#๐ฅ๏ฝanime
1 messages ยท Page 287 of 1
Yeah, it uses ComfyUI portable in the background
Whatโs ur lot favourite model is currently? All these images look great
what i i2v?
image to video
illustrious for me. These models rarely fail
^
damn that really captures the style of girl on left
delicious in dungeon lora?
Just WAI-Illustrious model v11.0. and marcille in the prompt
Made a little fix. Soft inpainting works great with Illustrious btw. Edits seamlessly if you do it right
thats really god. Which model are you using, Kino no tabi?
WAI-Illustrious-NSFW. Here's params:
2girl, shalltear bloodfallen, marcille, [anime coloring, looking at another::0.2], cowboy shot, dungoen, (khyle.:0.7), arm over shoulder, pointing,
Steps: 20, Sampler: Euler a, Schedule type: Automatic, CFG scale: 5, Seed: 2406498166, Size: 1152x1024, Model hash: c364bbdae9, Model: waiNSFWIllustrious_v110
sank you, kino, no tabi @wooden sail
Like the fin tunes of it or the default model of it
modified models are good, you can pick what style you like more... for example I prefer hassaku 1.3 B for most tasks
also using naiXL_mmmmix v4.5 for more 2d look
love that art style
using leoAI for initial images again and finishing it off with illustrious... so much more "creative"
Do you use wildcards or are your poses done with prompt?
Wildcards and dynamic prompting, so a bit of both 
i love this one ngl
coffin boat ๐คญ
trying to make a solar eclipse version now, I can hardly run XL tho so I crash out ๐
Now I just wanna make coffin boats.
Sailor Vampire ๐ค Interesting combo
all the coffins keep coming out short and squatty ๐ฆ
haha well.. nothing about a coffin boat makes sense ๐
ok i'm gonna go find some food or something...
Did they ever fix pony (or really, release a new one) so the weird score tags weren't required?
illustrious
kinda
Illustrious is peak <3
Yess ong pony still has its merits but i prefer illustrious
Pony def does style better but illustrious is better overall from what I've seen
Illustrious scrubs smaller details out it seems but in exchange hands usually won't be an issue
Uhhh I think you could get around that by training a lycoris as those are generally....better, than normal loras
Learns and maintains the style waayyy better so the few I've seen didn't have the details issue
I can't make any though it needs a really good pc and time ๐
Hmm no idea how lycoris training works but i might have to look into it one day
Was illustrious from the pony people?
And didn't it go commercial or something?
I never really mastered illustratious. Too reliant on artist tags which locked in. And often the results seemed sketchy or messy. I liked the colors from Pony, though backgrounds were a struggle.
I've looked into it a little?
Nothing I could run
You need kohya I believe?
Ah should be fine
Yeahh tho lora training locally gonna be a learning experience

Mmmm I bet it'd be fun ngl
Id be doing it already if I could rs
I was thinking of renting gpu to try it out sometime
I hear it takes a LOT
I mean fair, civit ai takes hours too
Mmmm wonder how long one of my loras would take
A character pack
Those are huge
Generally 300-500 images large
Hmm wait if you sort thru the one dataset
One day I'm going to hit civits max image limit on one
I'll try locally
Typo
I think I have a dataset lying around
been generating for years at this point, and have literally never trained anything
๐
I'll slide when I'm home
I made a few now. Pretty fun ngl
No rush lol i can only do it when I'm back from work
Styles so far but I'm gonna do concepts rather soon
I've made a bunch of loras
Like often i just want specific things from a style
Around 10
But
Those are the published ones ๐
I make a lot of different versions per lora
Oo 94 downloads, 23 ๐ already
And sometimes just don't post one because it's ass
Ayyeee
Ooo bet
Ya can slide em but be sure to pin em
I forgot I store my datasets there ๐
You should have access to that channel
It's literally called storage ๐
Rip
I hate when civit doesn't save my dataset
Sometimes when I post my loras it keeps teh dataset on the page where only I can see it
And sometimes it doesnt
I just hmm download the pre tagged ones
Saw an extension that downloads an image from booru with all the booru tags applied to it for easier faster dataset making
Mmmm
Found a dataset
Dms
It's uhh
Ultimate spiderman <3
My favorite cartoon ong
Oo bet gonna see if it comes out better if it's trained on the checkpoint we like to use
Should be tagged already and everything
Oo
๐
Bett ill yeet it back when its done so you can test/upload
Hella
yeah posters are fun
dall3 finally throttled me ๐
then you have the new chatgpt model... and wow..
