#🍥|anime
1 messages · Page 290 of 1
i like contrasting concept. Like tactical gear, but in pink color just doesn't make much sense, but it's cool 
it looks really good tho, even if it aint really tactical at that point lol
nurse with tactical holsters is a new one haha.
accidentally switched back to another sampler so it wasnt following my prompts too well ope
remaking them with the sampler i wanted
lol what sampler made those?
still looks dope haha
she made me wait so long i thought you were trolling lol
I know.. weird right? Just came out like that 😄
still tinkering with veo2?
yeah
I posted a pretty good dancing cat in #🏞|general-with-images
thats fuckin adorable
Comic-style wide shot of a dark alley, with the polar bear and the little man in the black hoodie locked in a tense face-off. The bear stands on two legs, wearing a red scarf and brown newsboy cap. The man is frozen in place as the bear’s massive shadow stretches over him, cast by stark moonlight. Mood: fatal finality. Use silhouetted framing with sharp light-dark contrast, emphasizing the bear’s dominance and the gravity of the moment. Let the moonlight carve out dramatic shapes in the alley, heightening the cinematic tension.
ok ill bite
Comic-style illustration with tight framing in the same shadowed alley. A close-up on the little man's nervous face — he wears a black hoodie, his face pale, mustache twitching slightly. His eyes are wide, fixed on the polar bear looming over him. The polar bear, standing on two legs in the background, wears a red scarf and brown newsboy cap. Mood: tense and still. Use high-contrast shadows and dramatic lighting to emphasize the man’s fear and the intensity of the moment.
@fickle oyster is that with the new checkpoint version?
yes!
Looks amazing!
downloading now XD i hadnt even got to try the last version yet, i was generating with Black Magic for the last day or so lol
hmmm my bear is just a shadow lmao
I appreciate it regardless! I'm glad it's been so good to you!
sending it through my nested wildcard prompt for a lil test
ill share whatever sfw stuff it makes here XD
illus v2 my fav
1
2
3 ah ah ah
THE COLOURS
evenin'
howdy
what's up?
tinkering with FramePack, since i just heard about it
Her hair reminds me of the chick from the Hercules cartoon movie lol i like it though
Megara?
yeah that was her name
hmm ok... lemme see
lol just googled it and the only part that i was close on was the shaping in the front, but for some reason thats what came to mind when i saw it
i imagine if she were drawn in a traditional anime style, that would be close. she had it pulled up in a ponytail with a red hairband
i guess the art style and body shape is pretty close too in all of those. i dunno if you even intended for that lol
closer...
I did a lot in this sketchy style for a while. I like it.
if this FramePack works out, and can generate 2 minute videos without taking all day, i may pick up a checkpoint that can do anime fight scenes. Cause i would love to see how it renders those.
is it running?
it is. its generated 70 frames and 2.3 seconds of the 10 second test i decided to do
4 seconds now. its not terribly fast, but that could be a hardware issue on my end. but it has a super low vram requirement to even gen, only needs 6gb
just saw this, thats one area i havent really tried. sketch or more hand drawn art styles.
what hardware?
i assume my gpu. since it needs only 6gb to use, and i have 10gb, i would guess that if it had say 24-48 to use it would generate quicker
No I mean what do you have?
oh, Nvidia RTX 3080
nice
Yeah its a workhorse. But its got nothin on the new series of GPUs lol
Never had any issues playing games, but its getting tested big time with AI
🫡
My 3070ti is still serving a purpose in my colleagues pc
Hmm i could've used it for ai as a second backend but feels overkill
I probably just need to build a dedicated rig for AI stuff so this pc can go back to being a gaming pc
But that sounds like work...and expensive
lol what ya got
fuckin nailed it
img2img?
yeah, but many iterations and tweaks 😛
lol looks pretty good!
happy eggs and bunnies everyone
What lora
it's 4:15am stop asking such tough questions D:
3 for me
I need to be up in 4 hours lmaoo
Uhh send it to my dms?
oh
I'm gonna lose it if it's here lmao
I'm not at my pc rn 😅
Cant download it
I'll have to find it later
npnp
ive returned....rested..and what in the good googly moogly is going on wit her eyes
Create a flying horse image
🐴
SCAm!
RELEASE THe HOUNDS
Noborumix doin work
oho, very nice
same character, same prompt, different checkpoint (RAMTHRUST'S PINK ALCHEMY). I love seeing how different checkpoints interpret prompts.
