[Stable Diffusion] Prompt Sharing and Learning Thread

Jimwalrus

Jimwalrus

Well-Known Member
Sep 15, 2021
1,045
3,994
Mr-Fox said:
Some testing.

You must be registered to see the links

You must be registered to see the links

View attachment 2435182

You must be registered to see the links

View attachment 2435189 View attachment 2435190
Click to expand...
Shirtlift and Skirtlift are almost magical - they pretty much just work! Interestingly with Skirtlift, if you nrompt out 'skirt' and 'dress', it will mould whatever's actually there into the rough shape of a lifted skirt, revealing whatever's underneath.
 
  • Like
Reactions: Mr-Fox and Sepheyer
Mr-Fox

Mr-Fox

Well-Known Member
Jan 24, 2020
1,401
3,802
A question of curiosity. Now that I have been doing this SD thing for a while I see that the folders full of generated images is starting to build up. How do you organize the image prompts you wish to keep, and how do you avoid the file buildup?:unsure:
 
  • Like
Reactions: Sepheyer
fr34ky

fr34ky

Active Member
Oct 29, 2017
812
2,189
Lately I've been making some stuff where I fix errors with photoshop and send it back to img2img and repeat, I'm thinking about dusting off my wacom tablet ;)
 
  • Like
Reactions: Sepheyer
Mr-Fox

Mr-Fox

Well-Known Member
Jan 24, 2020
1,401
3,802
Seph created this awesome prompt that I have used as a test subject and also edited a little but Seph did an amazing job and keeps on doing.:love: I hope people don't mind when I use their creation as a base or starting point for my own projects and explorations. I love collaborations and only have this in mind when I borrow stuff from others. I promise that I will do things from scratch also. :D
I switched to
You must be registered to see the links
. Sometimes it seems almost foolish to even bother with something else.
After I found a really good seed I used Plot Script to find the best Samplings Steps. Once again I observed that more steps doesn't automatically equal better results, sometimes it's the opposite. This one was generated with only 12 Samplings Steps.

(Enhanced with photoshop).
00117 copy.png

You don't have permission to view the spoiler content. Log in or register now.
 
Last edited:
  • Red Heart
  • Like
Reactions: VanMortis, sharlotte and Sepheyer
Mr-Fox

Mr-Fox

Well-Known Member
Jan 24, 2020
1,401
3,802
fr34ky said:
Lately I've been making some stuff where I fix errors with photoshop and send it back to img2img and repeat, I'm thinking about dusting off my wacom tablet ;)
Click to expand...
Sounds interesting. Could you elaborate? Do you mean using a tablet with a pen to edit in photoshop? I'm also curious about your process, going back and forth with photoshop and img2img. I have done a bit of this myself.
 
Last edited:
  • Like
Reactions: VanMortis and Sepheyer
fr34ky

fr34ky

Active Member
Oct 29, 2017
812
2,189
Yes, a tablet with a pen to edit in photoshop, you can draw in photoshop with it using the brush, etc.

This one I made it with mouse but it's an example:

1) I wanted to make dexter moms ass, so I created a prompt with the pants, apron and glove, but even when asking for the colors, everything was mixed up.

01.png
2) I took the first picture to photoshop and changed roughly the colors as I wanted
02.JPG
3) sent to img2img and kind of fuses all the things, you kind of force the picture down the AI's throat in img2img.
03.png
That's not finished, but you can see how to AI fused my rough edit and made the shirt look like one thing.

That way you don't have to make the final drawing yourself, you just edit the colors and position of the things and the AI finishes drawing.
 
  • Like
  • Red Heart
Reactions: Elefy, VanMortis, Mr-Fox and 1 other person
Mr-Fox

Mr-Fox

Well-Known Member
Jan 24, 2020
1,401
3,802
fr34ky said:
Yes, a tablet with a pen to edit in photoshop, you can draw in photoshop with it using the brush, etc.

This one I made it with mouse but it's an example:

1) I wanted to make dexter moms ass, so I created a prompt with the pants, apron and glove, but even when asking for the colors, everything was mixed up.

View attachment 2436648
2) I took the first picture to photoshop and changed roughly the colors as I wanted
View attachment 2436649
3) sent to img2img and kind of fuses all the things, you kind of force the picture down the AI's throat in img2img.
View attachment 2436666
That's not finished, but you can see how to AI fused my rough edit and made the shirt look like one thing.

That way you don't have to make the final drawing yourself, you just edit the colors and position of the things and the AI finishes drawing.
Click to expand...
Thank you. Also awesome image. :love: Let's get cooking...:D
 
  • Haha
Reactions: fr34ky
Synalon

Synalon

Member
Jan 31, 2022
225
663
atheran said:
./Stable Diffusion/webui/models/*depending what you want*

At least that's on Automatic1111 which I assume you're using. Text inversions go to ./webui/embeddings
Click to expand...
I'm trying different versions, Easy Diffusion and Invoke Ai right now. I saw Mr Fox pose about the safety checks nsfw filters taking up 2GB of vram and the fix but I can't find the "modules" folder to put the fix file into.
 
Mr-Fox

Mr-Fox

Well-Known Member
Jan 24, 2020
1,401
3,802
I did a large checkpoint test with the prompt and seed from this post, that was based on seph's awesome work.
The intention was to use the Plot Script and generate an image with all checkpoints I have. Because I kept getting runtime error that interrupted the test I had to go through the rest manually, about half of them. During this process I found out that some checkpoints has a config yaml file that I was missing so watch out for this. It was PRMJ in this case but I have seen other checkpoints that also use a config file.
In the end it was a very interesting and good test with some nice surprises and others that left a lot to be desired.
Keep in mind though that some checkpoints are trained for a specific niche and does that very well but not this prompt.
You don't have permission to view the spoiler content. Log in or register now.
 
Last edited:
  • Like
  • Red Heart
Reactions: Schlongborn, VanMortis and Sepheyer
Mr-Fox

Mr-Fox

Well-Known Member
Jan 24, 2020
1,401
3,802
Synalon said:
I'm trying different versions, Easy Diffusion and Invoke Ai right now. I saw Mr Fox pose about the safety checks nsfw filters taking up 2GB of vram and the fix but I can't find the "modules" folder to put the fix file into.
Click to expand...
It is not working, I got syntax error. If no one else knows something about it, don't waste your time. The article I linked to has other things one can try to fix cuda memory error.
 
Last edited:
  • Like
Reactions: Synalon and Jimwalrus
Synalon

Synalon

Member
Jan 31, 2022
225
663
Mr-Fox said:
It is not working, I got syntax error. If no one else knows something about it, don't waist your time. The article I linked to has other things one can try to fix cuda memory error.
Click to expand...
Thanks for letting me know, I just installed Automatic1111 and was about to try it.
 
Last edited:
  • Hey there
Reactions: Mr-Fox
Synalon

Synalon

Member
Jan 31, 2022
225
663
Where do you change the face fix to codeformer with automatic1111, I can't find the option.
 
You must log in or register to reply here.
Top Bottom