#🍥|anime
1 messages · Page 194 of 1
none
i think there is a workaround to vae on comfy but with a1111 not possible as of now
or use a 1.5 model with baked vae
and this is workig good?
My 1.5 without a Vae:
thats the thing ... whether set as none or selected you cant have two instances for one image
add a vae to it then
Art
I can't tell if this was because there wasn't a VAE
Or if it was just destined to implode regardless
check the console and see what it loaded or not
yea prob need to use comfy to do that 😔
I've been meaning to set it up but have never gotten around to it
Oh boy, time to extract ComfyUI
💀
if you are going to dive into comfy try this link https://openart.ai/workflows/academy
hold on a sec
this is a great tutorial to get it installed https://youtu.be/aL4xrbg4NjU
So how about I make an all around ComfyUI tutorial series that'll make you a God at anything in ComfyUI: SD15, SDXL, AnimateDiff, Stable Video, IPAdapters, Photomaker, FaceID, LatentID, and so much more.
Let's stop beating around the bush and really dive in. It's time to become experts.
The repos are located here:
https://github.com/comfyano...
easy peasy
i found the basics of comfy ui fun and easy to understand but later when dealing with more complex workflow i kinda gave up again
Aren't there like Blender nodes and shit on it
i dunno about blender but comfyui complex workflow requires lots and lots of noodles
I totally forgot about the VAE
But
It also got the outfit wrong again
So nothing of value except for one minute of my time was lost
Get offset lora
yeah offset lora is good
I'll have to look for one then
i gave you the link on top
But on other news old model woohoo
It doesn't work for SDXL though 💀
I am now on SD 1.5
This model is the official stabilityai fine-tuned Lora model and is only used as a carrying Source URL: https://huggingface.co/stabilityai/stable-d...
Yes
i havent been using base models a whole lot after i got hold of some commuinity models later in the months
i should try it out again with blue pencil
a cute woman indoors at desk writing on paper the context isnt all that bad with SD ...
do the xyz plot .. prompt s/r test
start with maybe 0.2
Starting with 0.25
where is that error from?
ComfyUI
hey @native halo the exchange between you and @raven pecan is hella fun to read
that from following consumpitons guide?
It's still there
I just extracted the folder I had lying around when I installed it
follow that video i shared, he explains it precisely and easy to understand
the portable zip file approach is fine and works but i find the video guide as fun as installing a1111
the color itself now looks as cute as her face
It got her hair wrong though
cant tell
It's supposed to be long
but whatever you did to the image contrast dont mess it up
did you prompt for long hair tho?
uh lol, shouldn't have to put weight for hair
Give me like
guy`s what mean this is loading in lowvram mode 1924.6048669815063
i am try use SDXL in Forge
AAAAAA
MY EYES
best quality, masterpiece, detailed eyes,
1girl, anime, petite, (long ashen-silver hair), hair between eyes, red eyes, black oversized t-shirt, black and red earrings, pale skin, gothic interior, livingroom, standing, night time, moonlight,
dark, caustics, dimly lit, gloomy background,
HDR, shadow, upper body, highres, looking at viewer, depth of field, realistic shadows, cinematic lighting,
<lora:add_detail:0.8> <lora:offset_0.2:0.30>
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, oversaturated colors, bad composition, too bright, border, cell-shaded, white light, random lights, poorly drawn hands, bad hands, outline, blurry, comic:1.5, chromatic aberration, rim lighting, low saturation, grainy, high contrast,
Steps: 37, Sampler: Euler a, CFG scale: 5.5, Seed: 2811616487, Size: 640x960, Model hash: 1449e5b0b9, Denoising strength: 0.18, Hires upscale: 1.5, Hires upscaler: R-ESRGAN 4x+ Anime6B, Lora hashes: "add_detail: 7c6bad76eb54, offset_0.2: 8e3e833226b3", Version: v1.7.0
@gloomy scroll
this isnt bad
WHY IS IT BRIGHTER NOW
i got it hahaha
what exchange you talking about, he did not exchange without moni 😏
and just to make him jealous imma send this 😈
without paying
What model is that?
Animagine
I'm not sure if that offset lora does anything with that low weight
`night time, moonlight,
dark, caustics, dimly lit, gloomy background its playing out a combination of those
try with this one https://civitai.com/models/199934?modelVersionId=325300
If you like my work, please consider supporting me ! For more information, please visit Patreon ! 本モデルの概要 (Overview of this model) 本モデルは CreativeML...
Can I see some samples from it
i can tell you guys are two fun real life buddies 🙂
theres only one sample of it sorry need more downloads so ppl can post more pics
Does it have these annoying rim lights
@arc the shadows and shades are coming from those keywords
So remove all of them or just the ones above
`night time, moonlight,
dark, caustics, dimly lit, gloomy background,
`
no i would say those are working as you prompted
its not too bright from what you are asking it to produce
some models have them,even when u dont prompt for it,best way to get the best model is to download a lot of models and do a x/y/z plot with same prompt to test them all
you could remove them if you wanted different result
Right so this is without night time and moonlight
try with (dark room:1.8)
Tried out with AutismmixSDXL
(dark room,night,dark:1.8) and in negative ( moonlight,sunlight,lamp,light,window light:1.8)
If you just remove all that has something to do with "light" and add like night, window
just try with the model i linked,the one u using suxs
with 1.5 no lightning at all
If only.
1girl, anime, petite, (long ashen-silver hair), hair between eyes, red eyes, black oversized t-shirt, black and red earrings, pale skin, gothic interior, livingroom, standing, night time, moonlight, upper body, looking at viewer, depth of field
im using offset and detail lora
I'll try it out
waste of vram those two loras

wtihout lora
really?
i hardly notice any difference w/ or w/o them
cdtuner can do both directly without using a lora
cd turner is what?
extension
Why did her hair turn red
this looks interesting Adjust color tone, brightness, and level of detailing using a different mechanism of LoRA or ADtailer
cdtuner with very high contrast
Man
I could just
Do that on like GIMP
but this one sends the dark color it directly to the noise it doesnt add it after generating img
so that extension is supposed to work with both sd and xl?
yea it works with any model, 1.4,1.5,2.1,sdxl,animatediff too
heres another with +20 contrast
nice, but does it get overwhelming with all the sliders or do you generally keep default settings?
@native halo 
if u use contrast,brightness,details or saturation to the max it gets noisy and gives u the heavy filmgrain effect
my goth gf 
you see when i use the offset and detail lora i only change the value to 0.25 which you can also set in your styles instead of having to change the values everytime, but im wondering the use case scenario with cd tuner, and if it requires changing parameters often
if u wanna use an old pic values just load the img on png info then send it to text2img or img2img and it will load the values from that pic into cdtuner
ok. i was looking at this video on youtube that shows all the sliders, tbh i feel a bit intimidated having to change those values for each image or so 😐
nah just change em once and they stay that way,when u close and open webui again just press the blue arrow and the values will load again
ok i think i could save those values to a1111 as well like every other extensions
yea it saves them in png so u can load them tru png info tab
i mean this under settings in a1111
it writes all the change you make with any extensions and it writes to a file so that becomes your default a1111 settings
Ah yes, back to the classics
@native halo Are you hurt anywhere mister
Oh shit it finally finished downloading
my 👖
Kek
here take this one with (dark,spread legs:1.8)
dark
very nice
Too small to see

