re: forever ai mistakes
i think pippa's just forever fucked to be like this, that body is fucked
gonna test a few more things, but i'll probably call it a day because it's a pain learning how ai works, because it takes forever to see the results if i'm experimenting with anything, from different tags, models, resolutions, samplers, steps, and so on
-
1girl, pippa, virtual youtuber, pink hair, hair ornament, braids, rabbit ears, pink eyes, unusual pupils, crazy eyes, (flat chest), ((naked)), (sex, cowgirl, squatting, creampie), pov, 1boy, penis, pussy, <lora:Pippa:0.66> <lora:PSCowgirl:1> <lora:add_detail:-0.3>
Negative prompt: ((blurry, monochrome, worst quality:1.4, low quality:1.4, low detail, bad anatomy)), bad res, error, malformed, cropped, pubic hair, censored, big lips, text, anal, futanari, extra limbs, extra legs, extra arms, extra feet, extra fingers, missing limbs, missing arms, missing legs, missing feet,
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 9, Seed: 3719497443, Size: 768x512, Model hash: 2fcdee6e9c, Model: based66_v30, Version: v1.2.1
gonna test a few more things, but i'll probably call it a day because it's a pain learning how ai works, because it takes forever to see the results if i'm experimenting with anything, from different tags, models, resolutions, samplers, steps, and so on
-
1girl, pippa, virtual youtuber, pink hair, hair ornament, braids, rabbit ears, pink eyes, unusual pupils, crazy eyes, (flat chest), ((naked)), (sex, cowgirl, squatting, creampie), pov, 1boy, penis, pussy, <lora:Pippa:0.66> <lora:PSCowgirl:1> <lora:add_detail:-0.3>
Negative prompt: ((blurry, monochrome, worst quality:1.4, low quality:1.4, low detail, bad anatomy)), bad res, error, malformed, cropped, pubic hair, censored, big lips, text, anal, futanari, extra limbs, extra legs, extra arms, extra feet, extra fingers, missing limbs, missing arms, missing legs, missing feet,
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 9, Seed: 3719497443, Size: 768x512, Model hash: 2fcdee6e9c, Model: based66_v30, Version: v1.2.1
re: forever ai mistakes
this is more sad than it is weird or disturbing, wacky and funny, or interesting.
maybe another day.
maybe another day.
re: forever ai mistakes
this ai can generate any sfw images of kiki, but as soon as i try to fuck her, it becomes a sushi platter
no sex, only sushi
no sex, only sushi
re: forever ai mistakes
these are vine flowers
-
1girl, Kiki, four-leaf clover hairclip, ears, brown hair, long hair, naked, (((sex:3))), <lora:Kiki-DICE:1>
Negative prompt: ((blurry, monochrome, worst quality:1.4, low quality:1.4, low detail, bad anatomy)), bad res, error, malformed, cropped, pubic hair, censored, big lips, text, anal, futanari, extra limbs, extra legs, extra arms, extra feet, extra fingers, missing limbs, missing arms, missing legs, missing feet,
Steps: 20, Sampler: UniPC, CFG scale: 8, Seed: 1033962643, Size: 512x768, Model hash: 7eb674963a, Model: hassakuHentaiModel_v13, Version: v1.2.1
-
1girl, Kiki, four-leaf clover hairclip, ears, brown hair, long hair, naked, (((sex:3))), <lora:Kiki-DICE:1>
Negative prompt: ((blurry, monochrome, worst quality:1.4, low quality:1.4, low detail, bad anatomy)), bad res, error, malformed, cropped, pubic hair, censored, big lips, text, anal, futanari, extra limbs, extra legs, extra arms, extra feet, extra fingers, missing limbs, missing arms, missing legs, missing feet,
Steps: 20, Sampler: UniPC, CFG scale: 8, Seed: 1033962643, Size: 512x768, Model hash: 7eb674963a, Model: hassakuHentaiModel_v13, Version: v1.2.1
re: forever ai mistakes
i didn't prompt cat girl kiki, but the ai gave me it anyway
which is pretty cute
which is pretty cute
re: forever ai mistakes
>kiki's model is so powerful, it turns the checkpoints that don't produce good results previously into good shit
or maybe the other models were just plain shit, what the fuck, why is this model so good
i feel like i've been playing with stone tools and finally got a CAD machine
it has its limitations in that it's the only lora that can be active, with others not being as effective, and i can't really figure out a way to get what i want with just prompting alone instead of relying a bit with a lora for positioning and details
