#🍥|anime
1 messages · Page 117 of 1
is it even listening to your prompt at that point?
Not a specific character, I only use black hair and denim style clothes as prompts
@vestal crypt She looks fabulous.

We trust this one, we will not hug without her permission, she is a small one.
Very beautiful visuals.
Morning 
This makes me wonder if drums would be a huge mess
maybe, and also it's probably because im playing with sdxl models


What is the speed like in comparison to 1.5
but this is without refiner
or even hires fix
Ssus
Gimme this prompt, just the positive o:
sign board with text saying : "SMOLSUS"
literally


This model is much more boring than SD

Can you do a few pokemons and a cat in this style later ? :D
sure, i'll try later, gotta go right now 😦

New person there?
is this one single gen ? >.>

New girls are still remarkable.
here with metadata if you want to try it out, but mostly design sheet working
The designs are fabulous. @lyric zealot
Is this a lora?
nope, its mostly just the prompt design sheet
damn thats impressive
@nova remnant the first of many
i just hope my model knows the pokemon well enough so i can do quite a few
farfetch'd leek is gone tho
kangaskhan and starmie 😄
quite a few duds tho, perhaps i could tweak the prompt a bit more
i did quite a few yesterday on #sdxl, see #✨|sdxl message
snorlax, oddish, tentacool
Yarn designs for Pokemon?
yeah ^^ he requested it 😄
no lora's so some pokemon are simply imposibble because of token contamination
It's IRL stuff too, @wispy canopy ?
i have no idea 😮
The images look so real.
these are 100% SD
but, i did a pretty good job mixing this model, even if a say so myself 😄
i reall liked my old japanese style eevie + evolutions from yesterday tho
damn, i don't even have to name them, but totally clear which ones these are 😮
aight, one more pokemon yarn run 😄
then imma need somethign new in my veins
growlithe, raticate, mewtwo

these were very cool
and the first duck, don't know their names :D
psyduck
no

I think its another one
The one with a little vegetable on hand
farfetch'd
yeee
all is right in the world. anime waifu still generates 🥛
but such is boring.
hmm... random artist names not enough... imma need something stronger -> BRING OUT THE RANDOMIZER V2
better, but still way overpowering. lets delay their generation 10%
*huffs in massive amounts of SD waifu* this is the stuff i'm talking about
I like this landscape one
moved back the waifu's to 15% want more of the rest of the prompt to be able to do something
Hey people 
What is a good site to find vaes 
Trying to find a specific one
Clear vae purified
Huggingface
Im stupid
Its built-in the model I don't need the vae :D
one minute of love to all people who bake in vaes I love them makes it simple for me
Is that Finn from the Adventure Time series?
yes
that reminds me of multiverse
that game was fun
Who is this guy?
nice tone, not too flat and not over baked!
not using weights most of the time, more using the weight timings
i found some gltiche with my negative prompts and weights
i just got done correcting that
now using (low quality, worst quality:1.6)
that weight seems to be working nicely with the tone im aiming for
i'm... not using negatives XD
oh? are you on SDXL?
yeah 🙂
nice what mod are you using?
mod?
model
ah, my own mix
is it up on civitai?
based of protovision, but protovision only wants to make pretty human waifus, so i mixed it a bit so it coudl do other stuff as well
see my meat gundams xD
oh you merged?
yeah
cool !!
i'm not good enough training, but i do understand how to mix a bit
i was tyring out reproduction anime mod earlier, its nice but i have issue with render time on SDXL
yeah, sdxl is slow
i hope that can be fixed to work better in next version
however, the output is pretty damn good,
although this was 8 images over 2:30, and the rest was just ... not really that great
then again
i like that anime you just posted above, also i noticed SDXL has very rich output by default
my prompt is: [anime:0.15] + <my randomizer prompt v2>
ahh
lets do a couple on songmaster (my song lyrics wildcard file)^^
rich, detailed, and contextually correct a lot of times
contextually correct.. yes absolutely
civitai must be doing some server work
she REALLY. but i mean. REALLY, loves your heart. not you... fuck that. just your heart
do any of the TIs work with SDXL?
so you'd have to start all over again
i know the last one is not anime, technically, but way too cool not to share
damn, i like these dudes.
looks like some manga villain
I liked to read chinese novels 
.
yeah but thats usually when a power unlocks under stress or a weapon manafests and bam boss is like "oh crap not that.. well here's goes nothing"
casual outfit to stay alone in a messy room checks out the usual
aight, one more spin on my addiction before i go to bed xD
when SD be like, yeah imma give you wallpapers
(well, i did prompt for it actually)
the detail in them looks great
that's sdxl doing it's job
they're pretty low confidence on the model though
lots of 'artifacting' going on
i see, but nothing that hurts the eyes
he embraced the inner madness
shotgun prompt: anime gothic horror, terror, bloody, elemental magick effects, fire, earth, air, wind, ice, snow, fog, mist, rain, grass, fall, summer, winter, rocks, tree's, vibrant undertones, water, stream, windy, outline style, Boss Fight, Demon, Giant sized, very dark overtones, fine art, soft colors, black outline style, elegant, very detailed Cel-Shaded With shaded detail's, soft shadows, Vibrant diamond ink style, night time, stars and nebula, 3.5D, unreal render, stumbling over a rock in a stream, cute, adorable, cyber movie poster, holographic, fine art, soft colors, black outline style, elegant, very detailed Cel-Shaded With shaded detail's,
Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, one tail, no extra legs,
Steps: 20, Sampler: Restart, CFG scale: 7, Seed: 759166187, Size: 512x768, Model: aZovyaRPGArtistTools, VAE: mangledMergeVAE_v10.pt, Version: v1.6.0
aight, imma prompt this one after 😄
are those from styles?
mostly
only part that isn't is: anime gothic horror, terror, bloody, elemental magick effects, fire, earth, air, wind, ice, snow, fog, mist, rain, grass, , fall , summer, winter, rocks, tree's, vibrant undertones, water, stream, windy, outline style, Boss Fight, Demon, Giant sized, very dark overtones
so yeah.. shotgun
splatter and see what sticks
😄
i have this 850 style csv file that has tons of content, i cant seem to focus on them lol
this specific prompt = 160 tokens -> that's gonna be a diluted mess
i played with some of them, they have interesting outputs for the minimalistic prompts
random is only really workable if you embrace the random bs output you're mostly getting from it
styles/specifics, i feel doesn't work that well with random prompting elements
lol right.. i had my fun but they are too random for me
no songmaster file for you 😛
im more focused on female aesthetics and trying to bring their expression creatively
I would love to generate wallpapers 
this is a good prompt ^^
big doggo
the proportion is epic
why is my prompt so different in 1.6
good nigh't!
glad you like it
how different?
can i see some image comparison?
Let me try some more
Who is that person?
its slow 
try 512x512 dont' hi res fix
if you want to upscale use the Extra tab to upscale a partciular image you like to keep
I do 512x768 nothing else most of the time
It's just how it is
upscaling in extras is not a replacement for hires fix
no its not, but if cat's using adetailer that might sort out some of the blemishes with face perhaps
What is she thinking about
Sometimes I do
@bitter latch 4gb friend 
ahh ok, i use it all the time for face
The UI on 1.6 is super smooth
I'm running out of vram on 1.6
Not sure why ):
uhh webui itself should not affect vram
could be some settings with gpu use
is your 4gb card nvidia or amd?
AMD
It works on 1.5 no problem
oooh ok, afaik amd kinda have issues with sd
not as optimized as nhvidia cards
Yeah

nvidia
I wanted to try the samplers in 1.6
3050 Rtx
Do x/y/z plots they are fun
How many it/s do you get on it?
Mines at 2.2-2.3s/it 
512*768 images
It just eventually runs out of vram
Hmm
I don't have anything open in my PC
Is there a way to import every settings/extension you have from another folder
check if you are already using in webu-user.bat file COMMANDLINE_ARGS=--opt-sub-quad-attention --lowvram --disable-nan-check
if not, try that line and see if it helps any
gotta i'm like this restart sampler
I'm using --medvram --opt-sub-quad-attention --disable-nan-check --skip-torch-cuda-test
you need to use lowvram
not sure I need skip torch D:
try that line i gave
remove the previous
I've been using medvram without problems 
its faster
on 1.5
for AMD gpu medvram would apply to 8gb i think
restart has some contextual issues?
i dunno if contextual is the right way to describe it but i notice some differences in following the prompt with restart also bit slower than dpm++ sde karras
mean to say I gotta love the restart sampler.. my brains scattered. blah
its good with SDXL that ive tried 🙂
--lowvram is incredibly slow and really isn't worth using unless you can't gen at all without it.
Do some samplers cost more vram than others ?
not by any meaningful amount. some will seem slower because they are effectively doing 2 steps for every 1.
devasting on performace is true ....i was using medvram with my old 4gb nvidia, but i dunno cat's having issues with memory completely

