[Stable Diffusion] Prompt Sharing and Learning Thread

sharlotte

Member
Jan 10, 2019
268
1,436
Has anyone tried to make older women in stable diffusion? Not sure what prompt to select. I used stuff like 45-years old in the prompts but it only shows me younger women - any idea what type of tags/prompts to use?
 

Sepheyer

Well-Known Member
Dec 21, 2020
1,528
3,598
Has anyone tried to make older women in stable diffusion? Not sure what prompt to select. I used stuff like 45-years old in the prompts but it only shows me younger women - any idea what type of tags/prompts to use?
I think this is really model dependent. In PFG you'll get a pleasant MILF while in Protogen you'll be looking at cougars.

Prompt inside.
00010-3365253823-40 year old woman from Sweden._(stunning European face)._Middleaged. Chubby._...png
You don't have permission to view the spoiler content. Log in or register now.
You don't have permission to view the spoiler content. Log in or register now.
 
Last edited:

fr34ky

Active Member
Oct 29, 2017
812
2,167
Man, I dream of the moment we'll be able to seamlessly import or generate girls into some 3d software like daz or blender and then put them in VAM, it seems like it's not very far and some very 3d savy guys might already be able to do it...
 
  • Like
Reactions: Mr-Fox

yodin

New Member
May 28, 2019
1
0
Re-running the generation with the same parameters, but setting it to run 30 steps of upscale x2 (to 1024x1024) with a denoise of just 0.1 seems to fix the hand (and puts a glove on, which is in the prompt, but not the original image):
View attachment 2359500
Effectively, upscaling isn't just there to improve resolution, it can also be used to fix, enhance (or completely fuck up) features. It's not as easily directed though, so it's much more of a crap shoot.

You don't have permission to view the spoiler content. Log in or register now.
 

Jimwalrus

Active Member
Sep 15, 2021
900
3,323
whats the outfit for the 1st pic?
Dunno, it was carved out of static by AI!

In all seriousness, it's not my original prompt, coming from Sepheyer.
Stable Diffusion just takes the prompt and runs with it, often tripping over itself as it does so.

So, it's not any particular 'outfit', just what SD decided to make at the time. The other pics are the same prompt, different seed.
Also, bear in mind, almost all the AI-generated images you see online are cherry-picked to remove the really scary stuff.
I don't know if you've tried creating your own images, but I can see a time soon when some poor sod gets PTSD from all the Chthulu-esque eldritch horrors that SD has delivered to them...

As I've previously recommended, if you want to try it out but don't have the GPU, get yourself over to unstability.party (yes, that is the URL!) for some NSFW AI stuff.
If you look at what people have made, especially when trying to create sex scenes, there's some scary shit among the good stuff!
 

Jimwalrus

Active Member
Sep 15, 2021
900
3,323
BTW, I keep mentioning unstability.party, it's quite a good place to hone your prompts early on.

As an example, here's one I made on with the prompts and settings below:
You don't have permission to view the spoiler content. Log in or register now.
A_photo_of_two_stunningly_beautiful_25y_old_lesbians_in_business_suit_jackets_blonde_topless_3...jpg

Tried to recreate it in SD on my own machine. The "ar: 9:11" is asking for a 9:11 aspect ratio (or as near as SD can do). I looked at the output image which was 1024x1280 and, knowing it had been upscaled x2, worked back to 512x640.
Had to guess at the denoising strength, went with 0.2.

This was the result:
00031-3028803833.png

Left nipple has gone inverted, Ms. PurpleHair has lost her highlights (and is now giving us some side-eye!) and the necklace is nowhere to be found*, but broadly it's giving much the same results without spending hundreds on a GPU.

I think it may have Xtensors enabled or something, it's hard to tell exactly what's going on 'under the hood'.

*it's not in the pre-upscaled image either
 
  • Like
Reactions: Mr-Fox and Sepheyer

fr34ky

Active Member
Oct 29, 2017
812
2,167
Interesting articles about negative prompts from stable-diffusion-art. They have many other tutorials also.





I've been testing a lot of different poses and pictures with some particular details, still can't make things exactly as I want, but I learned that using the same clutter of negative prompts is really 'negative' to achieve wanted results.

Negative prompts have to be crafted as you go while creating the picture, it's like rowing a boat, you have to alternate the use of both rows to go to the right direction, and if you use one row only you spin in one direction and then with the other spin in the other direction but stay in the same place. (I've been playing too much Sea Of Thieves may be)

I swear I'm not into chubby girls in really life (not that there is anything wrong with it), but for some reason ai art and digital art makes every body type look so hot, bellies, extra skin and all that particulary, I'm also trying many things with this concept, I hope I can post something soon that I like.
 
Last edited: