Pregnant AI Artwork
Bo27's Stable Diffusion work
  2 of 7  
  • 19 Vote(s) - 4.47 Average
  • 1
  • 2
  • 3
  • 4
  • 5
bo27
(Edited)
(Edited)
I don't use one model, I actually created and am still creating a very wide selections of models. I use the "base" models 1.5, f222, hyperpreg, sd 2.0, and now experimenting with Waifu, but I have multiple combinations at various strengths at increments of 5% from 5 to 95% ( a lot of GBs of checkpoints). I will go up and down the scale. Usually use the "more dramatic models" to get the theme, then in paint with the more "true to form" models keeping the de-noising low. like I probably use 4-30 models per picture over time, and these pictures aren't done quickly they usually take hours, but I think the results are worth it. My checkpoints are significant to me, and I have like a word document where I take notes. I might even experiment with mixing two bases, for example f222 and hyperpreg. That actually is my preferred go to, again ranging from 5%-95%. Where 5% is more f222 and 95% is more hyperpreg. The sweet spot in there for me is typically "25-65". Only use more or less for more dramatic effects, and i might just do it for the belly only, or the breasts, then in paint with 5% or even f222 itself. I used to use 1.5 and hyperpreg, but F222 usually gives me more true to form  results. I also sometimes will speed up the process by using photoshop to "get the shape/design I want or to remove extra fingers/artifacts, then send back to inpainting. Again hours and hours, and its an on going experiment.

For experiments i will take a picture, its seed number, then using the "X/Y" plot referrence different checkpoints and denoising strengths. So for inpainting or img2img, to keep the look the same. I usually have my cfg at 3.5-5, and my denoising from .4-.65. It rarely goes above or below that thresholds.

I think I've gotten better, and so I probably will remaster some of these older pictures, like the pregnant lady drinking the beer, I imagine she will look a lot better now. I can even in paint some cool stuff like Icarus Illustrations, but I don't post it unless the guy gives me permissions, he probably spent a lot of time generating that content, and I would feel bad psting my pictures using img2img and inpainting to get my results. However, I think if you had someone with that talent do a combination of CGI and AI painting, even with Deforum, he could probably pop out amazing stuff.
Liked by Fyra yu Chan 1999 (Mar 22, 2023), (Dec 18, 2022)
rotund_regina
Love these! Combination of a bunch of my favorites. Hyperpregnant with a combination of a pregnant belly carrying low and out with a round gut.
Liked by Fyra yu Chan 1999 (Mar 22, 2023)
Victorexcx
(December 1, 2022, 9:43 am)bo27 I don't use one model, I actually created and am still creating a very wide selections of models. I use the "base" models 1.5, f222, hyperpreg, sd 2.0, and now experimenting with Waifu, but I have multiple combinations at various strengths at increments of 5% from 5 to 95% ( a lot of GBs of checkpoints). I will go up and down the scale. Usually use the "more dramatic models" to get the theme, then in paint with the more "true to form" models keeping the de-noising low. like I probably use 4-30 models per picture over time, and these pictures aren't done quickly they usually take hours, but I think the results are worth it. My checkpoints are significant to me, and I have like a word document where I take notes. I might even experiment with mixing two bases, for example f222 and hyperpreg. That actually is my preferred go to, again ranging from 5%-95%. Where 5% is more f222 and 95% is more hyperpreg. The sweet spot in there for me is typically "25-65". Only use more or less for more dramatic effects, and i might just do it for the belly only, or the breasts, then in paint with 5% or even f222 itself. I used to use 1.5 and hyperpreg, but F222 usually gives me more true to form  results. I also sometimes will speed up the process by using photoshop to "get the shape/design I want or to remove extra fingers/artifacts, then send back to inpainting. Again hours and hours, and its an on going experiment.

For experiments i will take a picture, its seed number, then using the "X/Y" plot referrence different checkpoints and denoising strengths. So for inpainting or img2img, to keep the look the same. I usually have my cfg at 3.5-5, and my denoising from .4-.65. It rarely goes above or below that thresholds.

I think I've gotten better, and so I probably will remaster some of these older pictures, like the pregnant lady drinking the beer, I imagine she will look a lot better now. I can even in paint some cool stuff like Icarus Illustrations, but I don't post it unless the guy gives me permissions, he probably spent a lot of time generating that content, and I would feel bad psting my pictures using img2img and inpainting to get my results. However, I think if you had someone with that talent do a combination of CGI and AI painting, even with Deforum, he could probably pop out amazing stuff.
The beer one was pretty good already, do you take requests on this? Just curious
Liked by Fyra yu Chan 1999 (Mar 22, 2023)
bo27
(Edited)
(Edited)
I decided to challenge to myself to a theme, the idea is this lady ate a mushroom from the world of Super Mario and things went bad.
                                                           