100%, one of the most dopamine inducing things about AI
Agree 100% lol its what causes me to seek new models to try out some of my favorite prompts with, or new LoRAs to try and tweak aspects. This gen was with Dream Matrix. Still learning what makes that model/checkpoint tick since the creator didnt give a suggested set of parameters lol
If you use an image to prompt model, do you modify its text output to make it more closely to what you want?
you would modify your prompt, and repeat the process. theres several ways to achieve this. 1.add or remove things from prompt/negative prompt. 2. add weights to things you WANt to show up such as (expressive eyes:1.5), for example. or 3. send the image to img2img and inpaint portions to add things manually
well, that's from a scene of the anime "edgerunners"
really good anime, even more so if you played cyberpunk
ive seen it and played 2077. fuckin loved both
yup you definitely need to. img2img is basically just them adding noise to your image, and then denoising that noise. So you will need to let the "ai" know what to denoise into, and also how much noise to put into your image first
thought you generated that lol didnt realize it was lucy at first
her hacking in that scene is crazy fast
probably not now, but with time, possibly 
yeah. i might try and recreate an image like it and see if FramePack can animate it
that little gif, have text, which a lot of models are really bad with, and not just that, it's mirrored, since it's to the perspective of lucy. Then also that hologram screen thingy
just realizzed i was using the wrong checkpoint for all of those gens....FACK
yeah i dont think i can get anywhere close to that clean, but it might be able to generate something similar. worth a shot at least
Ill have it work on it while i sleep cause framepack kinda slow
yeah my 3080 was sweatin, took like 20 min to get a 10 sec video lol
you at least got it working. I used the one click, so those xformers or sagattention wasn't installed. I thought to just do the sanity check first, and it took like up to an hour to do the sampling, then it was stuck on "extending video" or something with my ram and gpu usage at 100%. I waited for another 1 hour, and just ctrl + c the cmd
oh damn. thats crazy. i did the one click install as well and it took a hot minute to download all the damn models and shit, but it worked after that. i skipped the sanity test lol
Domain Expansion - Painted Labyrinth
scam, dull doll dont click on this one
lol
ive built an amazing prompt for use with the INCS lora (and a neat outline glow and silhouette one i found) that is spitting out some random beautiful gens. Thanks for putting me on this LoRA and the Noborumix model, this has been so much fun
would be cool if they just add a bot that detects masked link, and warn users about it
example: it's my test bot 
hmmmm i could probably make that lol. but yeah, with where AI is now, this sort of bot should be on every discord server. i mean hell, my phone company has one, or a software system at least, that warns about scam texts lol
oh wait i just clicked the image, you already made one lol
i mean, everyone can make that tbh. It's easy to do so with ai all the more so
it's just, staff here are really against adding bot
this would be so nice to have
or just a bot that deleted masked links automatically and just mutes people who use them
i mean if their mods are on top of it, i guess it isnt really necessary, but you got people of all ages in this discord, so it seems like at least this sort of thing would be a higher priority
well
kids just be clickin every link nowadays
still there lol
poor dull lost acces to his steam and email adress bc of that yeah
automod would be nice
my bot logic is to give it the benefit of the doubt, and let people do this
Search the world's information, including webpages, images, videos and more. Google has many special features to help you find exactly what you're looking for.
yeah. i got 40k unopened emails cause i aint trusting anything sent from an address i didnt sign up for lol
oh yeah thats fine but if it didnt go to google thats not
like, that should still be allowed, but will still be flagged and people would see that warning 
yeah, then it will probably need some type of link checking, instead of just checking for masked links
we've been asking for essential bot here for ages. Like example, a bot that detects metedata in image posted here, and react with a 🔎 so people can press it to auto get metadata here on discord
swarmUI idea i see hehe
that just works for PNG (discord kills JPG EXIF on upload)
that kinds of bot is like everywhere 
well, i think most people uses png
i don't, JPG 95% uses much less space and i generate too much 😉
meh i mean if someone wants my metadata all they gotta do is ask. if i post it here i probably posted it on civit or will at some point. anything i dont want people to have i dont post lol
same here... and locally it's perfect, all metadata in EXIF, forge image info works and all
exactly why i said it's most people 
yup, Zenia gens more then me so it ends up being 15gb in 12 days
lol im still trying to fill up the first 500gb of this 4tb ssd, and its been a little over two weeks now
well, it's just a way for easier access to metadata
and that was including me offloading a bunch of models to the ssd for me to copy back over for later use
(i just put her user settings on JPG lmao)
hmm png is my usual tho
don't start with local LLM until you want to burn though space even faster , 15-20gb per model easily 🙂
something where you could plug in any image and get the metadata (help verify if it was AI generated or not first off, but also potentially recreate aspects of it if you want)
like that? 😉
well yeah theres extensions inside the webui that can pull that info. but i meant for people on mobile using on site gens that dont offer it. like a standalone app they could use for free to do the same thing
or, even better, a mobile connection to your local webui
you get my point 
hm, would be surprised if that didn't exist already... and mobile connection, well, that's what VPNs are for 🙂
never touched a vpn, lol but there isnt, at least not as far as chatgpt o3 could find
yup, there's definitely already discord bot that already does that.
your machine runs in your internal net. you install a VPN on your router and access it from anywhere. done. that's what a VPN is for
id always heard vpn was for masking your router location to another geographical location, for reasons.
didnt know you could use it as a portal to access an at home device from another device lol
vpn is "i go to the target network directly", location masking is a side effect.
