since i am a prompt nerd and don't know any photographer, i use other prompts:

1. "ideal" replaces "most beautiful girl in the world"

2. "expensive-profi-shot" replaces any photographer

3. "cinematic-style-filter" is used for anything cinematic like lighting and more

4. "warm-light".... no need to explain this one

maybe anyone else who is struggling generating good photos will find those prompts useful....

i honestly hate xl. in the past it was like pony these days.

no prompt was working. not even the most simple one.

but this checkpoint is proof that xl has changed to a higher and better state.

it really is nice to see that anything i prompt in positive or negative is getting attention.

this just proves that xl is in a very high state and pony still is the worst bullshit ever.....

okay, since playing around with this checkpoint has become very addictive, two things to point out:

put in positive

"redhead-hillbilly, blunt-reality"

and in negative

"denim, corpse-like, cadaverously"

negatives are only suggestions, but you will see the reason for yourself pretty fast.

another thing to point out:

if you want 3D toons, use this checkpoint instead of downloading a checkpoint for xl only made to generate 3D toons.

simply prompt "toonified-3D-character" or "cartoonified-3D-character". this checkpoint can handle it perfect.

EDIT: add "smooth-CGI-rendering" for better results

checkpoint or lora for "final fantasy cgi"????

use this checkpoint together with

(((smooth-final-fantasy-CGI-rendering)))

EDIT: and if you add "npc", you'll get very unique characters

checkpoint for pixar style for xl?

just use this checkpoint and prompt

(((disney-pixar-3D-style)))

as a base.....

EDIT: after i saw "judy hopps" and the style used, i just prompted it. prompt

"real-3D" together with the prompt above. it's working

wanna turn this checkpoint into an anime checkpoint?

use the following:

in positive: (((flat-colored-2D-anime-style, flat-anime-details, flat-colors-anime-world)))

in negative: (((2.5D, 3D, photorealistic, reallife, photoshooting)))

results aren't always 100% anime. this checkpoint is trained on realism and therefore you will have mixed results many times. i couldn't stop it with any prompt and it wasn't even what i wanted. it is just for fun to see what this checkpoint can handle.

and one more style is working:

(((dreamworks-pictures-3D-style)))

no preview image for this one, but:

since i stated on my article "using japanese words as prompts", that "kodomo" does not work, i gave up on this. but using this checkpoint, it is working like i always hoped for.

and it looks and feels like i'm flooding this site here. sorry for this.

entschuldige. es macht einfach unglaublich spaß, mal einen richtig guten ordentlichen checkpoint zu bekommen, der alle erwartungen erfüllt. und dann erwischt es auch noch so einen prompt-nerd der allerlei dinge testst daran.

es macht echt einfach irrsinnig viel spaß damit zu experimentieren. da steckt so unsagbar viel potenzial drin, daß ich meine notizen-liste weiter auffüllen musste für das testen meiner prompts.

2 dinge die mir bisher negativ aufgefallen sind, die man aber mit negativ-prompts lösen kann:

maskuline merkmale traten sehr häufig in verbindung mit "naked-apron" auf und ebenfalls sehr häufig waren spitze zähne.

die "cat-ear-maid" in meinem letzten upload hat die besagten prompts in negativ.

ist nichts weltbewegendes, nur eine kleinigkeit.....

cinematic shot lora???

just use this checkpoint and prompt

(((cinematic-shot)))

same style, same effects. all you need to add is your own fantasy

model only for cum on face? well, i already published a lot of prompts for this case, but never tried it on xl.

"messy-cum-shot, 1girl, excessive-facial-mess"

world of warcraft????

(((world-of-warcraft-ingame-screenshot))) 1girl-npc (for example)

something i never got working on any checkpoint was "60+fps". until now.....

(((Ultra-HD-details, 60+fps-details))) in positive and

(((3D-rendering))) in negative

results are ultrasharp resolution images that just look awesome.

eine sache, an der bisher jeder checkpoint gescheitert ist und auch dieser hier keine ausnahme darstellt:

eine zahnlücke. egal ob an kindern oder erwachsenen.

wenn jemand eine idee dafür hat, dann immer her damit :D

really nice to se that "reality-shot" is working exactly the same way like it does for standard sd.

this prompt generates your average everyday person and/or scene.

you will not even once see a beautiful model. all you get to see are people that wear (for example) worn out clothes, torn clothes, dirty streets, damaged cars, people with dirty hair and many other things based on reality.

minecraft?

(((official-minecraft-game-square-box-world-style))) 1female, outdoors

and absolutely no(!!!!!!!!!!!) prompt in negative.

this one, to be honest, was for now the hardest one. lots of trial and error, because it was very inconsistent at the beginning.