Do previews affect performance in a meaningful amount
If you pick Appox cheap with a slow refresh time, not really. On less than 6GB vram, that's what I would use. I'd try 5 for the "Live preview display period", if that still seems to be slowing gens down, then I'd do 10.
With 8GB or more, I would recommend Approx NN or TAESD instead.
Progressbar would be worth turning off on low vram cards too most likely.
I've been using 9 approx NN
Interesting
Why is that
I know in the past, the progressbar hurt performance almost as much as Full live preview. Not sure if that's gotten much better.
TAESD is roughly equivalent to Approx NN. Not sure on the technical differences between them.
She's going to turn against Hogwarts
TAESD gives much better preview, maybe a little more intensive than cheap tho
imo it's worth it for my 8gb vram
I shouldn't have them on
Half the time I'm looking away from the monitor while in bed waiting for pics x3
But it's fun to see the image come to life 
the AI is watching you 

When previews slow your gens, does it appear in it/s or time to gen?
i know right, that's why i put my live preview display period to 1, and progressbar and preview update period to 200ms

Really o:
https://i.imgur.com/6e0h9ie.gif like this
Ddim didn't do a tail, bad sampler :<
i wonder if .....
medium shot / cowboy shot / medium full shot / full shot
only work specific to a model or not
cant seem to get what im aiming for ... here is an example ...
(medium full shot:1.1) of a woman, indoors, (highly detailed, beautiful, aesthetic:1.1)
feet out of frame will work better for anime models due to that being the name for it
hmm you mean feet out of frame would cover the whole body?
Is this restart
probably not so much for anime model
ouch, so how do i control the camera angles for anime?
it's definitely much slower using restart compared to some other samplers
Welp this looks like a Modern Potter-world timeline that got screwed.
for anime models it's close-up, portrait, upper body, cowboy shot, feet out of frame, and finally fully body
How is she so fabulous, I wonder?
depends on what you really want
ok .. i think i will put those to test what Kether mentioned 🙂
feet out of frame,
in the case of having similar effect as medium full shot, you can do what @runic vault has said, and maybe implement elements to include the knees but not the feet
like example, kneeshigh
yup yup, that's probably the super cool effect of restart
fixes weird stuff, like the mangled fingers
Trying to get a more adult looking character out of cetusmix rn although I might switch up models and loras.
Kinda does
Maybe detailed skin in prompt?
Not sure how it is with full anime models
what's your prompt? 
not sure if you included it, but can try adding prompt aged up, mature female
and even giving more weight to those prompts
(wrinkles,grandma,old lady:2.6)

He probably still wants a girl
Not a fossil
im getting a lot of hit and miss, here is another example that didn't quite come out right...
a woman in standing position, indoors, (close-up), (highly detailed, beautiful, aesthetic:1.1)
controlling the cam angle has always been tricky for me
one another way is to use controlnet
yes but trying with prompting
i mean prompt should give correct outputs on those basic postures
Quite amazing stuff there.
Close up instead of close-up?
Sure I'll give it a go.
i'll try that .. i could also add close up view?
(obama:2.8)
Thanks its been almost a month since I took some serious tries. Changing cetusmix from its usual style (which looks like blue archive) is hard.
Close to the camera, near view.. even zoomed in has worked for me in the past xd
@toxic dragon Working on designing images for women in their 20s onward?
late 20s early 30s
i have a feeling the canvas size can interfer with that
You're going for 1:1?
768x768
Maybe wide angle or something like that
It worked for me on non portrait resolutions
look at this.. lol not so obvious but this works
a woman in standing position, indoors, (closeup portrait), (highly detailed, beautiful, aesthetic:1.1)
Detailed skin doesn't seem to change much.
zooming in, super closed-up, portrait, zoom the fuck in,
I have had luck i the past with meinamix v8 and older looking characters but its ability with complex outfits is sometimes less than meina 3.5

A Beautiful woman there.
highly detailed skin with weight might, but could also over bake
Once I did something like that
I got eyes being 60% of the pic
👀
Zooming in with these tags can get a little over details and switch to semi realistic.
So its currently now a game of managing distance to get it to do what I want.
These are a nice place
I mean
I'd say the first one (china dress) seems the best style wise.