don't trust chatGPT, just google for "access local stable diffusion via vpn" 😉
well to be fair, the question i had asked chatgpt was to search the web for mobile applications that image gen folks could use or would use that dont exist and that chat gpt could reasonably build lol
e.g. https://www.reddit.com/r/StableDiffusion/comments/xilfd9/can_i_remotely_run_stable_diffusion_on_my/ - the topic literally is "Can I remotely run stable diffusion on my computer but access it from my phone?"
it also suggested an openpose repo app that allows people to create and share open pose skeletons in a mega database for anyone to use
mandatory random image to avoid text walls 🙂
same
looks like i need to get a vpn then...cause the one thing that kills my generating and sometimes costs me ideas is the damn 10 hour work shifts mon-fri lol
oooohh new plant milk checkpoint version is out....time to test
based on V0.1?
o based on V1
i think so? i just see aesthetically pleasing art, and i download lol
o like using the latest so it knows more / new techniques
they didnt say, their another one of those creators that doesnt give a lot of detail on the models other than the general art style and performance. i try to push them and see how creative or clean they can be with various prompts. and how consistent. Though the Noborumix checkpoint you shared has set a high bar lol Rams new Pink Alchemy does really well too.
Any krita AI users? Just wondering where I should put the inpaint model
never used krita sadly, im an ex-A1111 user turned Forge
oh he mentioned it, just less obvious
i would think you need to change your SD model under SD Common Options to your inpainting model when you are inpainting
I am still trying to have comfyui connected to krita but it won't let me until I download and put the models in the right places. So I am trying to get through them one by one
no idea how to connect comfyui to kirta, since i don't use comfyui. And i also didn't use krita much, only back when i was using it for some event
I'll just keep trying folders one by one until the error message disappears
is your comfyui working in the first place?
Yep
then it should connect to your comfyui with api calling, or at least that's what i think
you don't really need to move your models from comfyui to krita
Krita gives me options if it wants to make a seperate ai folder within krita but I chose the option to do it from comfyui so I don't have to make duplicate models
yup, then you just have to connect your comfyui with krita, possibly starting comfyui with api flag
not sure how comfyui connection with krita works tho, so don't quote me on that
Yep it says connection established but missing stuff. I'll go through the list of missing stuff and try to get em all
Plant Milk Vanilla
thats with no hires, hires breaks the image. but oddly enough, in a good way if your going for something that looks glitchy as hell. just needs tweaking to still be clean, maybe less loras if you use hires
Yay it works! 🥳 🎊 Now I just need to make enough money so I can do art all day and not go to work
So it doesnt do well with faces in full body shots at wide aspect ratios, but adetailer fixes that
I can say the backgrounds and other objects are pretty solid, not the BEST ive seen, but compared to some other models much better.
Definitely leans a lot into semi-realistic style, outfit details are quite nice (at least for armor pieces and bodysuits, all ive tested so far).
A powerful anime-style girl standing confidently in a dramatic, fiery and magical battlefield background, with dual glowing katanas — one blue and one red — crossed in her hands. She wears a stylish outfit combining a brown crop sweater and a green plaid skirt, adapted for a fantasy warrior look with dark accents and subtle armor details. Her face and eyes are intense and expressive, with one eye glowing bright blue. Her hairstyle is a fusion of a sleek bob with flowing strands, with a dark purple to black gradient. The art style is vibrant, detailed, and dramatic, with a focus on lighting and energy effects, featuring shards of glowing crystal and elemental auras surrounding her.
Plssss
i just use gemini for this, so it's kinda meh

some high class lady for you
converted it to booru tags and gave it a shot on SD
Theres no such thing as too much muscle
supposed to say "crush you", my japanese is really bad but the AI thinks so 🙂
Isn't that the newest plant Milk, vanilla?
The newest plant milk seems heavily skewed towards more childlike facial structure, could be a side effect of the training data or my shitty prompting lol
thanks! and for the two others above them?
Fair lol
just add "grandmother" 😉
What's y'all thoughts on the new illustrious v2?
I'm gonna do this and we are all gonna see what happens, once I'm off work lol
btw, anyone tried HiDream yet? rather new, i checked the web demo some days ago, looked like flux-level (not a totally new quality but really good)
I love it. Hardly use adetailer or hires with it.
its pretty good, not the biggest fan though
it does some whack stuff sometimes with the same prompt
overtraining 🙂
masterpiece, best quality, newest, highres, absurdres, watercolor, traditional media, painting (medium), armored female character, pale complexion, white hair styled in a french bob, multicolor robe with detailed accents, gorgeous, beautiful 21 year old woman, ear piercings, choker, cleavage, flowers in hair, pop art, pop surrealism, conceptart, simple background, flat image, abstract,traditional media, scan,blending, cute, pop art
sorta followed it
ignores like half my prompt
tho stealing flux prompts do it pretty well
A close-up of a face adorned with intricate black and blue patterns. The left side of the face is predominantly yellow, with symbols and doodles, while the right side is dark, featuring mechanical elements. The eye on the left is a striking shade of yellow, contrasting sharply with the surrounding patterns. The face is partially covered by a hooded garment, cinematic_1940s cinematic_octane
just for fun with flux pro 1.1, not trained with booru at all, but very nice prompt following 🙂
i wish i could run flux pro locally ong
Hidream seems to love oversized heads hto
i use replicate for that - the pro model isn't freely available. but the results are spot-on most of the time so costs are reasonable, no need for many tries
flux pro 1.1, photorealistic of course
the mechanical elements are really cool
same prompt but the V2 illustrious noborumix
in the style of ck-mgs, nistyle, Special Ink-drawing mode, intricate linework with expressive contrasts, Inkplash art on rice paper, sepia, henna , Silhouette Art, magnificent, inksplash closeup portrait of Chinese woman in a rowing boat, sunset, calm lake, reflection
adding "flat 2d cel shaded anime style"
same prompt but on illus v2
genning at 1536x1536 gives better results but not what i prompted for
the model knows what's best for you 😉
like with most AI lmao
Just let it happen. You'll discover something new you didn't know you would like lol or have nightmares.