that text is all generated ๐
Interesting seeing what these other AI Generators come up with
evenin'
Mature girl with big breasts 
where is mature girl
mature latex girl
she s in another castle

wasn't here to see wut happened but you can do stuff like www.steamcommunity.com so might've been diff url.
2023ๅนด10ๆใใTOKYO MXใMBSใBS11ใซใฆๆพ้ไธญใฎTVใขใใกใใใฃใขใ ใผใณๅธๅฝ็ฉ่ช๏ฝๆญ้ ญๅฐใใๅงใพใใๅงซใฎ่ปข็้่ปขในใใผใชใผ๏ฝใใ
โผๅ
ฌๅผX๏ผTwitter๏ผ
https://twitter.com/tearmoon_pr
ใซใใจใฉใใใใๆญใใจใณใใฃใณใฐใใผใใQueen of the Nightใใซใฎใใฆใ12ๆญณใฎใใผใข...
idk why discord allows this.
it's really dumb and everyone shuld be aware of it

but i get what you meant by it
oic not even trying to hide it
yea as long as the link doesn't have https:// part it can be forged
yup yup, and, i should delete the image, just in case someone try to actually type that into their browser for some reason
i think it's allowed for most places that allows link preview
i think there should be restriction if it's at all formatted to look like a url
e.g. with periods & slashes it should disallow it entirely
they could still use commas or somethin but idk best way to fix it tbh. seems like a lot of ppl get hit by scams based on this
and o i mean in the one i posted, the www.steamcommunity.com is not related to the actual link it points to at all
i overlaid it with a youtube link instead
the dumb thing to me is that it's allowed to be laid over something that looks like a url
yup yup, i know about that
it's just a way to actually do something like this
there's probably no good way to stop people from misusing it tho
yeh, only noticing it more frequently 'cause of all those spam bots lately
a good way of actually checking, is to mouse over the link, and see where exactly the link is bringing you to upon clicking it.
example
yea, just gotta remember to do that every time. otherwise clicking too fast thinking you know the link and auto going past the screen that shows url again is a serious issue
someone resurrected the dire wolf, is that big ???
https://edition.cnn.com/2025/04/07/science/dire-wolf-de-extinction-cloning-colossal/index.html
or the size of a big dog?
It's how Jurassic Park starts.
dang that image has some clarity!
when cat grills?
the logical conclusion of this project is making cat girls real
Cirno
kino no tabi
HOw is this style?
i don't care what anyone thinks, there's no way this absolute shut in ever took a shower, actual stink gremlin
lore accurate
lmao that first pic
sometimes you just gotta rant
@dusky linden since when does illustrious even do text THIS good ๐ญ
Oh sometimes it just, hallucinates
It says there, blea_______ch
I will do that
thanks mania
what was the limbus code word again
limbusstyle
it doesn't work :(
show me your ui
red paint
??
I can't take this anymore
i cant help you fix it if you close your ui ๐ญ
I tell AI to do something and AI doesn't want to
its like ordering ice cream and getting lemon juice
๐ญ
I will go to bed and read my new romance manhwa
you should read it
bet
it's probable the model is bad
are these 4o?
yep. this too (but i inpainted a finger in naiv4)
I've never had success generating that many images before it tells me to go away. I get 1 or 2. ๐
oh really? I thought I saw it was still extremely throttled.
