#🍥|anime
1 messages · Page 24 of 1

didnt this one guy win last year with midjourney picture a price :DDD
he got it like 2-3 months before me
「BLUE HUNTER ART COMPETITION AWARD 2022」開催日時:12月18日(日)15時開場、15時30分開始 会場:WITH HARAJUKU HALL『BLUE HUNTER 真夏と時のカイリュウ』のキャスト声優が登壇!長江里加(日下部 真夏 役)、井上 麻里奈(須藤 藍...
this me
doxxed once again
doxxt'd
no, such a lora doesn't exist afaik, straight-on would be a camera shot not taken at an angle. I'm saying I'd like such a lora
Thought that kind of lora would exist by now
how did you get that hairstyle
same man
congratz man
no way lmao
how do you even find out about this contest lol
very nice but upscale pls
haha yeah
art contest
a game?
im guessing
randomally went onto myanimelist and there was like a thing there for it
it's a bit funny to have a random English speaker in this JP vid
did they know it was ai 
ah
I got like 300$ for it
fr
there you go
lol, most these days don't allow ai
thing is i dont have enough money for entrance fee 
PoW and CoW only gets you nitro 
u need to wait 1 year
I don't like this upscaler
then everything is cooled donw
what settings you use?
heyho

and my love 
what's wrong with it?

well there is the new animation thingy now with big money rewards
4x-AnimeSharp
sd upscale script?
Nah animation's way too hard for me
And I do this for fun anyways
RNG pretty much
pink hair long hair
it sounds like they want a music vid length animation too
i sent that to Extra tab to upscale
no no no no no
you doing it wrong

rip
i am trying to get exactly that
don't you also get some kinda badge for winning PoW?
role
Black role 
i mean, if you mean actual just upscaling, then doing what i did, is actually upscaling without changing the image
me first

prompt if u want it
((masterpiece:1.2, best quality)), wide shot, beautiful detailed pink eyes, pink hair, very long hair, freckles, 1girl, solo, small breasts, from above, cowboy shot, expressionless, flowers, looking to the side, side view,
stunning environment, scenic, breathtaking, 4k uhd, HQ, intricate, max detail, white background,
<lora:standingFullBodyWithBackgroundStyle_v10Offset:1>
Negative prompt: (worst quality, low quality:1.4), 2girls, jpeg artifacts, lowres, 3d, realistic, bad anatomy, bad hands, extra digit, fewer digits, missing fingers, nsfw, collar, shirt print, bad-hands-5, EasyNegative, cleavage, bare shoulders, bare back,
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 1108061458, Size: 512x768, Model hash: 1bbeb9eded, Model: CamelliaMix_Line_V2, Denoising strength: 0.4, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires steps: 20, Hires upscaler: 4x_fatal_Anime_500000_G```
but i get what you meant
You get "invitation" ||forced|| to the cult
Good point
no, they both change the image. just extras is worse
hehe except you
with extras?
or img to img
tbh, best current way to upscale probably that multidiffusion thingy
Yep
but sd upscale script good enough


thats what i do but it seems like its wrong?
what upscaler u use
very bad
in the script
If it looks good it looks good 
fatal/animesharp/nmkd

idk man seems like it
yes
i dont want half my image to randomly change tho 
are those others u need to get?
Lower denoising
you can choose how much to change brother
which one
yes
besides extra upscaler allows my photoshops to look more natural
download upscaler

why would make this a difference
my thinking for pow
it splits images into tiles, upscales tiles and combines, and somehow it is indefinitely better
ayo
let me make comparison for you
Goodness gracious

sure would like to see
If it's not for contest I'd say it's not very important

true
for contest, everything probably matters
but i don't see much differences tbh