"world" is necessary to tell this checkpoint it's far from reality, "box" is for the shapes, because this checkpoint always tried to make characters look real. and as you can see, you can't(!!!!!!!!) prompt "1girl" or "1woman". if you do, you get a real human and nothing minecraft anymore.

sadly adding anything else can make this prompt unstable as this checkpoint sometimes doesn't know how to recreate it in minecraft style. therefore you have to make batches before it's working.

one last thing:

nudity is absolutely impossible. this checkpoint gets super confused the moment you force changes to a human body.

dreamworks style?

(((dreamworks-pictures-3D-movie-style))) in positive and

(((unlike-finger-count, unlikenesses))) in negative.

make them naked and you get a lot of barbies the whole time. can be funny but also annoying.

one thing that works ways better on this checkpoint as on sd photo checkpoints are "flower girls on weddings". many sd checkpoints generate big girls with different dresses and no child girls with traditional dresses and a flower crown.

this checkpoint here does a wonderful job when it comes to flower girls. for starters, let me give you those prompts:

"flower-girl-scattering-petals-from-inside-a-straw-basket, participating-in-a-wedding"

and no, there is absolutely nothing illegal in even one of those prompts. result is a little girl wearing a big flower crown and a white sleeveless dress. sometimes when they get generated, they wear white flat shoes.

there is absolutely no reason for this to get reported or get deleted. i'm just mentioning this, because there are really crazy people here on civitai that go overinsanely in karen-mode when they read "little girl"...

one preview image, but not generated with this checkpoint and also anime. Image posted by themancalledmojo999 (civitai.com)

never even thought about that: tears....

the only prompt i can give you is a negative prompt:

(((wild-mess-of-tears)))

the rest is up to your imagination, because i didn't have even the slightest problem to generate them.

simply tell this checkpoint where those tears should be placed and what those tears are doing and the amount of them....

no need for a bad trained lora for such a simple task.

i noticed something on some of those images:

emma watson lookalike.....

and i didn't laugh that much for a long time after i put that name in negative and clicked that generate button.....

Image post by themancalledmojo999 | Civitai

edit: since i found out the reason for this lookalike, here that negative prompt which was/is the reason for that:

"japanese"

.

. .

. . .

"mix-of-every-human-race" isn't working on this checkpoint. instead i changed it to

"mix-of-any-human-race" and got it working like it should be.

and since this checkpoint gets too confused with some prompts, put

"runner, chaos, race-as-in-running"

in negative and everything will get fixed....

this goes for any realistic checkpoint:

if you get only one type of person generated the whole time, put "facial-similarities" in negative....

eine sache die mir echt sauer aufstösst und weswegen der "emma watson" prompt entstand:

der checkpoint wird mega wütend wenn "japanese" in negativ geprompt wird und generiert super hässliche menschen. Image posted by themancalledmojo999 (civitai.com)

EDIT_ es lässt sich nicht einmal "tarnen" in form von "japanese-similarities" in negativ. alles hat zwar bessere qualität und die mädels haben mehr farbe und sind models, aber die erzwungene hässlichkeit bleibt. hinzu kommt hierbei sogar ein erzwungenes "duckface".

best combination of negative prompts that work like a charm:

(((lack-of-different-joints, saggy, harelips, blurry, foggy, 3D-rendering, CGI, japanese-only, frogface, ingame-graphics, unrealistic, doll-likeness, sty)))

no need to add anything else about anatomy. "japanese-only" has to be. remove "only" and it will become freaky. that's why "frogface" is still there :D

bisher gefundene negativpunkte:

1. japanese in negativ generiert eine art drogenabhängigen dünnen hippie

2. chaos in negative verwirrt den checkpoint so sehr, daß teilweise anatomie gar nicht mehr als solche erkennbar war

3. doll-likeness in negativ führt dazu, daß noch mehr cringy freaks auftauchen

4. full-size oder full-height in positiv dienen eigentlich als alternative um closeups zu vermeiden, sind hier aber eher abschreckend und es werden sehr häufig gesichtsverletzungen generiert

5. inside-woods funktioniert nur mit extrem hohen qualitätseinstellungen, sprich "Ultra-HD-refined-quality-details", ansonsten sehen die mädels teilweise ziemlich absonderlich aus

6. man muss an manchen orten sehr häufig tricksen, weil sich die körper seltsam verformen.

hilfreich ist dabei in negativ "lack-of-different-joints", "lack-of-humanity" und "lack-of-precision".