but i feel like i've gotten closer to what i wanted
or maybe the other models were just plain shit, what the fuck, why is this model so good
i feel like i've been playing with stone tools and finally got a CAD machine
it has its limitations in that it's the only lora that can be active, with others not being as effective, and i can't really figure out a way to get what i want with just prompting alone instead of relying a bit with a lora for positioning and details
but i feel like i've gotten closer to what i wanted
re: forever ai mistakes
used img2img on a year-old drawing and experimented with models and clip steps
took like, more than an hour to generate all these images, but i've figured something out
-
loli, small breasts, young, sky blue hair, gradient hair, bunny girl, bunny ears, bunny tail, magenta eyes, unusual pupils, pink shirt, off shoulder, long hair, fluffy hair, sad, full body,
cherry blossom trees, flowers,
Negative prompt: blurry, worst quality, low quality, mutated hands and fingers, bad anatomy, wrong anatomy, ugly, ((mutation))
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 7, Seed: 1068999326, Size: 768x768, Model hash: a1535d0a42, Model: AnythingV5Ink_ink, Denoising strength: 0.75, Version: v1.2.1, Script: X/Y/Z plot, X Type: Clip skip, X Values: 1-5, Y Type: Checkpoint name, Y Values: "AnythingV5Ink_ink.safetensors [a1535d0a42],based66_v30.safetensors [2fcdee6e9c],BloodorangemixHardcore_bloodorangemix.safetensors [55cfe9bdf0],calicomix_v11Pruned.safetensors [20d3eca4c2],CounterfeitV30_v30.safetensors [cbfba64e66],feiwumix_v10Slightflat.safetensors [6505688a22],hassakuHentaiModel_v13.safetensors [7eb674963a]"
took like, more than an hour to generate all these images, but i've figured something out
-
loli, small breasts, young, sky blue hair, gradient hair, bunny girl, bunny ears, bunny tail, magenta eyes, unusual pupils, pink shirt, off shoulder, long hair, fluffy hair, sad, full body,
cherry blossom trees, flowers,
Negative prompt: blurry, worst quality, low quality, mutated hands and fingers, bad anatomy, wrong anatomy, ugly, ((mutation))
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 7, Seed: 1068999326, Size: 768x768, Model hash: a1535d0a42, Model: AnythingV5Ink_ink, Denoising strength: 0.75, Version: v1.2.1, Script: X/Y/Z plot, X Type: Clip skip, X Values: 1-5, Y Type: Checkpoint name, Y Values: "AnythingV5Ink_ink.safetensors [a1535d0a42],based66_v30.safetensors [2fcdee6e9c],BloodorangemixHardcore_bloodorangemix.safetensors [55cfe9bdf0],calicomix_v11Pruned.safetensors [20d3eca4c2],CounterfeitV30_v30.safetensors [cbfba64e66],feiwumix_v10Slightflat.safetensors [6505688a22],hassakuHentaiModel_v13.safetensors [7eb674963a]"
re: forever ai mistakes
tried to experiment with another image and it made a bunny in the background with a massive cock
re: forever ai mistakes
decided to jack up the clip skip to 12 and got these
it's a shrug from me; not bad, but feel generic
it's a shrug from me; not bad, but feel generic
re: forever ai mistakes
first nonad: the first tenma lora, <lora:char-tenma:0.8>
second nonad: the second, newer tenma lora, <lora:tenma:0.8>
i'm having fun with these comparisons, despite the fact i don't know what i'll do with this information
-
1girl, small breasts, loli, small breasts, young, curvy, tenma, pink eyes, fox girl, fox tail, black jacket, hoodie, <lora:char-tenma:0.8>
Negative prompt: worst quality, low quality, mutated hands and fingers, bad anatomy, wrong anatomy, ugly, ((mutation)),
Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 564620271, Size: 512x768, Model hash: 7eb674963a, Model: hassakuHentaiModel_v13, Version: v1.2.1, Script: X/Y/Z plot, X Type: Sampler, X Values: "Euler a,DPM++ SDE Karras,UniPC", Y Type: Clip skip, Y Values: 1-5[3]
second nonad: the second, newer tenma lora, <lora:tenma:0.8>
i'm having fun with these comparisons, despite the fact i don't know what i'll do with this information
-
1girl, small breasts, loli, small breasts, young, curvy, tenma, pink eyes, fox girl, fox tail, black jacket, hoodie, <lora:char-tenma:0.8>
