#πβai-images
1 messages Β· Page 44 of 1
lol dude what's wrong? why you are so into albedo last days? XD
actually, RAM, gotta say that these your pics are exceptionally good
I love them a LOT
wat
wat
wat
what exact program are you using to generate the image data?
comfyUi, he just told you :-P
ComfyUI
π΅
tried into pointilism, but can't really do that
still quite cool result
what even are you talm bout
Tell us here!
May I try to gender one of y'all characters in figurines
very alpha right now, but the idea is you can submit your art, it's automatically moderated by a vision model, and then a vote cycle is initiated and you can vote using the STAR voting system to cast your ballot
resulting in a way to present your art and enjoy other's as well!
all users can submit images for the vote, and the future plan is to use the winners for the puzzle game
yo
can i ask someone a question
does anyone know a ai thats unsensored, im trynna add a video and edit it a bit
Grok
Today, we are re-inventing the wheel
Again, like with my old post, this is made using Google AI Studio
Offline local Stable Diffusion.
The very image i sent is AI generated as well, more specifically, Z-Image
what, it uses a vision model to construct a prompt from a description?
how do i get dat
oh it literally just sends the image to ai which then returns a prompt based off that image
Nothing new
if you ask question like that probably you don't need it
BUT I still will try to explain
cumfyUI is what we use most of the time. look on it as on sort of 'environment' for stable diffusion.
You can get similar results in Forge, and it is much easier, therefore I recommend to go this way at the start.
so, just go ahead, ask gpt or grok or whatever 'how to install forge for sd' and just do it. (there is just an easy way going into forge's git and smash 1click installation, nothing fancy. if I remember correctly it will even install python env for you)
sd model - you need illustrious model, that's what we use most of the time. if you have at least 3050 with 8Gb vRAM (mandatory) or better - you good
for model just go to civit and search for whatever catch your eye
I think I'll just copy that text above and paste every time someone asks 'how to do that guys'...
you could just copy paste this or just tell them to install stability matrix cause it installs and manages multiple different UIs and everything in one app. It also tells them on install if their GPU is good enough or not.
I use it cause managing python virtual environments is a pain in the ass
lol I have read "It also tells them to uninstall their GPU"
idk why
I think I have some python envs from old comfy in my PC, and it takes off my disk space
but I'm not sure KEK
what if deleting it will crush everything I have
1TB is pain ((
well it's different if you just use python for comfyui and absolutely nothing else. I use python outside of this stuff so having automatic package management and an easier updater is nice
i click a button and it updates comfyui
instead of booting it up, running it through manager, or typing a git command
just qol stuff
yeah, seems good
cause I have to click 'update' and I know everything will blow up and I'll have to reinstall comfy lmao
it's just the easiest way
yeah i'm too stubborn where if an update breaks it, I go into the venv and fix the broken packages and/or scripts
if it's too effed I cave in and reinstall
shet
I don't have such perseverance in me π
it's probably the nerd side to me doing it, i only use AI for generating hot anime girls and i'm teaching myself programming and game development from written documentation and youtube videos :P
I could just use AI for everything but wheres the fun and creativity in that
I didn't see your success with game dev and other, but from what I've seen with hot anime girls I'm sure you nailed it
:3
well it's cause i don't share it here
or really anywhere until i'm confident in my ability
i hope by time i'm out of uni i'll have something im proud of
when I was a student I did pretty much nothing
not like I'm hardly regretting this now, but, well, it could be more productive
I'm happy that at least someone is doing better lol
lick
i'm honestly surprised my rtx 3060 hasn't blown up yet, ive been making images and nonstop using it since 2022
right around the time when stable diffusion was just releasing
I hope my shit will stand for at least as long XD
pretty much my mood now
just started my 12h shift π
I got home from my 8hr shift a few hours ago now
but i feel ya my first job I had 12 hr shifts too, I needed a change since i've been going to uni
it's killing people, these 12h shifts
You don't?
just joking
I'm on remote, so I'm pretty much free to go the f out whenever I feel like. Well, almost.
But still I have a damn lot to do every shift, so... it's like drowning in shit every time
gotcha, yeah with online jobs you usually have a status indicator kinda like discord right?
once I had to work for 15h straight, no breaks, had very much tasks to do
literally twice as much then usual
15 hours is crazy
nah, we don't have that
here we do for online jobs
i know π
i wouldnt work more than 6 hours
but hours here are more lenient
yeah, we too, sometimes. but here we don't actually need it. it's, hmmm, let's say visible if you are working or not
I'm happy that you have this option
I don't
do you usually just generate images or watch videos while doing online work, or do you 100% focus on work only (I know I couldn't)
whats ur line of work
nah, 100% is impossible. even in that 15h shift I had only, like, 85-90% at most
if shift is chilly then yeah, I'm here and there
nice, pay good?
for my place - not bad, but could be more honestly
for international measure - shitty salary
like, 1300$ a month?
that what '$' stands for, I guess π
yeah true
for me right now i'm getting around $1200 USD a month, honestly pretty good considering i'm only working 2 - 3 shifts a week cause uni
yeah, that seems good
but I really only can use like 350 dollars cause I have to pay stuff off
like me XD
I was at a point where I was struggling for money bad and didn't fill up my car's gas tank at all or put any in for a month
I was behind on payments
I only got through cause at the beginning I had a full tank
It was winter at the time, so there was no way in hell I was riding a bike in sub 10 weather
Freedom units btw not Celsius
lol, our courier fellas at -30 celcius:
"long shipment? SURE SIGN ME IN, x2 PAYING BABY"
they are like crazy
telling you
lmao
yeah that's crazy
anyways i'm off to bed, gn i got a shift tomorrow (technincally today)
damn i cannot type, i'm too tired
gn buddy
rest well
now I have to inspire myself for hot girls generating...
π
no, just unexpected, dont worry about it
true, and it probably shouldn't be a surprise to me at this point, considering thats kinda all that's put here
but hey
i havent been active here much