So very beautiful.
She is strong.
Oh well I guess for now I'll mess around a bit with detailed skin and distance? Cetusmix 3.5 may be an older model but I always liked its style.
Meinamix v8 does a good job too when it works.
Yeah mature female, aged up, older, tall, tall female
Negatives, have things like: childlike, child, young, and younger
I suppose I'm getting decent results since adding stuff in like lips also can help a bit.
try with age specification, like a woman, age 30,
Hello, can someone tell me if what I writed in #1072238304042438758 should be posted here?
if it's anything related to anime then sure
and, it's pretty forgiving here, that almost everything can be linked to anime
felt cute might submit later 
made this one too
Well my Minecraft skin looks like this and my prompt looks like this, I don't know what can I add more to have the best result
And here's the results. It looks great but idk if i can do better
so good
you need to either go to img2img and set a very high denoising strenght or use controlnet like controlnet shuffle
@toxic dragon Any luck at your side?
Img2img a Minecraft character sounds interesting
Im not a pro at generating things but I'll try thx.
Late teens to early 20s?
Hmmmm, why not ? Let me try that.
that will not have any impact on the output for u
Anyone here also a part-time roleplayer?
Idk if "double ponytail with "green elastic piece" and "bottom hair is weavy" are good prompt using only txt2img, im not a native English speaker so I don't really know how to describe my minecraft character's hair
very long hair,twintails,blonde,green hair ornament,
mhhh, looks better! thx.
It's too pixelated for SD to recognize it as a person I think
Sketch something on paint
Much easier xd
I do that sometimes
More of the same I'll try more tomorrow when I'm more well rested.
I use AI because my level at drawing is negative lol.
I am curious to see if you'd be up to do women in bikinis sometimes.
Wait, negative art?
I took a stab at making her. The hair ties didn't come out green, other than that, what do you think?
@worn nimbus my negative art into possibly worse ones generated by AI, mine is perfect!!

Omg, she looks so cute i really like her dressing! I think that if I get this result with more silver-blond dye, dark brown eyes and more older it would be perfect!
when i say im 0 im literally drawing house like a 4yo child lol
Looks like someone from Guilty Gear, I-No what you did, there.
the hair turned green 
Green Miku Hatsune. 😮
Best character I'll smash you with a guitar :<
ahaha she looks like an elf with modern clothes
slightly closer
I believe in you
Anyone can draw good enough for the AI to know the details!
Blond Miku Hatsune?
nope, just trying to get this #🍥|anime message
My pfp is a drawing I really like it :d
Shes creeper color theme
Differents style and I really like the way it tooks
I mean in the sense she looks like a blond Miku Hatsune.
not wrong if...
oh my, the head is lookin' perfect
is there a way to save the model as lora or somethin like that so i can do whatever I want with my base character ?
because even the outfit is perfect (remaining is only the socks but its fine)
I'm confused 
not the best but 
I sometime use Lycoris/Lora to generate things from manga's character, so Idk if I can actually do the same for what you are creating but maybe I'm mixing things up
well, i think you meant you want to train a lora based on what i generated?
or, if you want to use the same prompt to generate the exact same pic, i can give you the metadata embedded pic, and you just need to follow the same prompt using the same model + lora
SDXL?
nope, sd 1.5 based
You using a Minecraft lora?
nope
.>
i guess the model im using was merged with model that was trained with minecraft dataset

as I said, I don't know much outside the prompt, so I confess I don't know how it works in detail. But my objective is for example to have lora:Nykan and then input different prompts for the background but I may be complicating things
yup, for that to work, you need to be able to train a lora, with dataset that's close to what you want
is it difficult to achieve ?
well, i can't talk much about training lora/model, because i don't know shit about that
you will have to ask people like @supple raptor
Didn't require much, used 20 pics of a character xd
i might finally start training stuff after upgrading my gpu
O:


Mhhh okay I see no problem, if at least you can lend me information about how you actually have these results ? 😄
well, example embedded inside this image is metadata with information about the parameters of the generation of this image
so you just need to copy everything and you can get the same image
RTX minecraft
You mean the name of the generated picture ? Im sorry i'm not good xD
Wata looks good
1girl_solo_blond_hair_1.2_long_hair_wavy_hair_twintails_dark_brown_eyes_green_jacket_white_shirt_green_skirt_green (this)
nope, for example, you can download this image, and drop it in this website https://supagruen.github.io/StableDiffusion-CheatSheet/
and you will be able to see the generation parameters for this image
well, my pnginfo tab doesnt work
D:
but the website is also an easy way, that you don't have to open auto1111
Ahhhhhh okayyy


but got bored of it, shhhhh don't tell miku
Kek

1.6 Flashbangs me when I open the SD
dark mode 

What's aoi and tafisIsAllYouNeed in your lora's prompt?
that's two different lora
that you can download at civitai
how is your UI so different than mine?
it slightly alters the face generation, not too much since it's at weight 0.2
ohh
Oh I see and same for bad-image-v2-39000, negative_hand-neg on the negative prompt ?
yup, those are texture embedding
What's this mess of samplers 
you can take it as the same as lora
You spilled them out of the drop menu
back in my days, there isn't the radio group selection method for samplers, as such, i was used to it. 
I want more negative prompt because even with my actual negative prompt my character has a 3rd leg with no reason XD
well, that depends on your model. And more negative prompt doesnt mean it's good
sometimes it changes the style too much
I don't like when menus get shifted down the page
true actually
but im already used to it


I want to pick them up one by one and leave them wherever I want
negatives prompts can really have big interferences against checkpoint?? i didn't know
Like magic :D
i guess so, if not, how would it affect your generation?
for me, negative prompts are only used to remove things, not to change them completely
If you change something, the AI will think differently about how to make the image
If you do it enough, you can even see the style change
well, if you remove something, example color blue, it will change the "style" by removing blue. So in theory that will change them
These are the same prompt but different negatives
ah I didn't know the difference, I really thought that the negative prompt came after the generation and therefore removed things, but yes in retrospect it's not very logical and not optimized to do that
That little bubble really wants it
it put the ahegao face in the bubble 🤣
anyway gn, I'll try tomorrow to install every lora to try to have the same result than you, thx for your help!
She looks like Rias Grimory from HighSchool DxD for some reason.
I just went to google to see what she looked like again
and my search had "rias gremory png" 
What did I search that for lmao
Wonder what it'd be like to see her in bikinis, if I may ask.
I was aiming to hope for bikinis, but that one piece suit ain't too shabby either.
Miniature Rias Gremory, anyone?
Definitely :P
If her knockers get any bigger than they already are, she'd look even more like Rias.
@runic vault So far so good?
I think they are big enough >.>
Never said she was flat, though.
She looks even more like Rias Gremory than ever.
I meant in the second one, black bikini
Thanks for clarity there. :3
Does anyone feel up to do bikini pics for someone like Goldcorn, sometime?
Is that a real character, do you have a lora ? >.>
Yeah. (as in she's a real character, no on the Lora stuff.)
She's a player character of mine, actually.
@nova remnant Should I show you her profile sometime?

@nova remnant Here's the profile in case that helps, unless it's better to share by direct messages.
Character
Copy this sheet to your own google drive before editing. I won't share with edit permissions because of troll corruption.
NAME:,Goldcorn Lettré Heinie Avisé Jokinen
,LEVEL:,3
GOAL:,Uncover why Celestial Words are becoming bound to mortal races.,INFLUENCE:,4
DESCRIPTION:,Committed
With ...
bro pulled out the spreadsheet 💀
@nova remnant Okay, I have more pic references for Goldcorn sometime later when you're ready.
And yes, she has genetically natural H-Cup knockers, really
Is it bad?
When Cat with 4gb returns, Cat can ask me and I can get more image references in case it helps.
me and my man necoart
many people make the best
except steven