Negative prompt: worst quality, low quality, mutated hands and fingers, bad anatomy, wrong anatomy, ugly, ((mutation)),
Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 564620271, Size: 512x768, Model hash: 7eb674963a, Model: hassakuHentaiModel_v13, Version: v1.2.1, Script: X/Y/Z plot, X Type: Sampler, X Values: "Euler a,DPM++ SDE Karras,UniPC", Y Type: Clip skip, Y Values: 1-5[3]
re: forever ai mistakes
CUNNY AI MODEL
CUNNY CUNNY CUNNY CUNNY
I CAN MAKE CUNNY
CUNNY CUNNY CUNNY CUNNY
I CAN MAKE CUNNY
re: forever ai mistakes
it's a pretty good checkpoint
accidentally used the old pippa model, so i have two sets
also, seems like tenma's new model is more built for sex than the previous one, pretty lewd on the outset despite my lack of specification
accidentally used the old pippa model, so i have two sets
also, seems like tenma's new model is more built for sex than the previous one, pretty lewd on the outset despite my lack of specification
re: forever ai mistakes
well i didn't say this, but guess the clip step is just high or smth
also the guy's legs are morphed into a chair in these images, alongside the fact tenma has an upside-down cock and balls in one of 'em
also the guy's legs are morphed into a chair in these images, alongside the fact tenma has an upside-down cock and balls in one of 'em
re: forever ai mistakes
i seem to have forgotten and repeated my past mistakes
using loras is a bit of a pain in the ass, isn't it, since it has such a massive tendency to fuck up
i need to make more notes on how to use ai so i understand what to do after getting back into it after a while
using loras is a bit of a pain in the ass, isn't it, since it has such a massive tendency to fuck up
i need to make more notes on how to use ai so i understand what to do after getting back into it after a while
re: forever ai mistakes
the worst kiki generations is better than nearly all my better generations
re: forever ai mistakes
i removed the lora, and decided to use words instead
seems like the ai wants to impregnate pippa
seems like the ai wants to impregnate pippa
re: forever ai mistakes
i honestly prefer the old pippa model
for some reason, this new model, although it can generate good stuff, is kinda bad
i'm not sure how to describe it, explain it, but i feel like i understand somewhat of how it works based on what it generates well and what it doesn't
the old model, while inconsistent and generally has a lot of errors and is prone to it, has more variety than what the new one seems to possess, or at the very least, it has a good enough abstract of pippa to have flexibility, while the new model is probably built to generate pippa holding guns if anything, given the page's preview images
the new model is very tight up the ass on a particular style, and maybe that's because my weights are too high, but i'll need to experiment more to test that
old model (left), new model (right)
for some reason, this new model, although it can generate good stuff, is kinda bad
i'm not sure how to describe it, explain it, but i feel like i understand somewhat of how it works based on what it generates well and what it doesn't
the old model, while inconsistent and generally has a lot of errors and is prone to it, has more variety than what the new one seems to possess, or at the very least, it has a good enough abstract of pippa to have flexibility, while the new model is probably built to generate pippa holding guns if anything, given the page's preview images
the new model is very tight up the ass on a particular style, and maybe that's because my weights are too high, but i'll need to experiment more to test that
old model (left), new model (right)
re: forever ai mistakes
3 nipples, 3 legs
this ai likes to give pippa 3's of things
this ai likes to give pippa 3's of things
re: forever ai mistakes
the new models...
i really am unsure how to think of them, because they plainly suck and there's too much work to do to make these work
the phase connect loras just plain suck?
i really am unsure how to think of them, because they plainly suck and there's too much work to do to make these work
the phase connect loras just plain suck?