If only I had a button "generate hot chick"
LOL
I mean, I have this button... but my wildcard is pretty smol, so it comes out kinda boring
Hey, you have a real UI at least, be lucky for what it does have
can't blame you, i havent done shit in like a month
get what XD
π₯
that's still pretty cool
i think vision models can be very powerful
did anyone want to try this?
the next stage is the winners will be added to the puzzle game, so you can put together jigsaw puzzles of your generations
Dudeβ¦
Why does it say βWelcome allβ
And im pretty sure that's offensive to Christians
Christians know Jesus came primarily for the sinners, it is mostly a nod to the inclusiveness of DEI plus it is free to go to hell.
Well, what religion are you?
I believe the Universe may actually be a life form and everything is where it is meant to be, the only things which happen are meant to. It could be like Agnostic Buddhist. We are all part of the problem to the solution and it continues ad infinitum in a drain circling motion. If anyone gave more than a crap about anything than themselves there would be no need to persist in existence. So yeah a Nihilist.π
not offensive to me
Same but I feel like it is to other people

Don't wanna let you know.
dude you can like type "write a prompt based of the appearance of the person/subject in the image, do not include styles, poses, and background, do not use natural language"
that's literally how it works
i had AI create that specific webapp for me just to circumvent ChatGPT having a limit on sent images
cool stuff, as always
missed pics like this one
in my eyes they are like your Magnum Opus
All I have is only magnum anus
lmao
actually I'm pretty happy with my last wildcards combo
Pretty basic but gives me what I want most of the time
`masterpiece, newest, best quality, detailed eyes,
intricate details, cinematic lighting, dramatic lighting, deep shadows, sharp focus, bright colors, high contrast,
detailed textures, intricate cinematic background, best lighting, absolutely eye-catching, solo,
characters, as concepts, eye_color, colorful eyes, moods, color_chaos, composition, styles, outfit, poses, `
can't complaint
your stuff is sharp and cool
how's your little queen? kicking and screaming?
She's fussing for the moment but it's fake complaining
She smiles when you look at her now
didn't get myself much into it yet.
Well, only thing I'm doing now is exploring Grok's ability to roleplay lol
D:
Yeah grok was my go-to but it started getting sensitive on me with lewd materials so I went local
kawai, I guess
I don't understand it, I believe I'm just too young for it. mb I will be fixed only in case of having my own child
It's so interesting to watch her make sense of the world
yeah, I have some thoughts on that topic. I'd do great (I think) in terms of explaining, teaching and stuff like that. But standing out these screams, first days and nights and all of that? I'd die lol
:ramchanpout:
yoo is this the server for that one voice changer?
well, try to look at channel names
alr gng
Why did you ask for voice changer in this channel? 
yo its a question and its the first channel i've seen
evil person >:O
last one took like 10 tries to upscale. very picky one. still not very good
orig
Smart, OpenAI will be bankrupt soon
What AI Model or website did you use to generate this
Anyone here knew what happened to the z-image main website?
huh, it has website?
no idea
hmmmm
I guess I shouldn't ask grok about that...
aw, good one
lick
I won't even explain this one π
?
what
Reasons
well i want to know now
I want to make one of y'all characters in a figurine style
Like this
Alr bye
βWhos even 2bβ
Yumeko
me on my 16h shift lol
3rm
threerm?
Komichi Akebi plush.
Yo, may i try to generate you a figurine of a character you like
Cynthia from Pokemon hugging Kurosaki Koyuki from Blue Archive. 