eventually. all the problems were in the model
what prompt and sampler?
best quality, masterpiece, detailed eyes,
1girl, (anime), petite, short, (long ashen-silver hair), hair between eyes, red eyes, black oversized t-shirt, black and red earrings, pale skin, gothic interior, livingroom, standing, night time, moonlight,
dark:1.5, caustics, dimly lit, gloomy background, dark room:1.8,
shadow, headshot, highres, looking at viewer, depth of field, dynamic angle,
<lora:add_detail:0.8> <lora:offset_0.2:0.40>
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, oversaturated colors, bad composition, too bright, border, cell-shaded, white light, random lights, poorly drawn hands, bad hands, outline, blurry, comic:1.5, chromatic aberration, rim light, low saturation, grainy, realistic
Steps: 37, Sampler: Euler a, CFG scale: 5.5, Seed: 593500941, Size: 640x896, Model hash: a88dd878f7, Denoising strength: 0.18, ADetailer model: face_yolov8n.pt, ADetailer confidence: 0.3, ADetailer dilate/erode: 4, ADetailer mask blur: 4, ADetailer denoising strength: 0.4, ADetailer inpaint only masked: True, ADetailer inpaint padding: 32, ADetailer version: 23.6.2, Hires upscale: 1.25, Hires upscaler: R-ESRGAN 4x+ Anime6B, Lora hashes: "add_detail: 7c6bad76eb54, offset_0.2: 8e3e833226b3", Version: v1.7.0
this model works best with dpm 2m karras sampler, CFG 8 or 10 steps 30 or more,and this neg prompt negativeXL_D,unaestheticXL_Sky3.1,(worst quality:1.5),(low quality:1.5),(normal quality:1.5),(monochrome),(grayscale),(watermark),(white letters),signature, username,text,error,(manicure),(nsfw),(earing),(lipgloss)
dont fill the neg with too many words its useless specially in sdxl
Weird. to me it looks like 2.5D or 3D
anyidea how to improve eye details?
but i am want more Anime style
is this FlatpieceCoreXL A1818 or A1579?
A1818
are you using adetailer
try with this prompt 4k wallpaper,highres,kawaii,1 girl, solo,light smile, open mouth,ponytail,frost eyes, ,simple background Sampler: DPM++ 2M Karras Cfg : 10 Steps : 40
I used mesh eyes only but it didnt improve
add a description of the eyes to the prompt. but initially it is better to look at the eyes in the examples for the model.
but you can try adding “detailed” to the prompt to describe the eyes ("extra detailed")
now look more better! and what influenced? increase in CFG?
yea u also need this two negatives embedings negativeXL_D,unaestheticXL_Sky3.1
drop links pls!🙏
I tried it like this, ahh cant see a much of a difference, ehh, when its generating its pretty good, but after 100% it gets bad
Sky 3.1 https://civitai.com/models/119032?modelVersionId=175819 and negativeXL_D https://civitai.com/models/118418/negativexl
いろいろあるので好みのやつを使ってください Of course, since there are various options, please use your preferred one. 強弱や組み合わせもありです It can be strong or weak or a combin...
https://huggingface.co/gsdf/CounterfeitXL Negative Embeddings D…fixed
what about use After detailer extention?
after detailer or adetailer?
in list on extentions it name !After Detailer
@native halo first link how vertion better sky3.1 or last?
the sky3.1
it gets worse
try a different sampler or more steps like 40+
yeah I did with 150 but every time it gets like this afte 100%
try with sde sampler or unipc
what about more down setting in Adetailer?
while generating, its perfect!
ip-adapter....
after its booom
mediapipe maybe this is the problem try with faceyolov8n
@native halo btw Cfg more 10 only for this is model or any SDXL model`s?
idk i dont use other sdxl models 😁
genie in a bottle
hey hey this is AOM3B2 orangemix !?
yah
and what you want from eyes in this is model? 😁
hmm I just need them to be detailed, eyes....
just watch examples any orangemix model`s
I would say that such blurry eyes are a feature of these models.
@native halo i am buy coffee for you
you want these eyes?
7 min for SDXL on my 6Gb no bad....
hmm yeah kinda like dat
50 secs here with high res fix and 29 without high res fix
depends on model
rich gpu?
no just a 3060
he just rich boy!
ehh-? 3060? but I feel like my generations are slower than that, got 4070
i am better ! with my 1660
im using the new a1111 webui-forge the normal a1111 sucks for sdxl
true!
only for sdxl? or overall?
without new webUI in my case i am dont have any chance use SDXL
added whole bunch of words and kinda fine now
ovelall
in Forge SD 1.5 i am get pic 2 min
for sdxl the diff is huge for 1.5 the difference is not that big
in old WebUi it 4-5 min
the difference is bigger on older cards 
of course! as stated by the author.
and besides, at least I can try SDXL
Why is this?! 🥴

@gloomy scroll @native halo can you tell me the models or lora`s of watercolor style for SDXL
none
😭
Oh I forgot! this whole time I was using face resto, and turned off, bam its good
big bam

😁
@brittle rampart
watercolor \(medium\)
of course! this is the LAW!😁
What are we cooking?
life is hard!
Uninstall genshin ??? Profit
@brittle rampart she's saying food, probably

nah genshin is the only game I play
whoa whoa whats dat model
Pony diffusion
SDXL
interesting
Yes
now thats not food
cut off ears and sold them. ate. now requires love!


wdym?
"i'd eat it" is the first phrase that comes to mind
I see I see
All the food
FuNya Explorer
no boxing in the kitchen

what is happening?
CLIP SKIP+VAE
the preview images are less than accurate
Also you don't need that much negative prompt
I was genning with:
And it was fine
same
I only use specific keywords if i have to, otherwise nothing
on this i have 'hood' in the negatives just so it doesn't try to cover her ears
even then, not always needed
Wow cute
No u
You need SDXL vae
Civitai I think
lemme find a one
Just Google, should be easy to find o:

I'll make... Hamster girl
i did a mouse girl yesterday
try
It's literally you
Hand a bit bad Cx don't mind that