i can't even tell if one is better than the other
Art is subjective anyways 
look the same
2nd is darker
what settings you use for the img2img one
what settings do you recommend?
i see what u mean with tiles but god damn
i have wonder for a quite sometime now
if this method is possible
it seems like it is
thanks this helps!
now i understand what you mean
??? so much time?
Huge stuff
weird
Good lord the resolution
it overlooks the resolution when using sd upscale script
pieces
and then upscales every tile
??
this is what he does no?
hm maybe 6 pieces
im blind
That looks normal time for me
not very good the right one
cos i can't really tell
it is the extras one
Unless one has this turned off then sd upscaled would be quick
what gpu you have?
1070
what was original?
also maybe do one with the extras tab to compare
should take a lot less time
original
2048x1304?
i think the differences is almost like using xformers
well, a bit worse imo
if you uses xformers, there will always be differences in your generation even when using the same settings and seed
great job i like this style
seems like i will just stick with extras

at least its faster
like it adds more artifacts
which is which tho
right is extras
well, i mean, if you can tell, sure
i can't tell tbh
so
I'll stick with extra
and what i meant was
i know what this tool does, it splits the image in to 4 pieces, then it upscales every single piece once with the selected resolution
it upscales this
2k by 1.3k
and then the other 3 pieces also
then it combines it
how did you get the background like that
and this way ofc the image is way better resolution
and quality
but god damn 8min waiting
worth it
@brittle jay can you try to upscale this? i wanna compare
best quality, long black hair, dark orange eyes, high res, black ninja shoes, dark skin, highly detailed eyes, black pants, black sleeveless turtleneck, shoulder tattoo, big eyes, street, cool, clothes around waist, sfw, fingerless gloves, solo, good anatomy, face details, ((eyes details)) , ((1girl)), fingerless gloves, chill eyes, good anatomy, alley, ((neck fur)), gorgeous, serious, (slender), cloth around waist, casual clothes, pants, black wolf ears, (black red skin color), looking around, wide eyes, dark skin, tactical clothes, detailed eyes, ((dark alley)), (black wolf tail), sitting, cool print on clothes, looking around, lighting, (night), magical, illustration, prints on clothes prompt btw
I think it changes based on model and stuff
what do you want to upscale it
what upscaler to use?
2k?
the one you recommended
x1 upscale?
which is which
animesharp then fatal
thanks
now on extras
it changed the shadow of the hand to another finger
what did you use
ye i set the denoise quite high
nighty night good luck
goodnight! and good luck for test ! 
maybe this is a bit better
when i use script upscale lol
ye thats about right
lol
wow
jesus
awesome
relatable
Why we posting ETA here?
are you try to render movie
Just put max batch count with controlnet and will stop at some point
xD
Hunting for seed for pow
3711987009
708289865
2992465303
294661265
294661263
2961479642
try these
3711987009 is to nsfw for me to post 
wew
708289865 was okay
Just like this
i know right
I guess clyde hasn't gotten any updates
xD
That's ban here
so it is a nsfw image?
Someone got timeout using that method before
not our issue
And because of that, the link embedding was removed for awhile
ic
Just trying to let you know since you are pretty new here 
If you rly want to see it search "I guess this was problem" from my message history
Another infamous way of bypassing Clyde is to draw red lines over your art. It is considered bypassing filter and therefore not allowed
Even if it’s not nsfw
i think I need to play the "how many girls can I fit in one img" game again xD
i hate how good the quality is
Not what i wanted but i take it
Made it myself
ʸᵉᵃʰʰʰ ᴵ'ᵐ ˢᵗᵉᵃˡᶦⁿᵍ ᵗʰᵃᵗ ᵖʳᵒᵐᵖᵗ ᵃˢ ʷᵉˡˡ
You got stealth pnginfo extension?
go ahead lol
no?
as i sayed this is Image2image

Discord strips metadata from images
if u wanna know
Oh
we are getting there
But still your image doesnt have pnginfo since Discord removed it
ok, I'm gonna try the same game starting with 2
jeah this aswell

dude delete that sd upscaler its shit, get this
transform
hm
so cute!
what model do you use?
maybe I should try to make chisato and takina 
can vouch for Tiled Diffusion being really good
iits nijijourney
You use Nijijourney for this one too? I love how it looks like it's a screenshot straight from an anime, heh.
nono this one is lora and bluepencil
can you try this prompt? best quality, 1girl, long hair, green eyes, 4k, royal, female, lion tail, cat ears, detailed face, solo, detailed eyes, neck fur, detailed eyes, light skin color, anime style, emerald eyes, unique hairstyle, cool, gorgeous, ((fluffy hair)), long hair, series, Rome, (Romanian clothes), ((grey hair)), mansion, (red eyeliner:1.22), (in arknight syle:1.1), siege arknight, long hair, lion hair, (tiger hair prints:1.3), (black streaked hair:1.25), sharp lines, sharp details, gorgeous eyes, light green eyes, shiny eyes, cool, sharp eyes, streaked hair, serious

can I add in a style lora?
sure
So if I just enable Tiled Diffusion and Tiled VAE I get perfect images?
trying to get a result similar to this (closest one i got)

I kinda cherry picked this one 
Yeah so about those perfect images with Tiled Diffusion and Tiled VAE...
keep it between 3-4
Yeah I kinda forgot that I just inpainted
my world? 
Submitted it 
Did I get it right?
a bit far but pretty cool
alohahey
https://civitai.com/models/6209/dark-gemini-v3 preview images got prompts now, most of em
So controlnet of this
Used this for controlnet
invert the haircolor and run a long strand of silver on the side, an eye patch over a third eye
what eye patch
Arknights?
yea only lion character i could think of
the one you make
idk man she is not wearing eye patch
or has a third eye
She got helping hand
lol
lifting the cape
nailed the hair tho
the extra gold is nice too
but yellow eyes instead of green
Well I used your example in controlnet with depth map
i should use controlnet sometime
Oh yeah tail turned into hand
and she thinks I didnt know that her aaf
@gloomy scroll oh do you know how to turn a sketch into a drawing?
aye
and keeping the style
ControlNet
Looking for something like this?
I guess this should still do the job even though controlnet has gotten some updates
https://www.reddit.com/r/StableDiffusion/comments/119o71b/a1111_controlnet_extension_explained_like_youre_5/
aight thanks
i dont need to update webui right?
each time i update it, it gets way more laggy
welp never know until you try
abas asoso venitrusk ahatranitsium
true
ancient spell to curse your opponents to have diahrea at he most crucial moment of their life
"1girl, white hair, windy, fluttering hair, fortune ball, aura, white school uniform, flying papers, embarassed, magic"
also this
WOW
looks good to me XD
that a big hat
They are totally not compensating for anything
Didn't prompt for her
oh lol
Heh dark gemini 3 receiving a lot more downloads than 2.1
all thanks to high quality mid diffusion preview images and them being waifus
101 how to make model preview image
- Big

- perfect fingers with controlnet
- Your unique style
well it is pretty unique, I don't need perfect fingers, my images are still pretty eye catching, dont tou think?
@gloomy scroll
also the versatility
Yeah I agree that your model has nice style
there's a lot of styles in one too
even if the composition remains at the core equal
I mean pretty difficult to call these styles the same
I mostly remembered your model for those dark fantasy images
then there these not posted,
It's come a long way since then
a ton of versatility now
Chisato
Java
#1098413490332508270 message here we go
Whatever you do don't look at the hands. DO NOT LOOK AT THE HANDS. 😄
Otherwise this was almost good. lol
use bad hands embed'
model? looks very nice
bad hands is really nice
Pikachu is holding a turtle head in
it's about to pop out explosively
Alright enough of the nonsense, a more civilized image this time. lol
No more abomination Pokemon stuff from me for a minute.
i should finetune the model so it can do this style consistently
"pov, pov hand, holding hand, foreshortening, 1girl, white hair, witch hat, castle, magical old city, old town, flying petals, light particles, moon, night, cloud"
and these
thankyou
I feel like this is memeable.
New scp looking sick
Now I need one facing the opposite direction, with its hand pulled away and "no" as the text.
Which one looks better to you guys? I'm split.
2nd
2nd
Are you going for creepy? The first one. If not, the 2nd one
Hmm, just studying concepts for story characters.
They are kind of....dark. lol
I personally prefer the second myself though.
a Mahiro Chan piece I made. Very proud of this one
As requested
Also might want to get rid of that watermark before you get something perfect and look at it and scream (from personal experience)
Okay, I'm keeping the second image on the black clothes/red eyes version. Now to rework the lighter colored/blue eyed version if I want.
watermark?
Bottom right of the pic
Nothing worse than spending an hour imaging a piece just to notice you've had alien language in the corner since the 1st iteration
xd
Haha, alien language. Facts.
love all the shadows
say, ya'll
speaking of shadows
(I love shadows with leaves, personally)
But what are your favorite kinds of shadows to see on a character?
@dull jackalthaaks i try to experince with diffferent generators
speechless everytime i open this channel
😭frrr
Market stall