Need me to get you other image references for Goldcorn herself?
I'll gladly do that if you want.
@dawn jasper lets go in DMS :D
in life nothing is free
Let me guess you're an artist who gave up on art because it didn't get you money and now you're trying ai art
How is it s far
let me guess u dont have a gpu and came here to get images
What's gpu
whats art
Idk images I think
My friend does ai art but he's busy with irl stuff so I didn't wanna bother him
But I'm fr what's gpu
I don't speak gif
we arent speaking we typing 👍
Btw how long did it take you to master ai art stuff
im no master nubby and other ppl here who make loras those are the real G's
nubby is the whole Wikipedia at this point
I'm sorry, but as an AI language model, I cannot create things but i can help you with resources on how to create your own models.
nubby is typing ...
My laptop is broken
And I don't think I have the mental capacity to understand those things
🔥✍️
Those are some pretty good neg embeddings. I have a folder with around 20~30 of them, but I have no clue of which one is good or not. thx ❤️
looks like 2.1
the quality or the size?
the style
ahh im not sure, havent' used 2.1
but i used reproduction SDXL an anime trained model for this image
yea almost of all sdxl anime models look like that since they dont have the nai thing
goodmorning everyone
gm ☕
gm
☕
How long does that dream bot take to generate stuff usually
i would say pretty fast
it was generating images in less than a min for a huge crowd when SDXL was coming out
(medium full shot:1.1), two women on a couch in the living room, (highly detailed, beautiful, aesthetic:1.1), cinematic, hyper realistic, dynamic angle
used a bunch of NP tho
the quality isn't all that great for some reason, i wonder
Many people are complaining about how it's not generating anything
Uh this is frustrating just when I got excited about something
How do you change angles?
odd, didn't seem that way a couple of months ago
it do be like that with free services
first prompt
different types of angles
or it cood be Lora
could*
My brain hurts
was using unaestheticXLv1 TI for sdxl
wtih very limited np
the anatomical anomaly is rampant with just the TI... another example..
are you using refiner?
interesting
you must have done the merging like a pro for those images to come out so crisp
and the worst part? i'm still not satisfied
i know where the errors are, so right now downloading more models
ahh . lol .. im having problem with regular images trying to get the tones right
and weird body parts
on sdxl or sd 1.5?
no, i don't think it is
the best would be to use no negatives, and then add negatives once you find a nice prompt
at least, that's how i work
i got that trimmed to unaestheticXLv1, (text, logo, title, signature, username:1.1)
ah. nah, i'mma use none of those 😄
Pain of merging and making models
I’ve been stuck on the same Lora for a few days
I think I got it though..tomorrow
for me it's the part where i have to test all the models on all my prompts, then look for the models which actually perform better in the areas that mine doesn't -> then start making pre-merges, then start adding that back in using block weighting
that's at least 1000 images again 😮
and then im usually so braindead I can't really think
since it's all subjective Anyways