That's amazing, tysm
it is....
That's a dual answer
it is..... comfyui easyuse node, image remove bg
Ahh nice
that is impressive
probably gonna make removing layers or adding a transparent layer to an image for POD easier
obligatory "im home" picture
that ones a lil strange
i must say im extremely pleased with my wildcard prompts. i set em to a batch of 30 and go to work, come home and see some hidden gems i didnt expect
thats awesome!
this was cool
@dusky linden been playing around with your lora from civitai. I like the results so far.
oo neat i thought i recognised the lora lol
it gets closer to the ace glitch style if you use base illustrious\
It didn't do too bad with amanatsu xl. And that image had its weight turned down a touch cause I added a couple other loras in with it to see what would happen. I think that one was with either ramthrusts pink/black alchemy/magic as the checkpoint.
I also haven't turned the weight up to max with just the ace glitch lora on to see what it would do. But I like the aesthetic. And it pairs exceptionally well with my "glitched/corrupted deity" wildcard prompt I made.
I will say i had to increase me sampling steps to address some small artifacts issues, but once I did that it looked dope. I'll share some of my outputs on Civitai when I get home, assuming they've unfucked the site by then.
I'll have to try it at different weights then. I think at 0.8 it had elongated the necks a bit and Sharpened the facial shape and features. The above image was 0.5 or 0.6 but had other loras. I'm enjoying fiddling with the settings and testing the outputs and influences on different checkpoints. Also, there's a WAVE of updates coming to a lot of checkpoints with the Illustrious 2.0 model, so I'm stoked about that.
Hmm if mixing with other style lorasi found 0.3 to 0.1 works best
Oooooh so colorful
did a low effort text with flux fill
lol looks good. i actually like how clean the linework is til it gets to the eyes, then theres like a soft blur to them thats unsettling, but not in a bad way lol
You got flux fill to work locally? Gotta show me sometime
Never got deep into flux
i use draw things i just download it off the app and it downloads its clip and vae and whatever else then i use its flux fill preset
they have 5bit and fp8 in the app
Ooh, somehow i keep forgetting you use apple
Great software & hardware but i cant help but hate it
same
but then again
ive always been anti apple, not because i dont like apple, but because all of the people ive known that swore by it were dicks lol
I mean for a user, its really good
It just works™️
For me tho? I want to change and edit everything
Break the system
But apple says no
am i the first XD for portable stuff i like apple but desktp pc better
lol time will tell. But yeah after i built my first PC ive sworn by it. Apple sells convenience, but if you want power like Euro said, gotta build your own. Cause Apple dont like you changing their stuff lol
i built my own but it was for gaming so i went amd gpu
now im just prep for 5070 ti
so i can use my macbook to take around again for once
lol i didnt know shit about GPUs so i went Nvidia cause it was "all the rage" right before covid. So i lucked into a decent GPU for this
used 3090 is like 150 less than 5070 ti so ill grab 5070 ti
yeah, i heard somewhere the 5000 series are buttcheeks for gaming, but amazing for AI stuff. So im just gonna build a dedicated PC for both and call it a day
i dont have money for that but anything will be better than my 6600xt
well im 10 years into my career, if i aint making enough money for this i fucked up XD
have you tried the AMD Zluda Guides? a 6600xt is still okayish to use with that
i have it was a while ago that i had a forge issue that couldnt be fixed so i gave up
If the ace glitch lora causes too much drooling (and sharp teeth) i recommend adding it to negatives since the artist draws too much with that
Dataset is like 30% drooling
Unironically those are the things i want XD for the corrupted deity stuff the expressions section was all like...unhinged, mad, or devastated and crying expressions.
Oh yeah then go for it lol
oh okay, let me know if you want to try it again, or if you have info about what cant be fixed
it pairs nicely with the Urban Waifu as well, since the training set for that was apparently a lot of vibrant, toothy, crazy images. Finding the right balance and the right prompt is the key now. But im gonna run some with just Ace Glitch and see what i get at different weights
after my val game ill do it
alright, let me know in #🤝|tech-support then, maybe you can provide a full cmd log to start tracking what might be wrong
I lost my dataset of exactly what i used but i can link you my source
(reinstalled my pc(
anybody know a good pony model for anime image generation?
