[Stable Diffusion] Prompt Sharing and Learning Thread

[Mr-Fox]

I can't shake off this feeling that word "tiny" almost always gets translated into a girl's frame rather than into the clothes size. Would you have the same hintch?

Anyways, I replaced 1.6s with 1.1s and got this:

View attachment 2479957

Spoiler: More

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

Now imma try the "widescreen format" version.

Very nice. No I have not had tiny have this effect but maybe it is because I have been using heavy weight for the boobas and curvy figure etc. You can of course change and rework the prompt as you see fit in order to get the result you want.

 

[Mr-Fox]

I got really cool results when I used pink or colored highlights instead of blonde.


When using a specific color in the prompt though it has a tendency to also be used for the clothes. There's an extension for this but it has been a bit of hit n miss.

You must be registered to see the links

 

[Sepheyer]

I got really cool result when I used pink or colored highlights instead of blonde.

View attachment 2480007 View attachment 2480009
When using a specific color in the prompt though it has a tendency to also be used for the clothes. There's an extension for this but it has been a bit of hit n miss.

You must be registered to see the links

So, for a quick second ComfyUI handled color contamination really well. Then an update happened, and we are back to the usual color contamination BS.

And I did try that A1111 extension that limits token scope, didn't feel it was reliable enought.

ComfyUI glory days, alas they didn't last. RIP its ability to give 80% chance of getting the colors right:

 

[Schlongborn]

So, for a quick second ComfyUI handled color contamination really well. Then an update happened, and we are back to the usual color contamination BS.

You can git bisect to find the commit that introduced the breaking change, then tag the commit before that in your local repo. That way you can always go back to it.

EDIT: I assume it is a recent change, so it would be as easy as doing "git log", going through the commits and then just checking them out and trying them. You can checkout a commit directly with "git checkout <commit hash>".

 

[Sepheyer]

Mr-Fox's prompt in wide-screen format:

Spoiler: More

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

 

[Mr-Fox]

Spoiler: Prompt

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

 

[Mr-Fox]

Mr-Fox's prompt in wide-screen format:
View attachment 2480043

Spoiler: More

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

Really nice. I love the patterned bikini. Did you use any tag for this or was it just by chance?

 

[Sepheyer]

Really nice. I love the patterned bikini. Did you use any tag for this or was it just by chance?

It is your prompt 100%. What's your DeviantArt handle? Imma tag you as the prompt's creator.

The file has all the prompt info.

Spoiler: Prompts

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

 

[Mr-Fox]

It is your prompt 100%. What's your DeviantArt handle? Imma tag you as the prompt's creator.

I don't have an active account, I only have one deactivated from the BE morphing days (Bobb54). It's enough if you only add me in the description. I have been toying with the idea of creating an account for MrFoxxx.

 

[Sepheyer]

"Image prompt by the talanted Mr-Fox."

You must be registered to see the links

Actual prompt link

"Retouched" with GFPGAN v1.3:

 

[Mr-Fox]

"Image prompt by the talanted Mr-Fox."

You must be registered to see the links

Aaw chucks, you're gonna make me blush. Thank you for the compliment.

 

[devilkkw]

BodyPaint:

Prompt:
(detailed), ((TYPE haired woman in a colorful intricate bodypaint)), sunset, grunge, oily, dirty, ripped, post-apocalyptic

TYPE=describe haire


Question: how upload image without losing png data for prompt?

 

[Mr-Fox]

Go to "Stable-Diffusion\stable-diffusion-webui\outputs\txt2img-images" there should be folders with the date that you generated the images and inside this folder you can find all the png files with the meta data.

 

[devilkkw]

Thank you, i know it, but upload here strip out extra metadata, but i saw some here have it inside. i used "attach files" function and added my output files with all info in. Is this mode correct for upload?

 

[Mr-Fox]

Thank you, i know it, but upload here strip out extra metadata, but i saw some here have it inside. i used "attach files" function and added my output files with all info in. Is this mode correct for upload?

Just press attach files and find the folder I told you and select the png image file. Then select full image, resize it to something reasonable like 480p. This only makes the image appear smaller so it doesn't take up the entire page. You can also drag the image from the folder and drop it in the post box.
We usually pick the best one and hide the rest with spoiler flag.
Or you can use 318p to fit 3 images in a row with a space between each.
If you want them bigger use 478p for 2 in each row with space between. I think that you pressed thumbnail instead of full image, just my suspicion.

 

[devilkkw]

Yes, i've used thumbnail , because i think it show thumb and get full data when you press it and open at full view.
So now i understand i have to use full image. thank you so much. sorry for OT.

 

[Mr-Fox]

Yes, i've used thumbnail , because i think it show thumb and get full data when you press it and open at full view.
So now i understand i have to use full image. thank you so much. sorry for OT.

No worries we are only glad to help here. Really awesome images btw. I especially like no1 and no4.

 

[devilkkw]

No worries we are only glad to help here. Really awesome images btw. I especially like no1 and no4.

Thank you. i'm glad you like it. Also i'm reading all page of this 3d, your work is good.
I'm unable to produce big image, my max is 1024x768 without hiresfix.
When i upscale, i use extra tab, and use some upscaler( i like ultra_fart-photo for realistic production).
but if you have some suggestion on better upscaler, i'd like to try. Also used most know Remacri, but for me tend to vanish some effect.

 

[Mr-Fox]

Thank you. i'm glad you like it. Also i'm reading all page of this 3d, your work is good.
I'm unable to produce big image, my max is 1024x768 without hiresfix.
When i upscale, i use extra tab, and use some upscaler( i like ultra_fart-photo for realistic production).
but if you have some suggestion on better upscaler, i'd like to try. Also used most know Remacri, but for me tend to vanish some effect.

Thank you. If your GPU can handle it, Hires Fix is superior to everything else. use either 512p or 768p and then use Hires mutliplier to do the rest. Start with 2 and if it works increase to 2.5 etc until you get cuda memory error. Then backtrack 0.1 until you don't get the error. Now you know your Hires limit. I use either

You must be registered to see the links

or

You must be registered to see the links

for upscaler.
For Ultrasharp choose to download zip and then unpack in Stable-Diffusion\stable-diffusion-webui\models\ESRGAN.
NMKD Superscale needs to be placed in this folder also. You need to restart webui you can do so in extensions/reload ui.
Read these posts:
https://f95zone.to/threads/stable-diffusion-prompt-sharing-and-learning-thread.146036/post-10333003 (big tutorial)
https://f95zone.to/threads/stable-diffusion-prompt-sharing-and-learning-thread.146036/post-10306143 (cuda error fix)

 

[Mr-Fox]

How dare you propose me to sleep with you ? View attachment 2480608

Grant stop posting full size images it's obnoxious! Also if you are going to post back to back, keep it in one post.
Resize your images to something reasonable like 480p. Go to page one and see the rules for this thread. No posting without sharing the prompt. About your images. I don't wish to kill your enthusiasm but I think you can do better. Stop using simple upscaling, it only makes the image big but removes texture and fidelity, makes everything smooth without detail.
Start exploring Hires Fix. If you don't have a GPU that can handle it then see how big you can make the image without getting cuda memory error and focus on composition and use Restore Faces and explore post processing. I have made posts about it.