re: forever ai mistakes
i tried to use controlnet on my old drawings, and used an old sketch on the last matrix
the results are something
---
1girl, small breasts, loli, curvy, tenma, pink eyes, fox girl, fox tail, drunk, holding beer, katana, on stage, brown clothing, <lora:tenma:0.9>
Negative prompt: worst quality, low quality, mutated hands and fingers, bad anatomy, wrong anatomy, ugly, ((mutation))
Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 1185817708, Size: 512x512, Model hash: 8b1ae042dc, Model: akkaimix_v20Bakedvae, Clip skip: 3, Version: v1.2.1, ControlNet 0: "Module: lineart_anime, Model: control_v11p_sd15s2_lineart_anime [3825e83e], Weight: 1, Resize Mode: Crop and Resize, Low Vram: False, Processor Res: 512, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced", Script: X/Y/Z plot, X Type: [ControlNet] Control Mode, X Values: "Balanced,My prompt is more important,ControlNet is more important", Y Type: Sampler, Y Values: "Euler a,DPM++ SDE Karras,UniPC"
the results are something
---
1girl, small breasts, loli, curvy, tenma, pink eyes, fox girl, fox tail, drunk, holding beer, katana, on stage, brown clothing, <lora:tenma:0.9>
Negative prompt: worst quality, low quality, mutated hands and fingers, bad anatomy, wrong anatomy, ugly, ((mutation))
Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 1185817708, Size: 512x512, Model hash: 8b1ae042dc, Model: akkaimix_v20Bakedvae, Clip skip: 3, Version: v1.2.1, ControlNet 0: "Module: lineart_anime, Model: control_v11p_sd15s2_lineart_anime [3825e83e], Weight: 1, Resize Mode: Crop and Resize, Low Vram: False, Processor Res: 512, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced", Script: X/Y/Z plot, X Type: [ControlNet] Control Mode, X Values: "Balanced,My prompt is more important,ControlNet is more important", Y Type: Sampler, Y Values: "Euler a,DPM++ SDE Karras,UniPC"
re: forever ai mistakes
i spent nearly an entire fucking hour generating using DDIM as a sampler (instead of UniPC, which feels like it's more consistent in generating the good stuff) while creating a giant table comparing checkpoints and control modes
i think i gathered a good enough sample despite that; balanced and controlnet are what to go for if your lineart is peak, but prompt is more important to give the ai more freedom and let it do its thing
i think i gathered a good enough sample despite that; balanced and controlnet are what to go for if your lineart is peak, but prompt is more important to give the ai more freedom and let it do its thing
re: forever ai mistakes
this says lora, but is actually a lyco
amazing
amazing
re: forever ai mistakes
i was going to make that shiina animation that apparently wasn't made because likely management came down on it
but what i thought was an animation setting, is generating a bunch of images that don't look like animation in sequence
but what i thought was an animation setting, is generating a bunch of images that don't look like animation in sequence
re: forever ai mistakes
i'm not sure how to turn this into a video quite yet either, but this is the product of what was generated
re: forever ai mistakes
i'm trying to identify a way to hasten the video-generation process so i don't have to leave it open overnight for all the frames to generate, like reducing the steps needed to generate a satisfactory image, but i'm not sure how to hasten it
re: forever ai mistakes
guess i'll just leave this running overnight, see what the end result is
now that i know how to do it properly, i'll be looking forward to the finished product, whatever it may be
now that i know how to do it properly, i'll be looking forward to the finished product, whatever it may be
re: forever ai mistakes
the first 17 frames on loop
decided to do some calculations, but if this ai decides to go through each and every frame of this shit, it'll be roughly 1216 frames to be generated
and given the current rate, it'll take 2 whole days, and 8 minutes to generate the entire thing
which, is some bullshit because the majority of it is the same looped motion
decided to do some calculations, but if this ai decides to go through each and every frame of this shit, it'll be roughly 1216 frames to be generated
and given the current rate, it'll take 2 whole days, and 8 minutes to generate the entire thing
which, is some bullshit because the majority of it is the same looped motion
re: forever ai mistakes
by altering the original file to contain the main 3 loops in this animation, i'm able to shorten the generation time to likely 5 hours, 8 minutes, and 45 seconds total, by generating only 130 images
because the ai doesn't seem to take into account the camera motion, that's both something to watch out for and to fix if making an ai animation, and has to be edited post-generation
which might help to create a bigger generation
i might test out drawing a sketch animation myself, then having the ai do the hard lifting, if this ends up turning well, and seeing its practical usages
i would use one of the animations i've made in the past, but stick figures and tenma transforming into a geats gijinka, i feel, probably wouldn't work that well, considering the ai likely doesn't want to do effects or would probably fuck it up
might be possible to do if you capture just the body movements of the henshin, then add effects post-generation, but i'm just speculating
my old and short animation:
https://invidious.varishangout.net/watch?v=F7a37onFyzU
because the ai doesn't seem to take into account the camera motion, that's both something to watch out for and to fix if making an ai animation, and has to be edited post-generation
which might help to create a bigger generation
i might test out drawing a sketch animation myself, then having the ai do the hard lifting, if this ends up turning well, and seeing its practical usages
i would use one of the animations i've made in the past, but stick figures and tenma transforming into a geats gijinka, i feel, probably wouldn't work that well, considering the ai likely doesn't want to do effects or would probably fuck it up
might be possible to do if you capture just the body movements of the henshin, then add effects post-generation, but i'm just speculating
my old and short animation:
https://invidious.varishangout.net/watch?v=F7a37onFyzU
re: forever ai mistakes
i decided to switch to kiki's model to see if it'd turn out better, seeing if a better model would make a difference
the left contains all 203 frames i generated, the right one contains only the 35 frames, since i noticed in the left version, it seems to loop around the same frames (with only slight differences)
i went to sleep, and only generated 203 frames (almost 7 seconds), and the full loop is just 35 frames (slightly more than a second)
edit: mf one of the frames is off because i did the creation of this gif weirdly
the left contains all 203 frames i generated, the right one contains only the 35 frames, since i noticed in the left version, it seems to loop around the same frames (with only slight differences)
i went to sleep, and only generated 203 frames (almost 7 seconds), and the full loop is just 35 frames (slightly more than a second)
edit: mf one of the frames is off because i did the creation of this gif weirdly
re: forever ai mistakes
decided to share this fixed version on /pyon/
i know 4chan doesn't like loli stuff outside of /b/ iirc (and the only way around it from the last time i remembered was through catbox), and i'm still unsure about catbox and if it allows stuff like this if they find out, but eh
whatever happens, happens i guess, since i don't use 4chan at all, and the only times i went there was mainly to be in papercrafting and origami, and i haven't touched that in forever
i know 4chan doesn't like loli stuff outside of /b/ iirc (and the only way around it from the last time i remembered was through catbox), and i'm still unsure about catbox and if it allows stuff like this if they find out, but eh
whatever happens, happens i guess, since i don't use 4chan at all, and the only times i went there was mainly to be in papercrafting and origami, and i haven't touched that in forever
re: am i an ai novice now?
i'm thinking maybe i can try to make an ai animation out of someone else's koikatsu animation
i'm thinking gawr gura because she's a loli, and also, her model is probably one of the best that's trained, considering her popularity
unfortunately, it appears i like obscure lolis more than popular ones, and thus they were the first i went with...
i have put myself the entire time at risk of falling into the habit of working around bad models, rather than learning how to properly use stable diffusion
which i worry would've made me worse at any model other than the ones i'd work with, or like a cat with a bell or rock lee and his weights, i do ungodly amazingly making images and animations due to working in harder conditions to accomplish things after being dropped into the kiddy pool side of things
i'm thinking gawr gura because she's a loli, and also, her model is probably one of the best that's trained, considering her popularity
unfortunately, it appears i like obscure lolis more than popular ones, and thus they were the first i went with...
i have put myself the entire time at risk of falling into the habit of working around bad models, rather than learning how to properly use stable diffusion
which i worry would've made me worse at any model other than the ones i'd work with, or like a cat with a bell or rock lee and his weights, i do ungodly amazingly making images and animations due to working in harder conditions to accomplish things after being dropped into the kiddy pool side of things
Follow
re: am i an ai novice now?
@ninja8tyu that horse gura needs its own content warning jesus christ