Would recommend Illustrious based models instead. They are IMO easier to use.
For example Nova Orange XL
Nova Anime XL
wainsfw
For pony:
Prefect pony
really only testing out Pony because of the abundance of LORAs but will definitely check out the Nova IL ones as well
Is the waiNSFW model vpred?
cant run that request for you in here doll
lol i promise you the prompt was a lot more complex than that
probably unnecessarily so in hindsight
masterpiece,amazing quality,very awa,very aesthetic,best quality,newest,absurdres,highres,masterpiece, very aesthetic, absurdres, newest, flamboyant rival character, solo, Mystic's robes with symbols, Armored battle skirt, Soft leather shoes, Simple cloak pinned with brooch, rainbow vapor haze, from above, Red Angry Scar ,<lora:Urban_Fantasy_Waifu:0.6>,<lora:GlowingNightmare_IXL:0.3>,<lora:AceGlitch:0.6>,<lora:TrendCraft_The_Peoples_Style_Detailer-v1.0A-3_13_2025-illustrious:0.7>,<lora:incs:0.6>
I'm sure you can. I spent all day iterating through different combinations of lora weights and testing different checkpoints til I found the look I was going for. But I use a nested wildcard prompt to randomize certain elements. I adjust it for a specific theme as needed when I want something fresh lol.
Those are some nice boots
@toxic lotus do you play limbus company the game on steam too?
or just read the manga
I like the art style so I was thinking of playing the game
shes taking a discord break but she plays the game. (not on steam)
its the same game tho right
thanks for letting me know @dusky linden ;)
on steam its all in chinese
so I hope when downloading its at least english
nevermind, its korean
lmao no worries
woohoo, no more posting stuff on civitai... their new policies suck a bit too much for my taste...
Yeah right when i wanted to make more loras too
I might just focus on more local / closed circle
OnlyLoras
Meh, while im not saying ill miss some of the content that was restricted by these policy changes, the system changes seem a tad overreachy. Forcing people to share metadata is tantamount to removing privacy for creative properties (or however you would say that, basically forcing people to share the portion of the generation that was their own creation, since the model made the image. like stealing art isnt already an issue lol) and the new filter system is just ludicrously bad.
Could always lie in the metadata, unknown samplers, mix up the prompt, leave out resources
What, they can't force you to share your local tools
Sounds mean but I'm actively not incorperating in my metadata unless its for a lora
I'll even edit it out some style loras from the metadata, what will they do? Sue me?
Sorry maxfield, nothing personal lol
This is true, and it would cost them more manpower to verify any of that lol which flies in the face of the automated parts of the system. I dunno, a lot of it seems hand wringy and silly, and the stuff about children and incest doesnt bother me since i dont care to generate/view that stuff anyway. Or the celebrity shit, which is just goofy anyway.
Tbh celeb and "minors" were already weird to be allowed so yeah good
Incest? Like how would you scan for that
I sure hope nobody prompts with "brother and sister"
The biggest tragedy in any of this is that its gonna push away some of the creators whose tools i enjoy or might have enjoyed using from posting new models or loras because the site policy is driving away users.
And yeah i assume a lot of that is intended to be found out in the supplied metadata, hence that rule being added. But then, like you said, anyone can just lie since you can manually input or change the metadata on upload. So its pointless.
And tbh there's a reason why most people don't fully share their workflow
Im not worried about losing any generations since i operate locally, all my stuff is backed up when its created on my SSD for outputs lol
personally never understood this, but then its their right and their choice so it never bothered me.
If i spent hours making my workflow, tweaking it to where i like it. I don't wanna share that all over the place
I'll help others learn sure

But my own "creative input" is what i like to keep private
Sounds ironic since we're using open source software lol but still
thats fair. i guess in my mind the only reason i would hide the workflow or gen tools would be if i were monetizing the outputs in some way. Then i wouldnt want someone to copy my work. But since this is a hobby for me, not a source of income, im a little less concerned. But i totally understand your point of view
i will be the counterpoint here, and say that if people thinks like that, then they won't be sharing their model open source, and open source would be dead, but i get you guys as well
Oh yeah i totally get it
But i feel like personally theres a difference between handing someone the tools
And "heres what i did so you can copy my hardwork without the knowledge"
Yeah i think its strange to wanna just copy someones work 100% anyway, when the models are capable of so much more if you tweak and change things in different ways
Exactly, and if someone asks me personally how i did this etc ill do it

But not publicly where my uniqueness is gone
well yeah, but you can say the same for model as well? isn't it? Like if i "made" this model, i won't want to disclose that, that's my unique stuff?
or at least by that logic
I should make custom nodes for my self so this isn't a problem anymore
I do like being able to look at prompts for certain cool effects people get. Like the INCS Lora you shared. Id have never found that naturally, and i had been trying to get similar effects in other images and failing miserably. From that, i was able to learn a bit better how to prompt for it without needing the LoRA all the time lol
Well yeah, thats your right and while i could use your model
i mean, i do get what all of yall mean tho, i'm just trying to be a counterpoint, just for thoughts
well, a lot of people share that, and that's how the open source moved forward
correct me if im wrong tho
Yes but eh boils down to personal preference
yup yup, i mean yeah, it's all personal opinion about this
Id not be surprised if most of the people sharing models on CivitAI dont have others they havent shared that they curated or trained further for personal use tbh
In a circle of experienced users sure ill share
Bunch of new randos? Sorry the struggle is part of the experience
well, to me, i don't really care about workflow and such, since i'm using open source tools/stuff, i will share whatever, to give back to the open source community, but that's just me 
And its not like i dont share all my prompts/workflows, in my loras i share them
yup, i also get you guys opinion about this tho 
Simple ones yes but shared
This is more or less where im at. Like i said, some of this i dont disagree with. Some of it seems unnecessary. And some seems difficult to enforce properly without being overly stronghanded.
And if github was easier with accepting non creditcard payment methods id donate to a few groups by now
Largest advocate for opensource, biggest hater on spoonfeeding too
Now i gotta research that sd licence for. "Commercial" use since idk if education also falls under it
i do believe commercial use by definition requires you to be making money from the output
i think with civitai, they are just doing it to protect themselves from lawsuit by others tho
so unless your charging for the education on how to use the tool, your fine to teach at your discretion
oh absolutely. this is the reason for 99% of the changes to the internet these days. its all about covering their own asses. Which is fine, i get it.

Eh public education so should be fine then
Still, should research it a little
And with civitai, its ultimately their site so they can do whatever they want with it 
yeah id def familiarize myself with all the fair use vs commercial use stuff with anything AI. especially if you may potentially make money from it down the line
Its something id been wondering actually, whether when you create a model or lora you as the creator dictate the commercial use and on site use rules, or if they are arbitrarily assigned depending on the base models rules.
Ive seen some people outright forbid copying the model to sites like SeaArt, which i get. Those sites have monetized systems, but they dont create any of the models or content on them (except their SeaArt suite which was kinda mid tbh)
My gens and posts are separate so thats no problem
My school won't earn cash from it most likely
i only play the game
@autumn roost sorry you have to see this
yeah he's been everywhere 
me, you, and another have pinged max.
i pinged max when it started in #📝|prompting-help
well
fuck
i guess
i should
type like this
to
get that
off my screen
maybe a gif
will help
nice it's out of my screen
it's cleared 
ew
likewise! i very much enjoyed that little dump you did, got me wanting to gen some more stuff haha
Mind if I send you a DM?
go for it
Glad you like it!
ohhh i really like the pencil shading one. thats pretty
ah the colored pencil contrast is nice
yeah it does pretty well
are these all Dall-E?
really really well done 
it's fun to go back and look at old images and see how rough they are.
i'm just the prompter 😄
you should post that in the other server if you know what i mean 
I... do not. LOL
yeah i went back and checked some of my first gens....and it was uh...rough is a good word for it
Some are kinda neat... this is over 2 years ago.
Ok... I know the place... but not sure what you mean about posting there. 😄
I no longer have my first gens
They were uh special
You just think they'll lke that one?
i guess so, not sure to be honest
Ok, well I guess we'll see. 😄
More randoms from a few years ago
we've come a long way
no kidding
though i started like...a month ago with SD, and these are better than some of my first gens lol
I wouldn't say these were my first gens...
ill find some early ones...
that's about as early as I have...
oh the arnold one is nightmare fuel haha
oh.. no... I lied... found older
the tiny kitten with the big head is adorable though lmao
weird and rough
need to find the first one id classify as really anime... lemme see
but first, bruce willis eating pizza.
even that was 2023... come a loooong ways
Ok here we go... this is probably the oldest "anime" style image I have saved. It's... bad. 😄
After that... funny enough... probably this "Tifa". My first Tifa attempts... I thought they were so cool.
lol you've come a long way indeed
We've come a long ways...
I mean... come on. This was just about a year ago. What a difference.
just hidream
my laptop has a virus :((
it shows that I have to send BTC to a weird 18d812ghe120ed9jqdwmn81 number
why
ohh its a ticket
thank you!!
I can't login to my device until I send the BTC and there is a timer :(
I will include it in the ticket
you've been clicking weird stuff again, isn't it?
yes, but @split finch helped me with the ticket discord server
well, you need to learn to stop clicking weird links
😭
i don't even know if whatever that person send you as a discord ticket thing is even legit or not
who is the mod here? 
??? Fair warning for everyone. Do not click on invites of anyone claiming to help
If they want to, they can do it here
i seriously am confused as hell what is going on. @woven gust just came here to say that there's some virus, then someone just reply to dull doll saying to reach out there at another discord server? 
like this 
holy, actually are you including your login details in your ticket?
Hes getting scammed

Again
there is just no fucking way
I keep telling him, do NOT trust people online
i am almost certain
the people are on that server, telling him not to listen to us, or just do as they say
buddy please
for the love of god
stop clicking on random stuff
i screen shot your pic and show it to gemini for advice 
It's a fake warning: This is not a legitimate message from the FBI or any other law enforcement agency. Their official communications would not appear as a sudden lock screen demanding payment.
The purpose is to scare and extort money: The message attempts to frighten the user into believing they have committed serious crimes and that their computer has been legally locked. The goal is to panic the user into paying a "fine" or providing personal information to regain access to their computer.
The alleged violations are serious and designed to cause fear: The message lists illegal activities like accessing cp and sending t-related spam. This is intended to maximize the user's fear and increase the likelihood of them complying with the demands.
It mimics the appearance of law enforcement: The use of the FBI seal and official-sounding language is a tactic to make the threat seem real and authoritative.
The Bitcoin address: The text "bc1qumhetmcr5cuqlkf5garusc3wwwheg43qx8sghh" is a Bitcoin wallet address. This indicates that the attackers likely want to receive payment in cryptocurrency, as it can be harder to trace.
"This is the problem I have": The user "panni" is indicating that their computer is displaying this lock screen, meaning their system has likely been infected with malware.
What should "panni" do?
Do NOT pay the ransom: Paying will likely not restore access to the computer and may even lead to further demands or expose them to more threats.
Do NOT provide any personal information: The attackers will use this information for malicious purposes.
Try to restart the computer: Sometimes, a simple restart can bypass the lock screen, although this may not always work with more sophisticated ransomware.
Boot into Safe Mode: Try to boot the computer in Safe Mode with Networking. This can sometimes allow you to run antivirus software or access tools to remove the malware.
Use antivirus and anti-malware software: If you can access the system in Safe Mode or another way, run a full scan with reputable antivirus and anti-malware programs. Make sure these programs are up to date.
Seek professional help: If the user is unable to remove the lock screen or malware themselves, they should seek assistance from a trusted computer repair service or IT professional. They will have the tools and expertise to safely remove the infection and potentially recover any lost data.
Consider data recovery: In some cases, ransomware can encrypt files. If this has happened, data recovery might be possible, but it's often a complex process and not always guaranteed. Professional help is usually required.
In summary, this is a common type of online scam designed to extort money by frightening users with false accusations of illegal activity. The user should take immediate steps to remove the malware without paying any ransom.```
read it if you want
how to remove it?
how to remove the malware
then it's ransomware
tbh I'm too tired to deal with this right now. It's dragging me a bit down that I almost fell for a scam again
you are talking to be thru other ways right?
yes
like
we told you so many times not to anyhow click on links
but
well, not the time to talk about this
first
already shut it down
do you know how to start your pc in safe mode?
what pc do you have
maybe i can find what to press when starting up pc, to get into safe mode
nah I dont want to bother with it
it's thinkpad
but don't worry
I can just buy a new one its ok
lenovo thinkpad?
yes
start your computer, but repeatedly press F8
^ what smol is saying or worst case scenario, new windows install
Dull its easy and simple
A new laptop is thousands of dollars
This could fix it free
No digital wallet or links but you do you
just, if you want to do it or not, just remember to not click any links under the sun. Learn to doubt people more, especially people online. If you really want to click the link, maybe use some online tools, like https://nordvpn.com/link-checker/
or whatever, it's better than just clicking link blindly
okay, thank you. I will try to keep it in mind for the future
please do
catching up with the whole thing, gimme a sec
this is the user_id of the person saying to go to the other server 1256804986310889546
Gave the boot to to two of them (including the guy with basically zero message and trying to clear the name of their accomplice)
To be clear once again.
The only "trust me bro for sure" people in here are
- Stability staff / admin , check the role on their profile
- Community guide , which is a role granted by staff themselves to a few others. So they should be trustworthy.
For all the rest, be careful out there. SD discord is quite big so it attracts a lot of scammer / bots. Especially because this a tech related server and lots of people come in here asking for advices.
@woven gust this does look like your typical ransomware scam. There are ways to identificate which one it is exactly and maybe even to recover the data. Worst case, a reformat of the drive should be enough.
thank you for the heads up. I will try to not get tricked again
Cant help you more with your situation at the moment cause I m at work. I m sorry, be careful out there folks.
you don't have to. It's ok. I get it fixed at some point... I bring it to a shop on Monday. Worry not
That s one way to do it.
It should save you a few headaches for just a couple of bucks.
I'm a regular there. I get discounts
I see :p
my first ones (circa november 2022), SD came a long way since 1.5 / waifu diffusion days :p
haha yes, those fit right in about that time!
I’m getting better with ideas (Ofc other prompts help me place this with that and then add my own thing etc)
hidream dev running locally, with "anime style studio ghibli, a typical manga cover" - not half bad 🙂
the same beast, same prompt, same seed, just in different styles...
@toxic lotus I didn't know that Mili made songs for Limbus Company
I started the game and listened to this banger https://youtu.be/XfTWgMgknpY
🎵 LISTEN BY STREAMING: https://linkco.re/FTUMav06
WE ARE LOOKING FOR AGENTS/REPRESENTATION OUTSIDE OF ASIA.
Please contact us at http://projectmili.com/aboutcontact .
For the upcoming game "Limbus Company" https://twitter.com/liberarelimbus
Follow Mili
Spotify: https://open.spotify.com/artist/0K05TDnN7xPwIHDOwD2YYs
Apple Music: https://itu...
Mili is one of my favorite creators
actually was my first introduction to her, and the vocals in that are amazing.
ok, have both...
Thanks
this got....wild...plantmilk vanilla has some uh...well its something special
hmm
plantmilk vanilla... just random words?
new illustrious model drop
was a topic a while bak
but didnt you run mostly gpt/cloud stuff?
me?
yeah
dall3 is my fave but i run local also 🙂
ahh im not a fan of cloud only solutions ngl
they have their own pains-in-the-butts
theres a creator on Civitai that has made a series or suite of models they call "Plant Milk" that are each named after weird milks, like vanilla, oat, hemp etc. Their very creatively trained checkpoints, but man they uh...they get too creative sometimes lol
yes I see that haha
actually ive been using invoke a bit recently
the interface is more pleasant to use than swarm... but I miss some of the ease of use in swarm... invoke is a little.. odd.
been enjoying this model lately.
Don't click on these, dull doll 💀
the scams are out of control
soulmates?
So I'm seeing something about civit closing or something? Gonna have to see what's up.
oh snap
mm nevermind... they're cracking down with some rules but not closing
not sure on all the details
interesting style mix
indeed
(with hidream, really really good with text now)
doing all right. kinda tired
"anime style flat 2d cel shaded, a cute anime girl waving with her right hand, left hand behind her back, short black hair, red short folds skirt, black high boots, white hello kitty logo t-shirt, oversized pink jacket, standing in a neon subway car, happy eyes shut"
plus "expressionist oil painting broad brushstrokes minimalist cubist" - hm, anime dominates the styles
hehehe it was an interesting combination of loras that got that effect
and photoreal as the last version
with more modern models, i almost never see the need for a lora anymore
flux?
i dont know if any modern models will hit some of the styles ive blended without copying the image lol especially my wildcard images. those prompts get nutty
hidream in this case (a bit better subject separations and better text) but flux would work as well
but of course, i couldnt do half of what i do without the more talented people creating checkpoints and loras, so i cant really take any credit lol
well, without our prompting the models and loras would just be dead bytes on your disk 🙂
"girl, waving hand...."
I thought hidream was an animation thing....
thats TRUE!
guess I'm out of it again
yoooo who did she steal that from lol
fighting dragon? sitting on dragon? what is happening here? 😄
petting the anjanath, though apparently with the outfit there was some style bleed. its a girl wearing normal anjanath armor petting a fulgur anjanath lol
she was SUPPOSED to be sitting on her knees, but the model had other ideas
flux
only 20 steps though
..because waiting on flux makes me nutty
yeah, flux has been my go-to for the last months, just playing around with hidream.
i recommend flux fusion and just 7 steps, good quality and really fast
yeah if it takes me more than 20 or so seconds i get annoyed lol
probably why i havent been dabbling in framepack animations, takes waaaayyy too long
that's the downside of hidream, 34s for such an image... 7 step flux fusion is 5s
haven't tried that one. Don't those "lightning" models have a tough time with prompt adherence, though?
still much better than SD1 or SD2 🙂
totally her own hand, i swear
"anime style, cute kitsune warrior female kneeling and petting a black dragon, dragon blue fire breath, snow forest background"
hmmm... blue fire vs blue dragon, gotta catch them all i guess
now we got it
aint no way lol
is this fusion? or still hidream?
and just like that...he was gone
I mean i can do that, but id need to generate a bunch first lol
and i need coffee before i do that
also, thats a girl in nergigante armor with arkveld on her lap. again though, style bleed between the two monsters so its not perfect
I'm not sure that armor is all that effective.
lol it is not, but then anime has never cared for effectiveness in female armor. not hystorically at least.
Look out for detective pikachu!
I dont do steampunk as much as i should. That came from the wildcard which is probably why the background doesnt fit.
Steampunk can be fun. Here's an old random one.
yeah i need to do more steampunk and baroque pirate. Those are two styles i dont dabble in much.
give it a whirl 😄
I miss the old 1.5 models sometimes...
I like the whale
yeah that and the fire fox spirit wearing a mask are two of my favorites lol
the old kenshi and xpero models were just fun..
Flux just doesn't give me the same interesting things... this just feel meh.
Yeah the only time ive used flux was with genspark, and it was alright but not near as creative as id have hoped. But i hear its really strong for realism and semi realism, and it has good prompt adherence.
yeah, I'm posting some flux stuff in general
Though with genspark i couldnt set parameters, only prompt
yeah those backgrounds are killer
nice owl 😄
@keen hollow there ya go, fresh off the press lol
thanks lol
actually surprisingly good. has a very western comic book color and shading style
Daily dose of Tifa
cookin a batch of aerith too for good measure before i get back into more weird and abstract...though i might do some more monster hunter stuff
I thought about doing the bathing suit ones, but figured that might veer off into nsfw territory lol
Welcome back! I did a thing, Omnia made me do a image dump
do you need more? 🙂
i see no problem here


forge?