7. outdoors allein geprompt führt häufig dazu, daß nur japaner auftauchen. ebenso bei allen prompts die ländlicher gegend sind. stadtaufnahmen sind perfekt fehlerfrei. wizigerweise verlieren bei "inside-city" die mädels brustfett.....

das war's erst einmal. teste gerade diverse "interior". "indoor-swimmingpool-interior" liefert richtig krasse ergebnisse und bei "classroom-interior" sollte man wählen zwischen schüler oder lehrer, sprich "1girl" oder "1woman".

gta v style?

(((GTA-5-ingame-style)))

only working with "ingame". without it, you get plain 3D

you can get an insae amount of styles for all kinds of games, but only modern 3D games are working. just replace GTA-5 with any game you want.

something small to point out:

if you want your images smoother and don't want to add tons of prompts, just try

(((Ultra-HD-rendering-details)))

Image posted by themancalledmojo999 (civitai.com)

it replaces a lot of details that photographers would add :D

this one is working best with this or other xl checkpoints. standard sd has insane problems generating correct images:

"hommage"

instead of style or theme will tell your checkpoint directly what you want. for first tests, i used the most simple theme "star wars". prompt it

(((star-wars-hommage)))

and you will get this Image post by themancalledmojo999 | Civitai

opulent odyssey? complicated algorhytm? long days of training?

just prompt

(((opulent-Ultra-HD-details)))

and you'll get the same in seconds and it'll save you space on your harddrive.

Image post by themancalledmojo999 | Civitai

EDIT: this prompt reminded me on a checkpoint that i found months ago that was made of this prompt. basically. won't tell you the name here. but one thing in general about that checkpoint: it generates an awful amount of details that, if you prompt those details in negative, generates crazy horror on purpose. mostly deformed nsfw horor....

since this checkpoint hates it if you prompt "doll-likeness" in negative and generates ugliness on purpose, i found an easier way to remove that doll-likeness. simply prompt in negative

"barbie"

"dynamink"???

just prompt

(((dynamic-ink-on-paper-style)))

Image posted by themancalledmojo999 (civitai.com)

EDIT: if you want to generate something more complex, put

(((photo-same)))

in negative, otherwise it won't work

Image posted by themancalledmojo999 (civitai.com)

do you guys know this meme coverpower.jpg (1200×675) (kym-cdn.com)

it was the first thing i had in mind after i saw this

Amazing Workflow To Generate Stunning Images! - v1.0 | Stable Diffusion Workflows | Civitai

that's why i keep my prompts short and simple and still get the same results

everything isometric? maybe game-style?

simply prompt

(((isometric-ingame-style)))

paper cut style?

(((multilayer-paper-cut-style, multilayer-paper-style, made-of-paper-style)))

1girl, 1man, 1woman is not working. instead use female or male.

diabetes? :D

(((3D-candyland-style)))

(uploaded images are stuck in an "analyzing" loop....)

reflections? mirror images? no checkpoint can handle this properly. but you can get proper results with this one here and using this prompt:

"mirrored-self"

it still is a hit and miss, but if it's a hit, it is a good one....

sigh .... i already had this topic more than 5 times already, but since this checkpoint can handle any style, once again:

"futuristic motorcycle generator concept"???

simply prompt

(((futuristic-tech-style)))

as a base and after that anything you want to see.

tarot cards?

(((medieval-renaissance-tarot-cards-style, flat-drawn-2D-style))) ""

this one is a bit special, but only for the use of prompts. you have to put it in quotation marks to get it to work. it's working without them too, but results are not consistent.

lol... "kodak ektachrome style"?

(((kodak-ektachrome-style)))

boah! dein checkpoint ist echt wegen geringen kleinigkeiten mega angepisst.

hatte "vector-style" testen wollen und bekam als antwort nur nackte ärsche like "leck mich am arsch"...

seems good aiming for better result generating realistic looking images , keep it please and share more of achievements

and please 'say more about the basic checkpoints of other people here' . i would like to hear from your experience , if you won;t say names don't but keep sharing your experience , achievements as you aim for better model as you seems to . from a real photographer sure it will be something good away from fake looking generations enough fake not real distorted images ..it's exist everywhere and the good result are a bit low chance..

about isometric:

you don't need special trained "sour" models. isometric is the basic of basics on any checkpoint.

just think of what you want to get generated and prompt it. a base could be

"isometric-style" for simple things or

"isometric-world-style" for more complex things or

"isometric-miniworld-style" for, well, isometric miniworlds (cuteness overload :D).

but anything else can also be prompted. just make sure to use "isometric".

best resolution is 1024x1024

EDIT: for super complex images with lots of detail, better prompt "isometric-miniature-world-style". that godzilla one took me more than 20 tries to get a consistent style.

a miniature world that's only based on jars without anything annoying from the outside getting mixed in?

