/gen/

(345 KB, 1024x1600, 1.jpg) (337 KB, 1024x1600, 5.jpg) (321 KB, 1024x1600, 10.jpg) (342 KB, 1024x1600, 14.jpg) (337 KB, 1024x1600, 17.jpg) (372 KB, 1024x1600, 19.jpg)
what software did you use to make these?
(515 KB, 512x800, 00097-989195540-2023-02-17.png) (511 KB, 512x800, 00096-989195539-2023-02-17.png) (494 KB, 512x800, 00095-989195538-2023-02-17.png) (507 KB, 512x800, 00093-989195536-2023-02-17.png) (517 KB, 512x800, 00091-989195534-2023-02-17.png) (533 KB, 512x800, 00090-989195533-2023-02-17.png)
>>26558
Stable diffusion with UI called Automatic1111.

In stable diffusion you can use whichever AI model you want to, the default is quite bad for this purpose. I used a model called Hasanblend. There could be other which are better, haven't experimented with that.
>>26569
>>26579
amazing stuff, could you share a bit about your workflows and prompts?
Do you use Hassablend 1.4 or 1.5?
Also what differences did u notice between the two?
(69 KB, 712x242, Untitled.png)
>>26581
Thank you very much!

If you want to check the prompts for some specific image here you can download it and check the exif data (https://www.metadata2go.com/view-metadata) The prompts and model information is there for almost every pic.

Prompts are guite generic. I have refined them little bit by generating bunch of images in bulk and then checking which ones I prefer/dislike and then try to refine the prompts from there.

Keep in mind that stable diffusion is not language model so all it knows is that words have correlation to certain visual aspects. So when I use prompts I think of more as tags and not so much as a verbal guidance.
(416 KB, 512x800, 00166-1190159296-2023-02-17.png)
>>26581
If you wish to create a progression similar to what's in the first message workflow is a bit harder:


>generate one average girl
>use img2img for the average girl to make her bigger or smaller. Don't change the seed and it's best to use same prompt as for the first girl and just make one or two edits in it.
>at the end, inpaint face so that their faces would look similar. Use same prompt and seed for every girl.

Most time is spent when trying to make the growth proportional everywhere. I found it's best to change very general terms in prompts (eg. chubby) rather than specifying separately each body part (eg. thick thighs).

Often times you just can't get decent progression and you have to start again with different seed.
(548 KB, 512x800, 00158-20230215012308.png)
>>26583
I'm using Hassanblend 1.5

I haven't used any other model so I don't know the differences :p.

What I couldn't do with hassan often times was to add little bit chub to specific body parts. I think there could maybe be a model with bbw focus that would know what "belly roll" etc means.
What an oddly specific thread
(433 KB, 512x704, more curvy.png)
bit more hips and realistic texture. this is fun.
>>26590
Nice job, but these gals are Finland-chubby at best. Can you do a few that are US- (or at least UK-) chubby?
>>26607
Man, if you could take pics of women and add say 20 pounds each stage until they were BBWs it'd be pretty fricken sweet.

As it is this tech is pretty amazing, great thread BTW. Love me some progression, being able to do this to pretty much any pic you want is fantasy tier
>>26545 (OP)
great thread, thank you for sharing with us
>>26654
Thank you sir

>>26591
i’ve been ”suffering” of yellow fever since my teenage years ;)

>>26607
Unfortunately I moved abroad and left my PC back in Finland so someone else has to continue from here.

>>26593
Great one!

-OP
Wow, they're all amazing, fantastic work. They look so real!

As someone who has literally no knowledge of these things, who difficult is it to learn how to create these? After ownloading the mentioned programms.
Oh wow I genuinely thought these were real at first!
Wow! Gr8t work! Thanks!
>>27666
The 6th photo looks like a hyper-realistic doll, or like Anime came to life.
>>27699

Only the best for the queen
>>26545 (OP)
can you make more of these compilation gifs/videos its so hot

Back to top