Yo guys I've the problem that when I speak that my voice is so perfect and very clean which make others sus so how to fix it? And what does Exlusive Mode do abd F0 and pitch sommthing factor? Also I've index rate at 0.50 and format at 0 but what does Format do`?
Use #β¨βai-help for help about Vonovox.
Yes, AI Images is totally the place to ask
Where else would they ask for help
clearly you dont
Cap
Can it be any ai or just anime because idk how to do a anime
It's not letting me send any images
Or video
its just AI images in general no specific kind, and you need to get to level 5 for media persm
*perms
You don't. 
whooooo are you guys talking tooo
(ofc I know who, he is the only muted person here for me. but I don't see to what you reply lol)

Any image that's AI generated. To send an image here, try chat in chat until your username turns blue.
Do you actually watch anime
I've been wondering about this as a lurker. Good to know.
oh lol, yeah, there ya go
The mf with Batman profile is a stalker for sure. 


LMAO
anyone else using Z image turbo?
getting back into local generation and realizing this is basically like Flux except better and more efficient
i also did some research, supposedly Flux 2 is trash and terribly optimized
Well, it's not trash but it is 64gb
ZiT is where it's at right now
Someone made a lora of m'cat
95% my cat
Yes?
I do use Z-Image Turbo
The original Z-Image is something I wish to use again
However that is gone because their main site vanished
gendump 2
And then I should get back to actually drawing
awesome, what's your experience with it, overall decent?
your pixel art looks really good
Too finetuned for realistic models
it is not my pixelart
this is the ai images channel
My actual pixelarts is in #π¨βart
okay i see, who generated those ones you posted?
me
okay so let me rephrase, the pixel art you generated looks really good
it's hard to find models that can actually generate proper sized consistent pixels
even with powerful models without LoRAs the pixels become warped in size
yea there are a few slight issues but overall still pretty good
the point being that you want it pixel perfect same sized pixels throughout
Wouldn't be an issue if I resize the image
Literally just bilinear downscaling then nearest neighbor upscaling
that would be cool to see done
i don't really care for anime girls, but if you can find methods to generate pixel perfect characters that's pretty massive
and then you could even do sprite sheets
like this
(not AI generated found online)
when then lays the groundwork for high quality animated pixel characters
and THEN, you'd have something people would be really interested in
π
yes
pinnacle of basic editing
For photorealism to say the very least
SDXL, more specifically Illustrious is the go-to for anything anime
yea i like that ZiT can do this pretty easily first try
what did i post there again? i forgot
I'll try and find it when I get back to my main account
why yes, I make models.
cool art style animation combo
i look forward to seeing more stuff like this because i think it's something that without AI is very difficult to achieve
like pixel art animated in a three dimensional way, causing a weird optical illusion but satisfying product
like none?
huh?
no time
4-step bruh
and it was shit
may I ask your wf? I want to test if I'm able to run it kek
it's in the full vid
Miku plush riding a big bird. 
Finally, this image generated using "Nano Banana Pro" instead of regular Nano Banana image model on Google Gemini. 
To anyone that was hoping to use the base version of Z-Image, it has been hosted on Tensor
Yuuka moment
hi, does anyone have any experience training sdxl lora's that could help me rq?
not me π
I think it will be more productive if you ask your question as is. in that case if someone knows the good answer they will advise something
good idea lmao