Liked by 16 members: Firefox12 (May 5, 2023), Inflationwolf (Apr 28, 2023), Vika (Apr 24, 2023), Fyra yu Chan 1999 (Mar 22, 2023), ahhah55 (Mar 16, 2023), (Mar 12, 2023), Iganinjaartist (Mar 12, 2023), Njameson (Mar 11, 2023), Bellyinflation93 (Mar 7, 2023), BalValor (Dec 24, 2022), mr-enigma (Dec 14, 2022), Tcorker (Dec 8, 2022), jokerpaso (Dec 3, 2022), Baelthar (Dec 2, 2022), thojm (Dec 2, 2022), AyyLmaoAlien12 (Dec 2, 2022)
bo27
(December 1, 2022, 11:11 pm)Victorexcx
(December 1, 2022, 9:43 am)bo27 I don't use one model, I actually created and am still creating a very wide selections of models. I use the "base" models 1.5, f222, hyperpreg, sd 2.0, and now experimenting with Waifu, but I have multiple combinations at various strengths at increments of 5% from 5 to 95% ( a lot of GBs of checkpoints). I will go up and down the scale. Usually use the "more dramatic models" to get the theme, then in paint with the more "true to form" models keeping the de-noising low. like I probably use 4-30 models per picture over time, and these pictures aren't done quickly they usually take hours, but I think the results are worth it. My checkpoints are significant to me, and I have like a word document where I take notes. I might even experiment with mixing two bases, for example f222 and hyperpreg. That actually is my preferred go to, again ranging from 5%-95%. Where 5% is more f222 and 95% is more hyperpreg. The sweet spot in there for me is typically "25-65". Only use more or less for more dramatic effects, and i might just do it for the belly only, or the breasts, then in paint with 5% or even f222 itself. I used to use 1.5 and hyperpreg, but F222 usually gives me more true to form  results. I also sometimes will speed up the process by using photoshop to "get the shape/design I want or to remove extra fingers/artifacts, then send back to inpainting. Again hours and hours, and its an on going experiment.

For experiments i will take a picture, its seed number, then using the "X/Y" plot referrence different checkpoints and denoising strengths. So for inpainting or img2img, to keep the look the same. I usually have my cfg at 3.5-5, and my denoising from .4-.65. It rarely goes above or below that thresholds.

I think I've gotten better, and so I probably will remaster some of these older pictures, like the pregnant lady drinking the beer, I imagine she will look a lot better now. I can even in paint some cool stuff like Icarus Illustrations, but I don't post it unless the guy gives me permissions, he probably spent a lot of time generating that content, and I would feel bad psting my pictures using img2img and inpainting to get my results. However, I think if you had someone with that talent do a combination of CGI and AI painting, even with Deforum, he could probably pop out amazing stuff.
The beer one was pretty good already, do you take requests on this? Just curious
What do you have in mind, I could try..
Liked by Fyra yu Chan 1999 (Mar 22, 2023)
bo27
And more
                           
Liked by Firefox12 (May 5, 2023), Inflationwolf (Apr 28, 2023), Vika (Apr 24, 2023), (Mar 12, 2023), Nickrider (Mar 2, 2023), preggohotties (Jan 1, 2023), mr-enigma (Dec 14, 2022), Tcorker (Dec 8, 2022), (Dec 5, 2022)
bo27
Some more for ya
                           
Liked by (Mar 12, 2023), Chronos999 (Mar 8, 2023), caddyova273 (Dec 25, 2022), mr-enigma (Dec 14, 2022), Tcorker (Dec 8, 2022), awakenedhero (Dec 7, 2022), (Dec 7, 2022)
bo27
A new addition
   
Liked by Inflationwolf (Apr 28, 2023), (Mar 12, 2023), Bellyinflation93 (Mar 7, 2023), mr-enigma (Dec 19, 2022), thojm (Dec 18, 2022), Tcorker (Dec 18, 2022), azarafa (Dec 18, 2022)
Victorexcx
(December 18, 2022, 5:21 am)bo27 A new addition
Any updates on the beer girl
bo27
(December 22, 2022, 1:43 pm)Victorexcx
(December 18, 2022, 5:21 am)bo27 A new addition
Any updates on the beer girl
I wasn't working on her, but I can revisit. Heres my next update.
   
Liked by Inflationwolf (Apr 28, 2023), (Mar 12, 2023), Tcorker (Dec 24, 2022), hugojps (Dec 24, 2022)

Related Threads Author Replies Views Last Post
269's Diffusion Dump | A1111 and ControlNet 2692205852541213 7 4,710 February 15, 2024, 2:47 am
Last Post: 2692205852541213
My AI Work jackiepg9 1 1,587 January 1, 2024, 2:59 pm
Last Post: jackiepg9
Asian preggo using AI (stable difusion) bla12322132 1 1,539 August 7, 2023, 5:16 pm
Last Post: awakenedhero
Playing with stable diffusion hugojps 15 12,762 June 6, 2023, 2:10 pm
Last Post: hugojps

Users browsing this thread: 1 Guest(s)