I put adorable in the prompt so.. It's adorable right
hmm look like a mouse girl
hamsters look like mice

Or do mice look like hamsters??
idk why your imgs look like that,heres a few i made with it
anyway - mouse..
maybe because of the long negative prompt?
maybe because im using cfg 11
both
same
XL models don't "work properly" with long negatives, and they don't need them
hell, even 1.5 models these days don't need 90% of the garbage we put in by habit
1.5 dont need them either
my own try
oki-doki! try again
ohhh
.
😬
Clip Skip =1 ?!
Thanks
hmm, someone gimme a prompt to try
Are the capacities idle?
Clip skip doesn't matter for sdxl
really?!
testing a new merge i made yesterday
Yas
clip skip "should" be 1 in sdxl
1 girl, sitting on a bench, drinking water from a water bottle, under a tree
but it doesnt matter
BREAK
(illustrattion of preteen neko girl:1.65 (walks along a country road with a basket of groceries in his hands:1.45)),(puffy small cat ears:1.5),cat tail,brown cat ears,curly hair,hair between eyes,((ponytails,bows on ponytails)),(cat bell:1.4),(glowing lightbrown eyes:1.35),(black maid victorian clothes:1.3),(arms outstretched:1.35),picturesque,perfect correct anatomy hands,serious,country road,sun day,
BREAK
(light romantic theme:1.4),(chataster illustration:1.2),(intricately angle shot:1.4),obscure light and shadow,((soft shadow)),chromatic aberration,ambient occlusion,cinematic light,depth of field,golden ratio```
lol, i copied @junior quest prompt from the image
wtf is dat
oh Im so confused
haha
Have u guys ever tried using ur own name for prompt
apparently i can't post the prompt because the server bots don't like it, but it's in the image
no,i only create anime chars im not an anime char
oh good idea
tell me, can I “draw” prompts?!
Tell me what things u use to make ur stable diffusion faster
money
I'm also on 4gb of vram so
Use another PC instead of your own
👍
steal a gpu 👍
steal the whole tower,even the monitor 🤠
Kidnap the people who can make good pc
seems like it only cant do my prompts, so picky
Steal the parts
Something is different

light romantic theme, chataster illustration, cinematic lighting,```
yeah but good enough I would say
why can't I do the same!
?
it turns out that the positive prompt shouldn’t be very long?
well, it can be if it needs to be, but the more words you add the more it can get confused, sdxl especially needs less to work with and a lot of the "master piece, high quality," type keywords aren't needed at all
so it just waters down the prompt
i always starts with the basics, like "a girl walking down a country road wearing a maid outfit" and then add details IF they are needed
in this case i just took out anything i didn't think was needed from what you pasted
is dat it? not good as this watercolor
it's missing a few things like box in the hair and a bell around her neck, but that can be fine tuned
understand...
wats dis model?
depends, your girl is not watercolor
the background is though
both are too coherent to be watercolor
ahh its not good enough

because that model is no good
wdym? dat model is not goood?
this is also from dat model tho
animagine
its good if you like images to look like they were over polished with oil
yeah I need
oily skins
then yes it depends on your standards
but but when I exlpain more, it gets worse

it looks like the generic poo you see on google imgs when u type "ai art"
oh ma god

Anime looks like AI

@gloomy scroll
@native halo
make this style with ai 
Easy

VivaldiXL Open DallE 3 Anime Model This model was fine-tuned with DallE 3 output. The base training model is the amazing Confetti Comrade Mix . Tha...
Try Autism
Search for that in civitai
Kek
i already have autism 🗿
🗿
is it XL?
It's a pony merge
that artstyle kinda hmm nt for me
srry i dont feel like downgrading 💅
Then prompt for another

What anime model is currently better than Pony based ones?
Horse based when

the ones that can create thousands of anime chars,runs faster,can do 4x or 8x upscales,have access to all of controlnet models
Who needs anime chars anyway
anime models,otherwise they cant be called anime models if they cant prompt anime chars 
From what I have seen SDXL models are better than 1.5 but yeah 1.5 models have more stuff trained and doesn't require as good hardware
also steven seagal
yeah, for loras and embeddings, vae, etc, 1.5 is way better
anyone know how to turn hugginface to dark?
but, a well trained sdxl model doesn't need a lot of that stuff since it can generally cover more areas
but theres always a tradeoff
I believe you need an account
But it looks ugly af
just reduce your monitor brightness to 0.10 🧠
always down to 0
and always night light on