mksks style, masterpiece, best quality, extremely detailed, detailed background, 1girl, solo, (traditional media:1.2), hands up, (fantasy adventurer:1.2) clothing, bangs, (embarrassed:1.2) expression, blush, (vibrant colors:1.2), sidelocks, french braid, (sketch:1.25), ([graphite \(medium\)|colored pencil \(medium\)]:1.05), shouting, (angry:1.1), fantasy village background, market stalls with vendors selling their goods, (monochrome:0.7)
Wait i thought its everyoneshira
Nope. That's on AuroraONE

That is a mostly unmodified prompt that I made for EveryoneShira though. I just added the stuff before 1girl.
Anyone wanna try making a banner sized image using my current profile color/theme, I feel like we could get some cool gens out of it 👀
Nice banner rn tho...

that wasn't a microdose
options
Pixelart version
Hey these look amazing, using png info doesn't give me anything.
mind sharing the prompt?
Sure, not sure what's up with PNGINFO 
Yeah i have no idea
Let me send
(realistic,) mature, amelia watson, virtual youtuber, short hair, blonde, choker, masterpiece, best quality, cowboy_shot, standing, realistic eyes, head tilt, smug, looking_at_viewer, mountain, holding_flowers, gentle, kind, sweet```
```ts
Negative prompt: KHFB, auroranegative, (Worst Quality, Low Quality:1.4), claws:1.4, border, anime:1.2, cartoon:1.2, outlines, grayscale,```
```ts
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 3, Seed: 2220907856, Size: 800x600, Model hash: b36d9ac70e, Model: AuroraONE, Denoising strength: 0.5, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires steps: 20, Hires upscaler: 4x_fatal_Anime_500000_G, Hashes: {"vae": "df3c506e51", "embed:AuroraNegative": "e5b5173e82", "embed:KHFB": "adac8597da", "model": "b36d9ac70e"}
tysm
Its a diffuser that goes on top of a model and for most anime models fixes the color and Details.
Here you see how it improves the image.
Normaly models have the vae already included but since most anime models are based on one specific one the vae needs to be added on top manually
VAE : Base would be more fiting, there is always a VAE, you' can't just watch the latents, but nice comparison
Vae base instead of 8400 ?
Yea that added XY script
yeah, it's just auto doing its labelling ^^
Yea but i dont think its base if the vae isnt in there
fixed
no, if using a model that removed its VAE, it will just crash, that's for sure. just to make the comparison valid I mean
Your right. I can test that with shark later
sexy
yeah, when coding my "diffusers comparison tool" (a grid X/Y/Z tool to compare diffusers before making your ckpt), I did play a lot with those tensors. Still can't read the pixel myself through them though :p
I forgot that there is always these two files included in every model.
it's the files I replace in my base model, with the updated VAE
so I don't need to ask people to download an updated vae
Exactly, in shark that can be done manually by drag and drop. Because they convert the models back to folders
Will VAE affect the structure of the final image, or is it just some light and shadow details?
I've never seen the structure change in VAE outputs. but I mean theorically ? it could if you really wanted to create a VAE that does strange things
You see it in the comparison, it doesnt affect the structure output
yeah I see thx
after a major screw up i finally fixed and republished my lora... it is actually a lora now 😅 was accidentally a whole ckpt earlier
How to generate such images
.?
how to use vae's on a1111?
@vivid torrent @outer sundial there is no bot on this server to generate images, check out #1072220168534642768 #1080946152318443610
put any of your vae's in this location: \stable-diffusion-webui\models\VAE
yeah i did this
its not automatically applying
you can click on show all pages and ctrl + f "VAE"
Do you have "sd_vae" in the quicksettings list in User interface settings? If so you will be able to change vaes at the top of the ui like this
how do i enable this to be in the UI?
she wants a headpat
thank you ^^
Hello, i'm still looking for a model that can do cool creatures/fantasy stuff if anyone has a good one 😮
so anyone had the chance to try dark gemini 3 ?
Pretty good
Man if I had the money, I would make an image using a model trained on a certain artist, use controlnet + regional prompter + ton of settings. And then ask the artist to sue me in US court.
nice, harley quinn meets my madman
the madman in question, or one of em
once youve tested it enough please leave a review
hopefully not a condemning one
but do test both complex prompts and simple ones
"Create an image that shows a hand-drawn world map with different icons and elements representing the diversity of the world. The icons can include landmarks, animals, objects, foods, and any other element that represents a country or region. The icons will be connected to each other by a route of suspensive dots, like a treasure map. The image should have an artistic and detailed style, as if it had been hand-drawn with colored pencils. The color palette should be vibrant and diverse to reflect the cultural and geographic diversity of the world. Use an AI program to generate the image."
ah nice going gearpunk
very neat 💖
aye, you should try my model @frozen olive
its the one they using
yeah Im a fucking salesman right now
im trying to master the picture creation, i do use a friends model already
i'm very happy with
so liking it then?
"nevermind the blood, would you like a kiss?"
1girl, woman, gorgeous, (vampire:1.15), (medium neck length black silky hair with red highlights and faded ends:1.15), (red shimmering eyes:1.1), black and white gothic casual dressing, (inside vip club section:1.1), fangs visible, blood stains on lips and cheek, evil smile, mischievous eyes, neon lights, strobe lighting, (digital art:1.2), (digital illustration:1.2), (gothic fantasy art:1.15), (splash art style infusion:1.15), pseudorealism, (2.5D Illustration:1.1), overhead pulsing lights, ZBrush Render, Autodesk Maya, HDR, 8k
(low quality:1.15), no quality, bad quality, crap quality, low resolution, lowres, no detail, low detail, (deformation:1.15), (deformed:1.1), (mutation:1.1), (mutated:1.15), (extra fingers:1.05), (extra toes:1.05), (extra arms:1.05), (extra hands:1.05), (extra digits:1.15), extra limbs, extra eyes, ugly, incomplete, text, title, description, logo, watermark, image artifacts, spaghetti body, conjoined twins, conjoined body parts, big breast, massive breast, major breast, big tits, massive tits, major tits, hentai, eroge, nukige, pornographic, porn, XXX, erotica, skimpy, slutty, nude, nudity, naked, bad_hands
Steps: 15, Sampler: DPM++ SDE Karras, CFG scale: 7, Seed: 3110081806, Size: 1024x640, Model hash: 021972d7ff, Model: DG3_FP32_Full, Clip skip: 2, ENSD: 1, Eta: 0.66
Midfusion upscale:
(Dark Fantasy Art Style:1.25), (splash art style,:1.25), (digital art:1.1), (digital illustration:1.1), (pseudorealism:1.2), (masterpiece:1.25), (best quality:1.25), (highres:1.25), insanely detailed and intricately super detailed, Gigapixel image, (2.5d:1.2), Beauty Filter, 8K, (HDR:1.2), unreal engine 5
Steps: 30, Sampler: DPM++ 2S a, CFG scale: 7, Seed: 3110081807, Size: 2560x1600, Model hash: 021972d7ff, Model: DG3_FP32_Full, Denoising strength: 0.32, Clip skip: 2, ENSD: 1, Tiled Diffusion upscaler: R-ESRGAN General WDN 4xV3, Tiled Diffusion scale factor: 2.5, Tiled Diffusion: "{'Method': 'MultiDiffusion', 'Latent tile width': 96, 'Latent tile height': 96, 'Overlap': 32, 'Tile batch size': 1, 'Upscaler': 'R-ESRGAN General WDN 4xV3', 'Scale factor': 2.5, 'Keep input size': True}", Eta: 0.66
https://civitai.com/models/6209/dark-gemini-v3 go on and try, you can get pruned FP32 for lower file size or FP16 for even lower
you too Vil
Wow
I would appreciate reviews too
gonna get some more vertical images for it though and some wild styles
hey @rough violet @gloomy scroll @brittle jay @frozen olive - she's inviting you all
lol

omg insane trailer for this channel
still testing out dark gemini 3?
yep
let me know if you want prompts for this style, though the ratio for good and bad is about equal and some ends of the style might "look like it would be bad" but it's the more gritty smeared oily style
@rough violet
death by a thousand cuts against the youma causing a lot of blood to gush and burst:1.28)>, rain of blood fills the air, a red mist reaches the knees under the soft dim lighting of the moon and medieval lanterns that fill the sides of the streets (digital art:1.20), (magepunk illustration:1.25), ( Multidimensional art :1.28) Everything and the surroundings freeze to absolute zero and not a single thing that is not frozen:1.25)> (best quality:1.27), (masterpiece:1.27), (highres:1.20), (digital art:1.20), (disaster art:1.25), (apocalyptic art style:1.28) <(feral daylight zombies sprinting the streets and eating humans a silent archer launching arrows into the heads of zombies)> under the scorching sun <(creating wildfires in the forests surrounding the city and smoke to fill the skies and streets (best quality:1.27), (masterpiece:1.27), (highres:1.20), (digital art:1.20), (digital illustration:1.20), (anime-art:1.25)
header, trademark, words, letters, alphabet, copyright, text, title, slogan, subtitle, publication, subheading, description, logo, watermark, extra leg, extra arm, extra finger, no head, no body, half body, cropped, cropped head, cropped body, image cropping, missing finger, extra thumbs, no thumb, noodle fingers, contorted fingers, deformed face, mutated face, melted face, spaghetti face, squished face, unintelligible face, missing eyes, no eyes, goonies eyes, big breast, massive breast, major breast, big tits, massive tits, major tits, hentai, eroge, nukige, pornographic, porn, XXX, erotica, loli, lolicon, skimpy, slutty, nude, nudity, naked, shotacon, [bad_hands0.50], [bad-artist-anime:0.70], [bad_image:0.80]
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 1777468004, Size: 1024x640
for anyone who cares and has my model
it is from a version or two ago in the beta stage
so it might look slightlt different now
Im gonna try regenerate on it
damn
anyone knows how to train eyes lora for inpaint, do I have to just crop her eyes from images then that's it? want to train kurumis eye clock.
just train normal lora then use it to inpaint eyes
should I be proud of my model?
considering its the third overall that I ever made
though it took awhile admittedly
no
jk
you judge for yourself
being proud should only depend on how you see your own model
not how other people view it
not entirely correct. Having my peers tell me I did a good job is an indicator that I've accomplished my goal
So I'm asking those whose opinions I respect
when I do that the eyes of kurumi look horrible idk why
how can you expect other people to know what your goals are specifically?
it's supposed to look like this
first, gather dataset where the eyes are like that, no need to crop only the eyes, just keep the entire image, though make sure there aren't any other kinds of eyes.
then train lora
I can expect others to know aesthetics and model utility, performance, ease of use, whether it generates good or bad images
ohh ok
I see.
let me think about it then
well I'd prefer you test it after reading guide and learning the model and trying out prompts and upscaling with mid-diffusion to see the final look and if that is nice
I think the model lacks a bit of fine details and has too much artifacts. other than that, amazing model
and when I say fine details I mean specific kinds of details
just as eyes
you've just not prompted right then for eyes
cause it can do eyes really well
it can prduce artifacts, yeah
especially in complex prompting, which requires a lot of fine tuning
I am looking at the images you used at the civitai page
these eyes not detailed?
nah, these eyes are good
all depends on the prompt. I mean the primary photo has dead eyes cause she's well...not happy
I mean like in this image
what about those in that pic?
a werewolf and a werewolf in human form, with glowing eyes
those are fine, only tiny artifacts you would look over
what artifact?
you mean the two teeth under his eye?
the tiny broken wing on his arm?
the only artifacts I can see
in here this
I don't see the artifact
no artifact
I see her eye being a bit..wonky
yes details
but you said artifacts
I said on other image
oh under her lip, yeah
I should probably pick another upscale
though if she has fire on her anywhere
that was actually fine in my opinion
thats cause she's a pyromancer
oh wait
this upscaled?
then nevermind my comment
I am like 99% sure upscaler did it
though gotta say, most images have hints of looking burnt
burnt where?
idk how to describe
are you sure it's not your monitor?
cause I can't see anything burnt
anyone else in here provide a second opinion? Either my eyes or monitor can't see it, or it's not there. If it's there. I have to know though
so I can figure out how it got there
it looks like the effect high cfg does
cfg is 7
what's burnt here? except the dragon wood statue
or here
the color grading of the previous "burnt" image is supposed to be flat and depressing though, matte etc
look this comparison, first high cfg second low. do you see what I mean with high cfg effect
happy?
the girl smiling
wdym
that different art styles and prompts will produce different colors, shading etc
from all the preview images I get a slight "burnt" feeling
(Dark Fantasy Art Style:1.30), (Anime Horror Art Style:1.35), (extremely detailed horned half-demon woman with light grey skin and crystal eyes and a (highly detailed beautiful face with (smooth contours:1.16):1.34) and (red wings made of silk on the top of her back:1.20) and relaxes in a massive (victorian gothic mausoleum of [lovecraftian:HR Geiger:0.70] influence:1.16):1.40), solo, cinematic, dark background, dark, backlit, dim light, low light, low key, (digital art:1.20), illustration, (splash art:1.10), (masterpiece:1.35), (best quality:1.35), (highres:1.35)
This prompt, has dim light, low key, horror art style, grey skin, lovecraftian, HR geiger, dark and low light
I don't see it
how is this burnt?
this?
this?
if I could only see it
I mean it this burnt?
this a bit yes
how?
this yes
if only I could describe it accurately
I don't see it at all
@rough violet @gloomy scroll @lyric zealot @quick hill need more eyes
can you guys help out?

are my images burnt?
what img?
all of them apparently
https://civitai.com/models/6209/dark-gemini-v3 all of them
ok one factor in this "burnt" effect I noticed is "oversaturation"
most images aren't very saturated
ia agree with kazoo that is more oversaturated than burntin
burn-in is probably more like setting CFG really high
cfg is 7
and can see the effort immediately
I just call it burning cause it's hard for me to find an appropriate word
also if you zoom into this image, you will understand what artifacts I am talking about. though it seems like most of them are from the upscalers.
I know a few are purposely saturated
that's a negative imo
cause some of them I wanted to give a flat matte feeling
saturation helps
but most of them aren't saturated
that was his prime example of oversaturated
yes
but what I didn't like was being told they were burnt. Not liking saturation is an opinion and I don't really care beyond appreciating them
and sometimes using them
but I thought I had actually generated them wrong or poorly
this image really is not too overly saturated but it still feels a bit "burnt", like these parts
also....this isn't saturated
i don't think that's burnt
it's probably artefact
but not sure tbh
oh that, discoloration and laziness
I knew it was there, I wanted to go to bed, I stopped the generation
as I said, I am just using that word cause I don't have a good appropriate word to use
i appreciate all styles, and are willing to accept any effort put into generation
yup, i will probably call it artefact
the original is colored differently there cause of the light cast
but not sure tbh
No idea, too subjective
its my oc 
super cute OC
ty 