How's the prompting? Tough?
natural language
u can also when u upload images for i2i
u can be like take element from image 2 apply to img 1
etc
uhhh
i have one easily gettable prompt example
but the image has blood in it
But can it do Tifa? ๐
Actually I tried some Tifas and they all came out pretty rough.
of courz?
lol
also if it ever can't do a char
u only need 1 image
i made this with 2 images in i2i
char has no representation online
and it's nearly perfect
Interesting.
ill throw a one-shot tifa example out
using same prompt i used for nao tomori
iz almost done
Dall3 basically jsut needs "anime style" in the prompt, and the rest is gravy.
for $20 it's bascially unlimited gens
between sora & chatgpt
diff limits
sora is like 600 a day, takes a min each to gen an image
so = 10 hours
and chatgpt has sep limit from that
and resets every 3 hrs
the plus sub includes both sora & chatgpt
this pic had this prompt
sora is video right?
n
but also yes
sora is both
that's why i'm saying it's what gives u unlimited gens xD
if u go between it and chatgpt
then u almost literally can not run out of uses in a day
hmm I thought sora was their video thing
it is.
but their video isn't gud
veo 2 is best video rn
sora is just another access point to their images
as well
So if I fire up chatgpt and just ask it to create an image, what am I using? ๐
makes the best gosh darn conan i've ever seen too
4o
they're both the same
sora and chatgpt are 4o to gen
I literally never used it until this new image thing, so i'm new to it. ๐
Did you use an input image for Tifa, or it knew the character?
i've been surprised
at how well it does stuff like this too
no input images for anything i've shown so far
except
the scouter
that scouter pic i used a base pic for
and it came out really good
and here only bottle miku had 3 source images
oh yea
it can also gen transparent
I'll give it a try with a simple prompt similar to yours... Tifa Lockhart is sitting on a wooden tower, gazing out at the starry sky. Retrospective. Anime style.
I love that you spelled her name wrong and it still knew. ๐
pokerface.png
lmao
only ff i play
is ffxiv
u reminded me
i wanted to take a screencap of my eso char
and do something with it
cuz it literally 1-shot my ffxiv char
like 1:1
ogod i hope my eso outfit isn't too skimpy
xD
this was an i2i
when u i2i it keeps largely the same style
but then u can add a bunch of other details/change pose/etc
or evne just completely transform it
You tell it what to keep from the image input, right? I don't think it's a true i2i with denoise levels.
yeh it's not true i2i but
i2i and t2i are just fast ways to type
oh wow
for the first shot
that's gud
my raid markings are so anime xD
and nah it's slow period
1 min 10s or so
my raid markings look so anime
when actually turned into anime
mine's still crankin'
I was using the text 4o the other day to tell me some stories and it was insane. And I hear they have an actual creative writing focused model still coming.
they wanna reclaim the throne from grok 3
lol
well they already did with 4o beating the newest grok 3 in text & massively img now
Do you get concurrent gens with plus?
u'll burn thru ur limit faster
but yes
2 concurrent on sora
i need to specify i'm a wood elf.
probably xD
but the overall design is good
oof, this one came out pretty meh
I feel like it tends to gravitate toward the ghibli style with revisions, which is not Tifa's jam. ๐
specify a specific studio or use an i2i
I don't know any studios. ๐
I found a random image and just referened it. We'll see how it goes.
meh, didn't work... she's a squatty disaster. ๐
out of gens now.. got like 3. ๐
Better Tifa with dall3 still. ๐
just gotta work on prompt
cuz u can get like really good outputs just t2i
but gotta explain a bit more in prompt sometimes i think
but u can tell ai to take the reign on certain aspects
ah yes
her :3 smile
i'm still so much in research phase i haven't been touching up very much stuff xD
from my experience with it, it can render solid anime concepts and styles, but pushing it to try and capture some of the polish of certain illustrious models in anime usually results in it veering off into semi-realism/2.5D territory. But it can copy studio styles well enough. Its ability to generate complex images with words and interesting formats and layouts is where it really shines. It probably wont achieve the same level of fidelity you can with stable diffusion or other pure image models, but for people who want to make say, web based manga or video game concept art etc its gonna be a game changer. For now at least.
@finite herald @autumn roost scam phishing link ^
Quiet 'round these parts...
Anyone around that could answer some questions for someone new to stable diffusion?
just ask here, if people know the answer, they will reply to you
broski ran
morn'
always quiet here and makes me grumpy.
Behold. Tanuki-Bro.
this new google video gen seems legit.
Looks good. Was that a one shot? Also, only issues i saw were the frames in the winks, but you can clean that up pretty easily i would think.
genspark didnt do too bad
Not sure. This one was a re-share. I tried to generate some but my prompting isn't quite dialed in yet, and you only get like 3 before you're rate limited.
This is good. Genspark is just a front-end for other models though, right? Which was this?
Kling i believe, though it has an option to use multiple models (costs a lot more tokens/currency) to generate images and video. Might test it to see how it does, if it passes the video along to another model for post processing or clean up or what. Probably wont be able to tell unless i do the same image and prompt with every model one at a time though.
Yeah the rate limits are a real buzzkill while in the learning and testing stages. Ive got the hardware to locally house a video generation model, i just feel far too lazy to actually set it up and i need to increase my main SSD memory before i do.
Ah, are you on a paid plan with genspark?
Yeah, I hear ya. I got a little tired of maintaining local environments.
Yeah, i decided to pay for a few months to see how it does. Plus it kinda solves the backend issue of setting up local environments to test other media and models. I do my image generation locally, and intend to get into training loras and checkpoint merges eventually. But for the time being i had some spare cash so i figured fuck it.
Nice, how do you like it? I saw that the "pixai" site did well at the anime-style animations that would be suitable for some game styles. But I'm interested in all the various offerings, will have to explore genspark a bit more to see what they offer.
Also considered setting up on like a runpod type thing, but seems like that is just as much of a pain as local setup. ๐
benefit of that is you can have better availible hardware (no censored ai) while at a lower cost in short term
though its not for everbody
Well it created a functioning, albeit simple aesthetically, random prompt generator with logic that prevents redundancy in the tags, populated with danbooru tags(it searched the site to create the database), sliders to increase or decrease the number of tags, and a functioning set of theme buttons that influence the random generator to prefer tags that fit those themes,in html in a couple prompt passes. So ive established it can code. Its research and analytical capabilities (or rather, that of the other models it interacts with) are pretty top tier and quick, though i havent attempted anything close to complicated for that yet. I will say its damn convenient being able to do all of that, and generate images, build workflows, and generate videos in one place without needing a lot of backend or hardware.
oh and the mixture of agents system seems to just have multiple agents work on a prompt for comparison, it doesnt run a single prompt and output from one to the other or anything like that. Though i imagine you could create a workflow for post processing and touch up to do that if you wanted.
The autoprompt feature actively interprets and reasons through your prompt (for your clarity on what it thinks your asking and to show how its viewing it i guess) which is neat, cause then you can avoid wasted generations if its completely out of line. Though i imagine the image/video model is still gonna do what it wants
yeah there's another one people were using but can't remember...
vast I think
vastai yeah
you tried that one?
nah im a full local enthousiast
right on
Flux 1.1 Pro Ultra aint too shabby. I havent done much with Flux but i might look into it after this.
Thats so pretty ๐คฉ
Moxxi inspired? Or a steampunk character?
looks dope. better than most of the video generators ive used.
Not gonna lie, Veo2 is kinda nutty
Yeah it's neat. Wish I could do more than like 2 videos. ๐
well nevemind, today I get 0 I guess. 2 got randomly blocked for no reason and now i've exceeded quota.