(((miniature-worlds-captured-inside-jar-style)))

anything is possible. Image post by themancalledmojo999 | Civitai

was curious what would happen if i add "isometric"

(((isometric-miniature-worlds-captured-inside-jar-style)))

the only change is an isometric perspective. it's still inside a round jar

lol! prompting detail prompts in positive and

"babyface" in negative

generates super cute girls on purpose. Image post by themancalledmojo999 | Civitai

just a reminder for everyone who is struggling with this:

to remove nsfw and still get nice or sexy results on any realistic checkpoint, prompt

(((fashionable-fashionista)))

in positive. results are most of the time posing or seductive looking girls/women.

i was curious if sd snd sd xl know how to generate "agent provocateur lingerie". well, sd generated plain lingerie, but this checkpoint here could regenerate that style just fine. fishnet-like with many intricate details in different styles and shapes.

long story short:

"wearing-agent-provocateur-lingerie" Image posted by themancalledmojo999 (civitai.com)

i was curious what would happen on xl checkpoints when using "alienated" in negative. on standard sd it generated freaks on purpose, here, on xl, it gets even wilder, vishnu-like wilder.

so i highly recommend to not use that prompt :D

want a very unique isometric style with "captured miniworlds"?

put in positive

(((isometric-worlds-captured-under-sealed-glass-cube-style)))

and in negative anything that gets generated outside that cube using "-overloads:2". in my tests i prompted "new-york-city" and in negative "skyscraper-overloads:2" and got this Image post by themancalledmojo999 | Civitai

that negative prompt does NOT work with brackets. those are too weak for this style.

starting tomorrow, i will be back at my original place and not posting anything here fo a few days (only if i find something quick, new and useful). there are a lot of other prompts that i want to test for sd, too. my list has already grown way too long :(

getting pixelart on this checkpoint is nearly impossible, because it's trained on highest quality images, but it is not completely impossible.

1. you better work with numeric weights starting at 5 going up to 9

2. the same goes for negative prompt.

3. "2D-sprite-pixelart:5" in positive

4. "real:5, realistic:5, photo:5, reallife:5, photo-realistic-likeness:5, lifelike:5, 3D-ingame:5" in negative

result like this Image posted by themancalledmojo999 (civitai.com)

it is very inconsistent at this point, because i simply can't figure out what i'm missing. but it's an improvement to show you that anything is possible on this checkpoint.

something useful for this checkpoint:

if you get only japanese girls the whole time, simply prompt "outside" or "outdoors" or anything else where your character should be and you won't see any japanese anymore

a very big "OOF":

prompt

"9.11.-in-new-york" Image posted by Stapfschuh02 (civitai.com)

wanna see something funny that i mostly got only on this checkpoint? prompt

(((assimilated)))

prompt is similar to "adapted" or "adjusted". you get "commoners".

but, and now that funny part, you have to prompt

borg:2

in negative.... no preview image, because i don't want to ruin that fun part.

some more knowledge for you:

instead of weights, you can prompt "halfly". "halfly-isometric" for example generates a slightly isometric view on things.

not recommended on anything nsfw-based. it will result in "fully".

found out what to prompt on this checkpoint to get nearly anything working. the answer is so simple, it's pretty embarrassing. simply prompt

(((non-human)))

tried to get a "3D-print-style" working using "girl" and "outside". it only ended up in fashion styles. after i used said prompt, i got this Image posted by Stapfschuh02 (civitai.com)

at first, i totally couldn't get the following to work on any xl checkpoint, especially this one since it's based on ultra high quality photography. but after some quick thinking, i got it to work.

prompt in positive

(((authentic-daguerreotypie-style)))

and in negative

(((high-quality-colour-photography-style, colors-on-vintage)))

and you get this Image post by Stapfschuh02 | Civitai

EDIT: okay... 4 images got deleted. again. those mods really are pissing me off. those images were totally harmless.....

teaching this checkpoint stuff that it isn't trained on, can become a nightmare. but it's never impossible.

this checkpoint starts to listen to you if you prompt

action-to-be-performed-"pulling-using-fingers-on-2bra-straps-up-on-2shoulders"

this absolutely wild example was the result of an insane amount of trial-and-error to get this result Image posted by Stapfschuh02 (civitai.com)

it's not even close to what i was hoping for, but now i know that i can prompt commands and instructions the same way like i did before on standard sd.