just remove your own eyes,darkness forever
Dark Reader browser extension
nope it doesnt work on that site
deym how did I miss that
Yes it works
nooo
literally me
oh wait it is
close enough?
it didnt work for me earlier, anyway thanks
no hands and no symbols either 😔
you didnt specify
the pattern and the color needs to change in every symbol theres also gradients in every color and colored shadows
all to say it's ugly in the end
to be fair, it was a 6 word prompt, i didnt try very hard

you can do it in csp in like 30mins
Make the book illuminate the room
CounterfeitXL?
😄
csp?
clip studio paint
who has 30 minutes? i could train a new lora and generate a few thousand images in that time
even with 1million imgs you still couldnt do those patterns ( yet )
i probably could, but it's not worth the time to figure out the prompt, it would be easier to train a lora to do it
a painter? or a generator?
in the future you probably could,right now u still cant otherwise we would have those styles in naiv3
maybe with a new release of ipadapter
regional prompts
only patterns u can create there are dots
its like photoshop
I see
a book with summon circle
was generating that before you said something 😄
need to figure out why this model is eating all my ram though
wats tha model?
i had a similar experience in the past.
maybe you need to trim the model
how do you even merge sdxl? I tried but failed
once i've tested it a bit, i'll probably put it on civit
probably
i made it for a specific thing my normal model can't do, and it does that well, but now i'm testing for more general use
take 2 models, punch them in da face till they became paste, mix them and punch them in da face again to merge them
lol, basically
to trim them, punch them in da face till they don't condense
it's the same as merging 1.5, shouldn't have an issue
unless it's a vram thing, i don't think i ever paid attention to that
ehh but I ran out of vram or smhing I gues
whats the gpu ?
yeah, i have a 4090, so 24gb
ohhhh sh*ttt
f me
deymmmm
got 4070 its going 100% even in sd 1.5
got any link for your civitai profile?
i haven't posted much on it yet, nothing released
my usual model has a lot of custom face training and i don't really want to release that, but this one i will if im happy with it
wdym by custom faces?
basically, if i generate a picture of a girl, it's always the same face
oh i see i see
if you search through this channel for my images you'll see most of the faces look similar, or identical
it's more obvious when i generate realistic images, anime is all kinda same-same
U need lots of ram to be able to use supermerger on sdxl
yeah only got 12GB
I'll join magic circle posting
model?
I'll need to find original image since that's old gimp edit
aurora_v10
This ones goood
If you just want model for pose then that's openpose
Whats control lora
what's vail?
Like why there are two inpaint
Controlnet models in lora format that can only be used in comfy and focus
hmm i dunno how to describe
that thing on her hair
Magical barrier

No it's her hair not veil
ehh no not hair
That thing on her head is veil
yeah
I think yes
this one
Yes
has to see if my model could do it 😄
some things like blunt bangs, hime cut, fangs, my normal model doesn't know or handle well
I don't use fangs most of the time since full body images don't rly like fangs
yeah, even with face closeups though my model destroys the teeth
Sometimes upper teeth, fangs help since it often does row of teeth and then adds fangs
did some merging for this new model specifically for more clean cute/chibi stuff and it handles some other stuff differently/better
lol
hi
thank you !
ehhh what is this model
which one?
oh, discord is scrolling me to the wrong thing
you mean the miku with the donut?
its a custom model i'm playing with
control loras work in a1111
ehh I see
its goood
I like it a lot
it's ok for certain things, i might just release it as is and let others test it if they want to 😛
just a controlnet model in lora format. smaller footprint, less vram used, but sometimes not quite as accurate. you use them the same way you use any other controlnet model (and put them in the same folder). they do work in a1111.
Nothing super special, just merged a few models together to get a specific look I was after for chibi characters. It can do more, but I haven't ful...
i uploaded if you want to try it