Thanks google.
oof, thats pretty cringe. veo2 is also the most expensive, token-wise, on genspark for video generation. ill wait til they add it to the gemini suite to use it more lol
This is really nice mainly cuz I struggle to do multiple subjects and itโs pretty good
Thanks! I finally nailed down regional prompting and decided to test it. Got a little wonky once i got to 6 characters in one shot, but up til that point it fairly consistently delivered what i asked for. Theres a guide on Civitai (the creator listed it as a model so he could include images for each step of the guide so it took me a while to find it) that i can share if your interested.
Please do that would be helpful, rn I use Wai V13 illus model
Works pretty well
Not sure what other models that are good
I also have been using wainsfwillus v13 a lot lately. I also recommend Hassaku's models (Hassaku XL, PONY etc) for a polished anime style. Theres a lot that also handle more pastel or dreamy styles better. Havent used them much though. For 3D or semirealism,ive been using HS UltrahdCG and Uncanney Valley a lot.
Ah thank u very much I do also like the 3D unreal engine render esc style I know a friend that managed to get a blender ish style
Yeah. Uncanney Valley definitely leans more that way. Hs Ultra will render anime and CG depending on how you prompt. Im always hunting for new models to try so if you see any that perform well and have decent consistency, feel free to DM them to me.
Will do so I love checking a lot for illustrations and Flux models
Let's see some of the outputs!
quick test of hidream...
Woah this is super cool what model do you use itโs a really good gen
it's an evolution of a leonardo AI doodle, but the model used for the img2img is illustrious Hassaku 1.3B with wamiarts and Uenomigi style loras from civitai
ah ok ty
no problem. The style is thanks to illustrious and loras (here is the original messy image... leo is great for colors and composition, but not the best at smooth and detailed images)
yh thats fair i use wai v13 for 2d images
3d render style Hs UltraHD is now what i use
ooof... my prompting isn't amazing, but seems like 4o/sora does not win at pixel art. ๐
Once i got it to register 6 girls, it started bleeding character details together. So ill have to fix that in big group shots with inpaint me thinks
good quality on that last one
Thanks. The model was in a good mood for that one i think. Cause i changed none of the parameters except the prompt between that image and any of the others ive done today ๐
Love that more people are doing wide aspect ratios.
They're harder to get results sometimes, but I think the outputs can be cool.
Bonus points if it's 3840x1600 ๐
(or 12:5)
I havent tried larger than 16:9 yet, i do most of them at 1260x720 then upscale 2x.
But i got inspired to try it out cause i wanted to make computer wallpapers for my monitors XD
Im about to dive into learning photo editing to add it to my work flow (backend after upscale and inpainting) to really increase my image quality, then maybe try and go through a printing service to make actual home decoration art with it. Gotta do some market research before that step though.
@keen hollow there ya go, 3840x1600 lol took a bit longer to generate. My poor GPU was cookin
That uno hand is pure evil
bout to get sent to the shadow realm
yes, cook that gpu!
im gonna upgrade to a better one before long, or get a second gpu. Havent decided yet and not sure which route is better. But my 3080 is truckin along for now lol
a100/h100/etc if you wanna use the latest ai image models
hidream produces super gud qual images on par with like imagen 3
but needs... like... over 42gb vram
for full power
lol yeah i saw a video on tik tok and was like....im gonna prompt that
make a video from it!
I wish I knew more about the commercial video models... I'm just out of the loop.
Veo2 is basically broken for me now so that's sad.
the landscape of video and image gen models is improving and changing so rapidly, i can understand this lol
yep yep
actually not mad at how this one turned out
lil too much muscle definition, but not as bad as i thought it would be
u had access?
i'm still waiting... everyone else is gloating about the new button on their ai studio :'(
Yes I did. (Do? Seems like it was down or something today.)
how was it while u could use it? still waiting for it to be enabled in my aistudio ;_;
it was cool! my prompting is terrible, though ๐
I was too anxious to try it... should have refined some prompts better ๐
you try to adjust the prompt but the rate limits are insane
i heard it was like 125 a day
given sora only lets me make like 200 images a day that seems usable
for video anyways
pretty decent
still better than what sora can put out in terms of video i think
by a longshot
they hyped sora up so much
but rip
I do all my google stuff in google ai studio... just seems like an easy way to access everything.
You can check there if you have it!
zero loras used
flux pro ultra with "cute anime futuristic warrior queen striding down a night beach with a blue glow and galaxies behind her, robes flowing with the wind, playful sly expression on her face" at native 3k, takes 9 seconds (well, big fat cloud GPU) ๐
another version
ๆฏ็
but free money :'(((. how dare they betray me like that
I LOST MY STEAM ACCOUNT BECAUSE OF IT
Omg, you should really donโt anyhow click links on discord, or any place on the internet
I will delete my screenshot, in case someone actually typed that in for some reason
he said he got "hacked"
Well yeah, a lot of people pressed those link, their discord account got hijacked, and the hacker will used those newly acquired discord accounts to spread even more of that scam
So basically 1 to 10, 10 to 100
Dont tell me its the giftcard scam
You can recover your account thru steam support
But dull doll, i told you not to click on anything that seems too good to be true
it was a friend telling me that they gift me a giftcard for steam
I know him irl
so I thought nothing of it
they changed my email too
no idea how but I cannot recover my account
So you lost access to your current email adress
Entirely?
Mate! Get off discord and contact your hosting provider like google or microsoft
Do account recovery
oh my god doll ๐ญ
? Contact them??
just stop clicking everything
quit talking to people ๐ญ
youre gonna get robbed for everything you have ommggg
get 2factor authentication on everything too man
or just stop clicking suspicious things ๐
That too
Let this be a lesson man, if someone you know ever sends something similar
Dont click
Call them
Contact on another platform
Just not
Random links
Parental controls?
basically ๐ญ
I can be your discord guardian dull if you need one
HEY
ya doll you should probably look up some stuff on internet safety
maybe ask someone before you click a link
every time
every
single
time
how? My Email is gone. They have changed everything
because this is like twice ive seen you doing some mad unsafe shit on the internet
You can contact your providers support
like some boomer level mistakes
like i bet youd click on the 'you won the lottery' emails ๐ญ
I'm a grown adult
no shade
What provider do you use
Id be yours too but you don't need it
Sorta invasive
I got invitation E-Mails from illuminati
of course you have ๐
Click on them and be sure to pay the application fee
Dull, for the love of all that is holy
that ^
bro if he could tell what links are real he wouldnt have clicked on a free steam scam?
Does the website lead to that we site
Ill teach
I work in a school after all

you have more patience than me ๐ญ
At this rate bro is getting his identity stolen so
Yss spam
he def clicked every link available
but it was so cool to read
๐
this is why you get hacked ๐ญ
you think its 'cool' or 'fun' or 'easy'
and end up
Dull, dm me the hosting provider of where you lost your email
Or here
for a moment of 'fun'
Because im trying to help you
no, I thought the gift card was genuine a gift from my friend
Well otherwise enjoy having no email address and losing all accounts connected to it
it's https://web.de/
Sichere Kommunikation, aktuelle Nachrichten und nรผtzliche Online-Services - alles an einem Ort. Einfach, kostenlos und zuverlรคssig.
Ah german
You want the protection officer or help snd contact
https://web.de/impressum/
Send them a email with a different account. Explain what happend and why you lost all access to said email adress
Then
Follow the steps
ok I will try
Steam is gonna be easier
Reach out to support, explain what happend, what account. Provide payment details you used before. Like 'i bought this game thru paypal on this adress" to prove it
I bought a steam key
omg
this year Who-ya is back!
I wonder if there will be another big artist
Nice. Which cloud gpu? 9 seconds is pretty good.
that is one low angle... don't think ive ever seen one that low
Oh I've seen lower, those are arts gpt won't tell you
i finally got access and every single thing i tried said i was doing a violation or something... even when i told it to just animate walking through the infinite archive in this image
i guess i got hyped for nothing i haven't gotten one vid yet xD
this is even my own original char with no copyright restrictions smhhh
what even is the point of using a seed image if it doesn't accept any of them even the ones without chars from existing anime xD
Yes I think it's broken right now for many people.
Ill try just a random one and see what happens. No seed image.
ok so it actually worked... that's good
but I've heard the seed images are all broken
(if there's a prompt you want me to try, ill probably get one more video)
it's ok i wanted to see how it'd animate seed images :\
i'm sadge how it doesn't work
i did get one image but it wasn't evne anime
forgot animation words. ๐
yeah it seems to not want to do the drawings, I dunno
happy to use my next gen for a prompt if you have one
iun really have a prompt right meow. i'll probly think of stuff before i can use it next
not sure, if in doubt they have clusters of H100s... and flux pro ultra is a MASSIVE model, no consumer version available
flower theme yeah...
Been messing with Invoke AI today.
need someone to fight all the spam
Takes a little getting used to, but interesting.
oooo that's a great idea...
maybe a daily
"Spam Fighter" - the superhero
trying to get spam, it's tough
I keep getting these like, weird biscuit sandwiches.
I've done .. i dunno.. a few thousand? ๐
finally getting upscaling to ultrawide working in Invoke... almost there.
it has one little tiny aerith artifact ๐
oh man...
so cool
annoyed cat collection
which ai you used here?
soo I went to talk to the guy on web and got my email back within two hours
but they told me that I shouldn't use fake names in the future
mania used to spam a lot too
maybe you know him
saaaaad
Hey there - this is Google's new "Veo2" video generation, used via Google AI Studio.
I am definitely not a very good prompter for this yet. Some of the videos people have made are quite good.
yay burger tifa ๐
is it available for free in ai studio or does one needs to add credit card?
Free, but heavily rate limited. You get like 2 videos.
Veo2 just did this one.
wanted a bit more movement, but you get what you get
tried to do another one, but accidentally clicked one of their example prompts. so it generated that instead, and now my quota is gone. ๐
Veo2?
Ah gotcha. I don't know anything about Kling. Is that the one that's like a mobile app?
Kling AI, tools for creating imaginative images and videos, based on state-of-art generative AI methods.
web-based, Pop in image, prompt and it makes some very good stuff!
Often, I mean
I hope Maxfield doesn't find me
ah i see... expensive per video?
They kept changing their pricing model around, you used to get like 3 freebies a day
not sure now
I bought their black-friday deal. 2.99/mo for better vids than I can make on my 4090 seemed like a good deal
If you want free....
I think there's a terrific one called hailuo or something.... I'll check
I think you get 3 freebies / day here
ๆน้ ็บขๆกๅ ็ๅนฟๅบ ็ๆๆๆๅพ
other kind of cute...
would-
@toxic lotus can you generate me?
woo lots of posts today
question for upscalers for anime images what do u guys use?
i usually use 4x_fatal_Anime_500000_G
@autumn roost 
at least mask your scam link properly 
this is just next level laziness for scammer
ok ty i can use this as hi res or upscaler only? kind new to that part of settings?
the time i ever fall for a scam thats the time ik its a good scam
usually you will do your txt2img like usual, but with hires fix turned on. Then you will put your upscaler to that, hires steps to a small amount, maybe like 10, denosing strength to like 0.4 or 0.5
I never use upscalers
upscale by value to anything you want, but if you go higher, it will take longer, and thus uses more vram. I leave it at 1.5x, since i only have 8gb vram
got u cheers got to get a nvidia gpu soon just vibin on mac rn
not sure what's the norm nowadays tbh
Does sd even work on mac? 