Basically new to training my first SDXL lora but the graphs dont rlly look right, anyone have any idea why this could be? i think my dataset is good but would be happy to show anyone that thinks they can help
grr
well, I haven't even seen these graphs ever before XD
but probably other guys here may know
yep
an idea ton make a super pro lipsync with a jpg image of a draw? for free ahah
Free tests of Magic Hour and Heygen are not good
for it, with my own voice on a wav file
it's for a experimental movie for a festival so I need a great thing
It's not a verry long audio, but I need to make it in french (with my own voice, I'm french)and with the english version (i have it too)
so, any reason why it should be asked in ai-images channel?
I just don't understand that
seems to be more like video to me
Oh sorry! Thank you, in French, the word "images" is used in a professional context, also for video work; I made a mistake.
no worries
lol, fingas hard
Not enough
this was made with base Z-image
without any style prompts
how do i make gemini use multiple reference images i dropped into chat cause half the time it only takes one and ignores the other. plus talks about public figures all the time
The regular "Nano Banana" won't work or sometimes worse on this one. For "Nano Banana** Pro**" mode, I usually type the prompt message to Gemini chat something like:
Make the character from the image "image_1.png (e.g. Hitori Gotoh portrait.png)" to dress as another character from the image "image_2.png (e.g. Akeshiro Rumi.png)". The image background is simple white, while image aspect ratio is square (1:1).
The image file names would be taken from ones you're gonna send.
In English context, the word "image" usually refers to "photo" as much as the channel #πβai-images name implies. The #π§¬βai-chat might be a better fit for AI-related topics.

it's honestly pretty nice to have a workflow that makes pixel art actual pixel perfect without using other programs
so, you just genning 1mp and DAMN DESTROYING IT with scaling? π
legit i just send the generated image to an upscale image by 0.25 node and then another one upscale by 4x both nearest exact, and boom pixel perfect
sure
just curious what you will get :3
aww
another
@tawny peak because my outputs are larger x0.25 and x4 isn't enough XD
x0.12 and x8 looks more like it
lol
ok, I suck, not my thing I guess XD
hmmm, these are lovely tho
Oh snap I can make trump images
(don't look at hand lol)
hahah
I looked through my window
and you know what? he is not there
so it's an AI
easy
patches
Joint Strike Forces
yea i like that style
Yeah local
π¦βπ₯
Hi
hello
goodbyen't
Look up mine in #π¨βart. 
lmao
how do you consider yourself as a smart person and useful member of society?
just curious...
yall hear me? im a nnoob w discord
loud and clear
by doing nothing
today is the green branch day
this channel was basically dead for months XD
wha happened
oh, so I just had to put weight 1.3 to cel_shading. It was literally all I had to do for flat image
this whole time...
hmmm...
@tawny peak you have spoiled me! XD
now every time I do something I'm like "hmmm, what if..."
What is the tool for changing facial expression?
what exactly do you mean by that?
there are few ways you can do that
you may just inpaint it in forge
or use detailers in comfy
or ask gemini or whatever big ai for that
all depends on what and how you want to do
fastest and easiest - ask gemini, I suppose
Gemini doesn't do the trick, what is comfy and forge? Are these models?
oh boi
ok
first, please tell why gemini doesn't do the trick
could it be just skill issue? no offence, I just believe that gemini is quite damn good
If you ever feel ahh, there's "Google Gemini ** Pro**".
mb you just don't prompt it accurate enough what I try to say
even free one is a mastuhpiece btw
I asked it to remaster my pics in other styles or color schemes and it was pretty good
i enjoy it :3
answering to actual question - no, forge and comfy are not models. It's... ah, let's say an environment or UI to work with SD models
omg what
what are you doing here XD
I didn't know our queen fox is here :3
im everywhere
(the other places r dead tonight)
it's so hilarious I can't
lmao ok so no gifs i see whats the image rules in here
ah ok like mild cleavage and midrift ok tho?
so, no bubz
weeeeell hehe
it's complicated
you have to chit chat a bit, till 5th rank
lameeeee but i get it
yeah, I know >_<
you could buy this permission with inserver currency but... you earn it with chatting lmao
whatttt thats odd
as you can see pretty deep and thoughtful mechanic
lol
yeah, but... I bet on good base generation
so, combo of right model and good prompting skills and style selection
plus good taste
in other words everything I don't have 
ripppp lol
i just make a new merge lmao its pretty
I got trump videos to work!!!
Should've just left it at 'magic' and keep everyone being just awed xD
hmmm
I have feeling that 80% of time vixon is already doing that XD
because... being mysterious
why not
(just fooling)
hi i m new here, just wanted to know new website for ai like today i got to knew about whisk for video
mahΕ kitsune π
This one's for @dapper linden because he always tells me to do more cute shit and less horrific monsters - so here's both :)
cuuuute
not like I force you to do it π
I just believe you are very good at doing cute things :3

hi
Hiiii~

mastuhpiece
https://civitai.com/models/2362176?modelVersionId=2656641
-# Sorry girls but I was too lazy to have "Support" for girls although you can experiment

<@&1159293140440723499>
177 KB for a 4K image is wild
really reaching that retro style
it's actually color dithering the image using Floyd-Steinberg and a PC-98 Color Palette
if you looked into this image i believe there shouldn't be any more than 16 colors
yep, sure enough, putting it into libresprite shows only 16
Tokyo is very nice
beijing
wozai
sheepyyyy
π
Hey everyone π
Generate a wood texture Please π π π
City sky
smug
Umm
what
Nvm
If I were a moderator, I'd just. 
what rank do I need to be able to post pictures here?
When your username turns blue, that's a simple indication.
lar thank you
What is ur best AI generator to generate images?
I use any software/service that uses Stable Diffusion.
I never tried stable diffusion.
Is it easy to use?
I only tried using nano banana and midjourney
it depends on a lot of things, like what you're trying to make, if you want to run locally, your PC specs, etc
Ok thanks for the replyπ
how to send images here?
Ur name tag should be blue (i saw other personβs question )
ok
what are alternatives to voyages.weights?
what happened here lol
I wasn't here for just 2 days because of hardest food poisoning in my life...
and this chat transformed into something totally different lol
hi
Hey ππ»
hiiii~
RamChan gud now
@warm dagger Hi Stacy, how can I halp you?
wha, DM again?
you bet
does anyone post actual interesting unique shit and not just copy and paste oversaturated anime slop
lol by all means, educate us on what is "good"
π₯±π₯±π₯±π₯±π₯±π₯±

there is no "good" its all subjective, however with generative ai its pretty obvious when something looks like shit. theres no soul or style to it, nothing actually interesting to look at. it so painfully obviously ai, and it just looks like a frankenstein of other artists work, (which i guess it is) once again it could be argued that it is good looking to some people, however i find the people who find stuff like this interesting have no artistic skill outside their AI zone, and it shows
you see a lot on twitter, and its just the same old copy and paste, nothing unque, no actual effort- whats the point in posting it? i guess thats my question
by my wording it sounds like I am against AI, but ill make it clear that I have no problem with it and will even engage in it if its actually unique/ interesting looking
So, can you explain about this one? 
explain what? if its creative?
Anything. 
If you wonder how I am an artist unlike some of these people in #πβai-images , look up my artworks in #π¨βart and also explain about these as well.
lol, fun guy
I could take offense as I'm pretty vulnerable to stuff like that
but even I see how much nonsense it is lol
as someone who extensively uses AI, I still hate the generic "AI slop" look that people seem to love for some reason. the Anythingv3 / pony styles that have been around for years but just more HD now. most AI images are instantly recognizable as AI from a mile away because everyone uses the same styles >_>
but someone do love it
and, lets be real, probably it is the exact reason why it's so widespreaded - it's just more popular than other styles
simple math and social reasons
it's totally fine that people who doesn't like it exist. it's natural. but this guy above was just... not pleasant with his behavior
that's why no matter how much his statement has a point I'll confront him. Being right in something isn't a good excuse to be a dick
when I unlock image perms in 5 million years I'll try posting some unique styles I've trained ^^ eventually.. >.<
come on! It's not that bad :3
just have a quick chat with some guys in general or wherever
and you can fraction your messages in several smaller ones to fasten that process
if you want I can jump in ai-chat to chitchat with you
Guess i'll be lurking and waiting to see these "unique" styles that he trained the same way as others
hey, don't be so sceptical
what if...
and, to be fair, more than half of pictures here lately are mine
and my stuff really isn't anything of unique or interesting
so it's understandable why people may think that it's the only what people do
I like finetuning the Qwen Image Edit models on styles / characters, then you can just take any image and just say "convert to this type of style / character / whatever"... like chibi or furry or something
it actually works perfectly
the actual finetune?
that's quite expensive
you can finetune it with enough vram and ram, i have a 4090 and 128gb
its just super slow
like insanely slow
x.x
yeah
I mean it's expensive in resources. time or $ for rental
but I'm the illu guy
tried ZIT and other natural language stuff - not my thing. it's more funny to me to play with tags
i still use WD 1.4 tags for captioning and my models work with both tags and natural language
not really intentionally, they just do
yeah, I know that works
still, it's... different

good point
what can I do
I have a weakness to 2d anime girls in generic style
and boobz
sorry
Hey, i can't argue with that
yeah I think I've proven my taste already here XD
i apologise if i came off as rude, however I never meant any personal offence to you in my message. and honestly i dont keep up with this server so i didnt know how much you were posting or whatever. if you had hand drawn this stuff i would obviously not sound as harsh about it and probably think over better words to use, my bad.
feel free to dm me I am interested in seeing what you have made
it's not a question of being rude or polite. It's about being mean and narrow minded, as I see it. I can't be offended by anyone here, so it's fine, no need to apologize
I know what kind of man you are
tell me

Is it local btw?
every my picture
stablle diffusion?
yep, illustrious
sure
http s://civitai.com/user/zerg212
you can see almost everything I have here, if you want
2k is crazy
my stuff is closer to 4k actually
but yeah, detalisation could be better I guess
I meant the amount lol
how much ram does it require?
ah, amount is 3.3k I believe
brobably you don't see xxx stuff XD
I have 32GB ram
It's enough
v ram?
16gb
gpu is 5080
I started with 3050
I'd say it's bare minimum for +- comfortable using illu
you can do it with older vcard but let's be real - you don't want lol
is a 3080 good for it?
yeah, totally
fr?
just it will be not 90sec for a picture, but ehh, 150? 200? can't say
well thats aint that bad
are u using comfyui?
I think it can't do the wf I use now, but lighter wf or in forge - no problem
yes
what's dis
no I guess
its a site for opensourced ai models
I forgot the name of the ai youre using
ah you mean model
no I use my merged checkpoint
you can take mine from civit I gave earlier - it's free
or take RAMTHRUST's pink alchemy or anti-nova model
this guy is king and his checkpoints are masterpiece
@steel haven oh I forgot to praise my lord today
please take my lowest bow to you, my lord
I wonder whats a checkpoint in this context
checkpoint=model
if you want to do local gen - sure π
I mean whit local I dont need tokens, and it wont have any restrictions I think
yeah you are right
for all little technical questions like how it works or how to install all of that you can freely ask grok or gemini or deepseek - they are quite good with explanations
and fixing easy tech problems too
yeah, whit github you can have free copilot premium
dude don't mention that disgusting thing at my face lol
Hello guys
I'm new to discord and this is my first day
Would anyone like to be my friend?
Hello, this is the #πβai-images channel, about AI-generated images.
For human-generated cringe please visit the regular #β¦βchat ;)
I love big black dicks
Umm
No. 
Ok
cant handle this machineslop smh
Why don't you go cry about it on facebook, or wherever you came from?
cand handle this chatslop smh
Ah, so you're the unimaginative and boring kind of troll. Good bye.
turkey sandwich
yo bro, can your ai do more anime style arts? like 1/1 straight out of anime like
give me an idea. or even the prompt
I'll try
it can I guess
It's just little another style, not something I'm familiar with
If use proper style tags, maybe some loras - why not
Annoying orange is the goatπ₯
I heard, native
Asked Gemini to turn my character Marshie (first image) into the style of My Little Pony: Friendship is Magic and Equestria Girls
(I had to manually fix the Marshie as a pony image as it generated with an extra leg)
An MLP anime by MAPPA be like
Can you create a monogram logo for VOLTRIX ?
<@&1159293140440723499> get his ass outta here
What happened?
Ah okay, also, I got a question, can I send images of Gemini trolling here?
are they ai generated images or just pictures of gemini chat tweaking out?
Ai generated pics
sure then
I asked Gemini to generate my character sitting and it tried to keep her hand wave and fucked the hand wave and now she's doing the forbidden hand gesture β οΈ
And when I told to Gemini that it made her do the forbidden hand gesture, Gemini just went
π
β οΈ
lmao
instructions unclear - it is what it is
but it really looks like gemini decided to troll you
hilarious
π
and, tbh, first one doesn't look really bad. for me it looks more like waving then something else. maybe it is her determined expression that is ruining good vibes, not just hand.
free gemini usually is little meh for generating. PRO one is much better at it as far as I saw
dewit bettah!
nice. what did you use?
i train lora using comfy ui
and it so simple to operates if you got the time man
I want to figure out how to use comfy UI
Stability matrix installs comfyui automatically & comfyui has workflow templates built in for a lot of use cases. So the barrier for entry is pretty low, you can pick it up pretty fast.
You just need a PC with a modern AMD gpu or RTX grade nvidia card
Intel arc works too, but too many workarounds to be simple
I think question was about... you know... actual using
which is still not correct. usage is pretty easy. you have like 2 windows and manager. Oh and big button "run"
not like there is much to ruin
what's for using it skillfully... I don't know shit about that lmao
it's just a question of time and experience. you try different workflows and nodes - and learn.
It's boring to me so I don't do that (and I'm ashamed of myself but what can I do, that's just who I am)
i mean you can pretty easily learn from the pre-made workflows and expand from there
Yeah, sure. That's, I'd say, the only way to learn.
to look at few WFs, understand how it work and - if you want more - start to explore new shit
what i see a lot of people do is copy workflows from civitai that have like a million nodes and they don't have manager installed (cause they don't know what it is) and they'll say it's too hard
I copied RAM's flow with quantillion extra nodes that flow doesn't need and call myself a genius
smart? π
damn smert
when i first installed SM they had the inference thing built in but I opened comfyui and dragged in the workflow it auto created and built off that
Oh I didn't show last mastuhpiece here
not my prompt, that's the reason of it's prettiness
hmmmm
I'll go to civit and steal few prompts
idk why, I have vibes for it rn
i got WAN 2.2 (a 23 GB video model) running on my 12gb vram GPU
huh
how
I tried wan2.1 on my 16gb AND FOCKEN EXPLODED MYSELF
5 min and pretty damn mediocre animation
I'd say pretty bad
surprisingly fast though around 137 seconds per image
obviously couldn't send it here
it's not perfect, still working out some quality improvements and i'm gonna try LTXV next since it's a little newer
LTX is infinitely better and faster
wan was just a starting point for me
short skirt & snow must be pretty cold
lets think she is another superhero or whatever
dammit
I already saw like 10 images, really nice ones
and all of them have hidden prompts
all from one author
shame
sad, run it through deepdanbooru
dammit
literally any decent picture I see is her
really cool stuff
shame it's between so low effort stuff
underrated
have you tried other models than Ram's merges or have you just been sticking with those ones?
my guess would be yeah
Japan's scenery is so peak