and then you're like "i'm pretty sure i'm not done yet, what if i didnt'find the optimal settings yet"
especially because the Lora I'm training is an OC / unique character with a weird dataset
NO LITERALLY
DOH had 0/6 instead of 0.6 ... sigh
I spent HOURS switching between prodigy and adaptadam trying to see which was optimal
bc people say prodigy but
yeah
blah blah
it's always "but what if..."
and then "I might as well do this"
+1 more hour and 1000 more images
sounds like me with prompt saved style idea's
but it's so hard bc it's all subjective and so random so getting "optimal" is either luck or how you're feeling that day
i feel like i should instruct the mod word by word a woman, on a couch, in the living room, watching tv
tho I did one merge and it came out pretty good
prose works pretty well if it's short
I use it for backgrounds
I find it especially better than prompt salad...which can happen
like one part prose
@wispy canopy it would be helpful to get other input on my stuff so I'm not going crazy, none of my friends/peers are into AI stuff in the slightest so
anyway
on tagging images for training?
not training, i meant for regular image prompting :p
hey nubby, that looks brilliant in tone, texture and expression
ty, it's just Kakigori doing what Kakigori do.
ah ok, but i dunno what kakigori is .. model, art style?
it makes the cute
oh cool 🙂 thanks
ok fixing that.. wow
gonna add that to my collection of mod 🙂
and of course I used my merge : https://civitai.com/models/115104/elegantrpganimemagick
huh and I closed boom3d Sound program before loading auto1111 and none of hte errors happened
Really neat concept! I think a pass through SD Upscale or Tiled Diffusion upscaling could work wonders on it.
wtf
hmm now wtf would a audio program interact with auto1111 or visa versa
shrugs
hmm my merge is giving me some interesting results. lets try a not so strange prompt
why does higher /IT speed numbers take longer? (confused look)
generating a small, simple image with a low level of detail may only require a few hundred iterations. However, generating a large, complex image with a high level of detail may require thousands or even millions of iterations.
goes back to counterfeit model withsame prompt and seed
using restart sampler as before
takes out that other hyper network that was not working before
i made rice in the microwave
its like an instant rice package thingy im about to see if its done
that image i just posted?
yea
im sure theres anime with srimilar style but i cant think of any i dont watch a lot of anime
thinking of an animation in that style just sounds cool
ahh that mod nubby made i just downloaded
ooh thats kakigori
yes
nice
i like it, very perky and bright
i need to learn to animate a stick fight in blender so i can learn to vid2vid in diffusion
yuss
like thjaat
oh there are some great one's out there and some people have made them into games that are greeat to watch others play. tho the one with the spiders is creepy. shivers
huh.. where did miku go?
where miku
pos promp: Miku, pov facing, illustration, 3.5D, cutout, transparent white background, rice paper, elegant detail, dry water color <lora:Ink scenery:0.5>magic touches, Space elements, time elements, multi-dimensional, multi-color Ink style, black outlined, back lit hexagonal backdrop-panel, lightly shaded, (magazine cover, detailed text, bar code)
doe sanyone play games like stellaris or anything here
eh not those types of games ,my adhd brain gets overloaded, to much to keep track of
there has to do ((miku))
i only play gacha games,i have a gambling addiction and i am in severe debt
based on what i know about steven segal that checks out
i like gacha games i wish guardian tales was on pc
hmm gacha games = loot chests with keys?
The action movie star is now available as a commander in the game.
World of Warships: How to Dominate with Aircraft Carriers in 60 Seconds
https://www.youtube.com/watch?v=FtktAlMqrvs
World of Warships: Winning with Battleships in 60 Seconds
https://www.youtube.com/watch?v=VdDTLk0IOhc
----------------------------------
Follow IGN for more!
--...
are you using illustrator?
The final brawl between two Legendary fighters
Watch Part 1 here- https://www.youtube.com/watch?v=d_qGO4GrbQM
Watch CENSORED version here- https://www.youtube.com/watch?v=XEtHpGSSTOU
Support~
Get your Merch here- https://jhanzou.creator-spring.com/
Become a Patron- https://www.patreon.com/Jhanzori
Socials~
Twitter- https://twitter.com/Jhanzor...
no thats from kagigori
i was trying to see if i could get something simular to what ur getting
thought you were using illustrator for drawing, which is kinda hard to do ..
with metadata if you load up on PNG Info
using these settings
Did you guys see this?
Like DragGAN except it's diffusion
I hope someone here is able to test it out
that helped a lot i had my cfg way too high
nice, usually you should keep CFG in between 7 to 11 for most tasks with SD 1.5, but for SDXL you might need to use cfg 10 as base standards
the lower you keep CFG value the more creative the Ai gets in rendering, and higher value would follow the prompt more strictly
Gotta look at the units. It switches automatically from IT/S to S/IT at 1. Once it reaches S/IT, the higher the number, the slower it's going.
ahh
