'm a lifelong fan of the Buffalo Bills and until recently, they haven't been good for like 30 years.
Yes, it is subjective. It is, however, an opinion you asked for when you asked "why?"
his sounded not like the opinion of one person but as a fact for the majority)
Fair.
Is this the right place to ask for help generating stuff?
Nvm just found prompting-help
yep why not