Forgot I did this image
figures civit breaks when i want to post things
og
Should have added like black eyes or something..it made the eyes red with the lips
Time to learn inpainting
Could have just edited it when I was doing some color correction
But I was lazy
color bleed is annoying, ask for red lips and get red lips, eyes, makeup, scarf, shirt, patch on jacket, lighting, canopy covers at the shops along the street...
looks like your CFG was a tad too high as well, her right eye is actually blue but has a red dot in it, and the spots on the jacket are from that
Was it cutoff that was meant to help with this?
I usually just lower the weight on the color until it goes away, like (purple:0.3) or whatever
I think between training and merges, some models just end up with very heavy weights on specific colors
otherwise you have to specify each color of everything else
I see
looks like the cutoff extension (if that's what you meant) actually modifies the token influence on the unet, which basically means it's modifying the prompt (kinda) to change the gen
looks useful for that, but i'd prefer to just weight things as needed
I was playing with weighted colors a couple days ago, purple (her eyes) likes to bleed a lot in my model so that's turned down, but white/black is also weighted for her hair coloring and using an extension like that would make this more difficult
sometimes the purple bleeds anyway though
Just add purple eyes to the prompt.
If you don't get the same results then just use img2img
🤓
I'm a fast learner with short time memory loss
right, anime little girl wearing a choker and hoodie, short (white:1.38) and black hair, multicolored hair, cat ears, (purple:0.2) eyes,
the problem is that purple eyes IS in the prompt, but doesn't stick to the eyes
Lmaoo
also, why would i waste time on img2img when i can just do the prompt properly to compensate?
especially if i want the same results for the next 20 images i generate
figured out face detailer in comfyui for full body portrait and faces far from 'camera' but I can't for the life of me find a good hand detector for deformed anime hands for SDXL models T_T
Why is my stable diffusion is faster on cpu than gpu
Like this one the most
This ones good too
whats your gpu/cpu/ram?
Cpu Ryzen 5 5500u, Gpu Nvidia GeForce GTX 1650
Ram 8gb
Vram 4gb
which webui are you using?
auto1111? ComfyUI?
Auto1111
and your using all the right optimizations for 4GB vram? https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Optimizations
oh didnt know they work on a1111 
how exactly is a control lora supposed to work?
I thought every CN required an input
idk how it works on a1111 but on comfy u just select the lora then connect the node to the input img
so it's a new node?
in a1111 you just drop it in your controlnet models folder and use it exactly like you would use any other cn model.
ye its like 128mb
idk if its better though since its way smaller than normal ones
In my experience they are generally not quite as good. In a lot of cases the difference isn't enough to care about though.
any suggestions to try to get male characters with pony?
does anyone know any good anime model?
this one
if you can use sdxl based model, autisimMix is good. If only sd1.5 based, then soushiki
thank you
Yep, trying to make her lora
anime girl moving
1.5

sexy 🗿

Noooo

Just for testing try the "shirt lift" prompt

Wait
What if I put lift on other tags
:smuggered:

a sfw version
What you mean
I only and solely post sfw
I am a good discord user
👍
surely
@native halo ain't that right Steve
no 😡

make pictures
@nova remnant @vital raptor
illustrattion of a underage child descends into a swirling vortex of darkness, clocks melt, tyrannical queen with oversized playing cards, sharp angles, distorted perspective,
damn I don't even know. be sad or happy?
i didn't do any upscaling, so yeah... 5.7s
illustration of a pixelated Alice glitches through a digital rabbit hole, encountering fragmented landscapes, morphing creatures, disembodied spirit,
😄
And in the distance is that a mega rabbit?
🤷♂️
she disappeared
Well, no, that’s not interesting... we need to be more serious
in time hole?
Here! This is a much more serious approach to art!
one more 
because he can't count rabbits
idk what happened here with the rabbit
looks like a normal rabbit to me
first person view, laughing doctors performing surgery on a patient, comic illustration
lol
idk whats this
dentist - octopus
no no... this is the birth of the Doctor!

took a while for dalle 😔
exactly what was needed!😁 👍
I did
41 downloads already less gooo
yeah, i woke up to 26 ~3 hours ago
deym thats some good feedback
i think it's mostly because i keep posting images from it, people are probably seeing them in the recent images
yeah, the whole site went down like 5-10 minutes after i uploaded the model
She listens to evanescence
i figure she'd be more Norwegian death metal or something, evanescence is a little too pop mainstream

I don't really know much about popular bands
First thing that came to mind
Emo band 
Her hair got a light streak on it
It's like someone threw paint at her hair through a window opening
Let's assume that this is a ray of light incident from the side
and on the front is a bear's button! 👍🏽 💖
Smol
Sus?🥴

this does not cancel out the kawaiiness! 
cat cafe
cat cafe 😄
nekooo
cafe owner and bartender
illustration, a little witch stands on a stool near a bubbling cauldron, the witch’s room is filled with various things, books, flasks, plants, she looks into the cauldron, stirring the brew with a staff
illustration isn't needed, it's an anime model
😄
Why did cook the witch?
yes, this is anime
@vital raptor I wonder what the negative token should be to get rid of human ears?