the same thing in cars with km/l or l/km, i suppose that's the one thing freedom country has better: it's always miles per gallon?
imperial gallon or us gallon?
oh fck me XD
how about we all go to metric already
Heavy equipment can be measured sometimes in gallons per mile, though at that point it's often more sensible to measure gallons per hour or hours per gallon.
As an American. This! So much this!
New styles, freshly trained after a good nights sleep 
Some came out good, a few came out bad, but they work 
that does actually make sense -> my car is rated in liters / kilometer, which is a totally nonsensical
especially once you start getting good economy
#7 thiccness
and i'm doing merging again, hoping to improve my model
How much is 1 mile?
Americans: erm, its 5280 feet 
How much is 1 kilometer?
Everyone else: 1000 
XD
why are they using their feet to measure distances
and how much water goes into 1 decimeter cube? -> 1 litre
how much does 1 metric cube of water weigh -> 1 tonne
it's almost as if they actually thought it out 😮
Thats... 
IHateSpicy style, V3
wait, it's all 1.5?
i watched a video earlier that was measuring global electricity usage by number of chickens cooked per second
They got some funky kinks
Exactly XD
always has been
(changed my question for you @magic wave <3)
The only American measurement system I prefer is fahrenheit. And that's just because smaller units of measure makes differences more immediately apparent. The 0 point is nonsensical, but also isn't really important.
Out of the ones I trained last night tho this one is probs my favorite
and the 2 anime ones, the harayutaka/anime screenshot style one, and overlord screenshot style one, but those are kinda personal
Fair XD
I don't understand farenheit, but I think celcius is fine, I think of it as like in ranges of 5, 20 to 25, you're chillin, 26 to 30, you're spicy, 30 to 35 you're boiling, 36 to 40 you're melting
All you need to remember is 32° F is the freezing point of water and 70ish is a comfortable air conditioned room.
I mean, for celcius
0° C water freezes, 100° C water boils
The nice round numbers really make it easy
xD
And 20° C you set your air conditioner and you're vibing
Its all nice round numbers
i just keep my windows open even when it's 10° C outside
americans when they go to a japanese hotel,they heat water to 70c and take a shower
XD
https://www.youtube.com/shorts/xuoaseF2xjQ?feature=share Reminded me of this meme xD
70f is cold to me i live in florida tho
i mea not freezing cold but like
im comfy in 80f even though my rommmate hates it
Celsius is a much more logical system, I don't disagree with that. I'm familiar enough with it since half my family lives in Canada.
as far as international news goes, i think the part florida might be the problem
Florida? 
Oh I didn't know Canada used Celsius, makes sense tho :o
i wish i had not been taught imperial my whole life
metric always feels like a conversion and imperial is standardized in my brain
Canada is a silly place
dont'worry about that tho. i mean, people should use the thing they're most comfortable with
i don't understand 'a cup of flour' tho -> i mean, i get they're somewhat standardized over there, but... ugh
you don't measure ingredients by volume xD
never thought about that b4
makes sense u can press a lot of flour into a cup or fluff it up
ping ping ping , this the answer
i understand grams well but if someone says their weight or height in cm or kg it doesnt click like with lbs and feet
likewise in reverse 😛
well, knowing a foot is ~30cm is pretty easy tho
Yeah we just have a collection of scoops of different sizes. It's not accurate, that's why family recipes in America are never as good unless Grandma makes it, because she's been cooking by feeling her whole life and just knows the right amount XD.
even worse, you can hardly get those scoops over here
reading freedom recipes always weirds me out bigtime XD
What's funny is that kitchen scales are not uncommon here, but most people just use them to measure out portions for dieting rather than actually accurately measuring dry ingredients
i used to be able to just fast no problem but now i get so hungry
I find smaller portions, eaten slowly to be a lot easier and more effective than fasting.
I drink tea instead of everything

I have the infinite hunger thing in me
I forgot the name
some chemicals on the brain go brrr and I feel like I want to eat even after eating lots of things
i get that
So I drink like
2-3 liters of tea every day 
(However many balloons of air that is in american units)
How well would these loras work well on a more 3d model
is she like one of those submarine girls
5 fridges worth of air
i should make some tea
Don't worry, we can buy soft drinks in 2 liter bottles so we know that one.
O, I didn't know that XD
I mean, they seem to be overtrained, trained with Anylora
Its a good model
But it ovetrains
So it probably works decently well
@magic wave
can I get one
Sure, which one do you want? I'll gen some pics with it and upload it
Like upload it to my civit page
This one

ASUS confirmed i had falty Intel i7 cpu so had to return the laptop
So Now i am thinking of buying a computer
Is this ok for SD?
Mainly asking because the CPU if it compatible
billions of dollars O:







