Welcome to Inkbunny...
Allowed ratings
To view member-only content, create an account. ( Hide )
Cayuse

ARrrrrrrggh

by
I spend too much time messing with AI. :P

I have no process.

I have no magical ability.

I make fifty prompts, just to smash three together in photoshop, and then transform that image twelve different times to find one that works, tweak THAT in photoshop again, transform the entire image in im2img, inpaint some parts, decide I'd like it to be a different color, paint over it in photoshop, img2img it again, try inpainting the last bits, get frustrated, end up doing it myself in photoshop, realize I spent four hours on one image, upscale it, and post it.

@________@

oh and make sure to img2img it in at least three different models because I don't like how any of them work, some are good for poses, some are good for getting rid of shiny, some are good at genitalia inpainting...
Viewed: 99 times
Added: 3 months, 1 week ago
 
dodgerpup131
3 months, 1 week ago
I feel your pain, I do the same thing with photo shop, I’ll generate each character separately because regional prompting is useless in my opinion, then generate a separate background, then mask my two characters into Photoshop blend them into the background, but I end up having to make several character generations just to get the poses and interactions right first. It’s a headache that’s for sure. my job sucks the life out of me too so I don’t have much energy. So, I just really take my time.
Cayuse
3 months, 1 week ago
Yeah idk how people get exactly what they want on the first try, or so quickly! I feel like I'm bad at this whole thing, but in general I do tend to like the end result ... it just takes forever. And it is fun to bring stuff to life that I've had trouble finding porn of before
CycilCollie
3 months, 1 week ago
Someone 'getting' something on the first try usually means it's simple slop. It's rare for an AI image to look real good on the first pass unless it's incredibly generic.

Also to reply to another comment

>I'll be honest, I don't really get the difference between original / fill / latent noise / latent nothing, and whole picture / only masked. I just click around until something works. (:

Diffusion works by starting with *noise* and then gradually removing noise every step based on a lot of complicated math. As noise is removed, it begins to look like an image.

Original = the base image is your starting noise
Fill = Averaged block of color is your starting noise
Latent noise = similar to txt2img, it's pure noise
Latent nothing = blank 'nothing', similar to fill

Denoising is the level of noise you are adding to what you're inpainting (based on your selection).

Whole picture = generating the inpainting as if it was generating an image of your entire canvas. If you inpaint after upscale, this can cause duplication artifacts on higher denoising.
Only masked = generate the inpainting as if you were only using your masked area as the base. Can make fine detail but may also just redo your entire image if noise is high enough.

This is a good reference: https://civitai.com/articles/161/basic-inpainting-guide

Cayuse
3 months, 1 week ago
True, and it is EASY to spot that 'slop' from a mile away too. And thank you, I'll give this a read!!
RemWolfy
3 months, 1 week ago
Inpainting is my savior
Cayuse
3 months, 1 week ago
I'll be honest, I don't really get the difference between original / fill / latent noise / latent nothing, and whole picture / only masked. I just click around until something works. (:
RemWolfy
3 months, 1 week ago
Sent you a pm ^^
Inusan
3 months, 1 week ago
We all started out that way.  Eventually you find out what some of the things do and get a bit better.  It's never easy though, particularly is you want good credible quality.  The models have gotten AMAZINGLY better, but it's still feels like you are fighting the models 1/2 the time.

It's not at all as easy at the media makes it out to be.
Cayuse
3 months, 1 week ago
Thanks <3 What're your favorite current models?
Inusan
3 months, 1 week ago
I am partial to the Nova line of models, Particularly, Nova Kemono (https://civitai.com/models/1641408/nova-kemono-xl) and Nova Furry (https://civitai.com/models/503815?modelVersionId=2104813)

But I am ALWAYS trying new models. I spend way to much time on CivitAI.  I also have WAY to many on my drives. (something like 1.4 Tb of them)

Cayuse
3 months, 1 week ago
Nice! I do use Nova Furry often
WhiteRabbitfan
3 months, 1 week ago
I know how that feels.
Cayuse
3 months, 1 week ago
Yepppp
WhiteRabbitfan
3 months, 1 week ago
Indeed.
" Cayuse wrote:
Yepppp
VarraTheVap
3 months, 1 week ago
Youuu are doing incredibly well tho >w<
Cayuse
3 months, 1 week ago
Thanks Varra <3
Rimusk
3 months, 1 week ago
I know little about the process but thank you for your hours or patience to make some of the best looking images I've seen!
Cayuse
3 months, 1 week ago
Ahhh thank you ;..; I'm so glad!
JacobMuken
3 months, 1 week ago
I mean, consiter how long it takes to make actual art, on top of the (very roughly) decade it takes to get to a place where your at a very good level, one piece can take dozens of hours or more of sketching, resketching, lining, coloring, shading, blending, shading agian, detailing, highlighting, cleaning. Making something decent of any kind takes a lot of time and effort and experience. Four hours isn't bad
Cayuse
3 months, 1 week ago
Very true! Real artists are still the best
TuxGen
3 months, 1 week ago
The extra work you put in and the deliberate choices you make really do make your work stand out! Your willingness to change things or work at them until they're right doesn't go unappreciated and the results you arrive at are gorgeous and really inspiring! Hang in there!
Cayuse
3 months, 1 week ago
I will, thank you!
babycrinkles
3 months, 1 week ago
AI in my opinion is just a cheap cheat way out to say "I made something" Not to say I don't like the pics, but, I like actual drawn pics more. if you do whatever, keep it up, K?
Cayuse
3 months, 1 week ago
I agree real artists are the best and deserve support :)
CloppyDisk
3 months, 1 week ago
My process for the past two years has been to just make lots of wildcards, do tonnes of batches, and just get lucky.
Getting my actual mental images properly generated is harder than what most AI artwork haters realize. Or I'm just stupid... very probable.

I hope you keep it up, really like what I've seen.
Cayuse
3 months, 1 week ago
haha thank you!! I just figure I'm dumb too. I just mess around. Maybe if I had a better understanding of everything it'd be faster. but I also do a lot of manual photoshop and adjusting too
Kovu1508
3 months, 1 week ago
Seems like a totally normal art-creation process to me :)

no art without a frustation spiral ^^
Cayuse
3 months, 1 week ago
Lol good to know!
New Comment:
Move reply box to top
Log in or create an account to comment.