- Mar 8, 2022
- 81
- 122
Disclaimer: As promised I want to show you how to use Stable Diffusion for generating all kinds of images (not only NSFW ones). It is supposed to be an intro to image generation. There are many forks of Stable Diffusion like Easy Diffusion but they are either not well maintained or behind a paywall. I explained a few technical things in my Easy Diffusion tutorial but I will give you a quick(er) rundown on what is what. (For the Easy Diffusion Tutorial klick here -> Easy Diffusion Tutorial)
What do you need:
1. Download Stable Diffusion and follow the instructions (choose the right version for your graphics card)
-> Nvidia GPUs:
-> AMD GPUs:
-> Intel GPUs:
2. Preferably a good graphics card and/or processor is very recommended. It makes the process (a lot) faster.
A FASTER CPU/GPU DOES NOT MAKE THE PICTURES BETTER OR HIGHER QUALITY!! But it does speed up the generation of images significantly.
3. Addons
-> ADetailer
-> Controlnet
4. UI Settings
Here are my UI Settings if you want to follow 1:1
"Settings/User interface"
-> In the Quicksettings list add those tags from the screenshot
5. Text2Img UI
This is the place where the magic happens.
Generating an image:
1. So lets start with setting up the generation
I will use these settings:
2. Let the fun beginn
Sidenote for Pony Checkpoints/Models. The prompts need the following tags in the positive text box "score_9, score_8_up, score_7_up, score_6_up, score_5_up, ". It is a Pony Checkpoint thing. You should always include that score tag when using this kind of Checkpoint. For more information look here ->
-> I am using an addon that is automatically inserting these tags. You should not forget to include those.
I admit I like redheads. So we will be creating a female with red hair, freckles and black lipstick. I want to give her a choker and a banger body. Maybe her stockings are torn as well. It would look something like this:
Very acceptable I might say. I used the tag "BREAK" to break up the prompt in more chunks. I did that because when giving parameters like colors or features Stable Diffusion tends to blend that into other things. Like when you want a redhead you would also get red nail polish or something like that. To make it happen less I tend to use BREAK at some places. We can also emphasize tags we want to stick out more or when a tag did persumably nothing. You can do that by either engulfing your tag with parenthesis or like this: (intricate details:1.2)
Every number above 1 will put more importance to the tag, just do not go overboard with it since it can make things worse. Still sometimes AI image generation does what AI image generation does and makes mistakes. Generate a few images unitl you get something you are satisfied with. Minor flaws can be fixed with inpainting.
Since our character has this kind of witch vibe, let us give her a witch hat and sit her on a log in the forest. It could also be night in the forest. She could also lean here chin on her arm with a mischievous grin:
Since this is a website for horny people we should spice things up and make it a little bit more lewd. Instead of sitting on a log she could be on all fours with a see through black dress, i forgot the chocker so lets add it now. I want to add some tattoos as well:
But I don't think it is lewd enough. The "nsfw" tag should just do the trick:
Okay, I like the pose, but what if she looked like a Disney Princess? I theoretically could wirte prompts and regenerate over and over again until I get what I want or I search for a LORA that has what I want. LORAS are like litte addons for a checkpoint/model. We will be using the
So, we made our witch but what if I told you we can also get poses from another picture and copy them to our character? The answer is Controlnet. The only issue is, that Controlnet currently only functions with SD 1.5 models. SDXL models are not supported at the moment. So lets change the model and Install the Controlnet addon. Furthermore you need some Controlnet models. You can get them here ->
We only need the openpose one. Grab it and put it into the "Stable Diffusion\webui\extensions\sd-webui-controlnet\models" folder.
Since we are not using an SDXL checkpoint anymore our MILF checkpoint will not work anymore. Some LORAs have to have either SDXL or SD checkpoints. Oh well, we can just revert back to the promt without it.
So my new settings are like this:
Checkpoint: Aniverse 3.0
Resolution: 768 x 1024 (since this is what the checkpoint creator suggests)
Not bad for a baseline. So lets find a picture from the internet that we can use to copy the pose out of. I chose one random from Google. So choose a picture with a pose you like and klick on "add image" in the Controlnet section. Enable Controlnet and Allow Preview to see the Pose it will create. If you downloaded the Openpose model and put it into the correct folder then Stable Diffusion should automatically set itself up. Klick on "Controlnet is more important" and then on the explosion icon. Then it should look something like this:
Now we can delete the "on all fours" prompt and regenerate the pictures:
As you can see it worked quite well. It would still need some tweaking, correcting and retrying but for a quick and dirty generation it worked quite well. For the little finale lets make her a little bit more NSFW and open up here clothes more....I need to go to sleep seriously haha
So thats it for now. If you have questions or want something specific feel free to ask in this thread. Have a good one!
Stable Diffusion 1.10.1 Tutorial
What do you need:
1. Download Stable Diffusion and follow the instructions (choose the right version for your graphics card)
-> Nvidia GPUs:
You must be registered to see the links
-> AMD GPUs:
You must be registered to see the links
-> Intel GPUs:
You must be registered to see the links
2. Preferably a good graphics card and/or processor is very recommended. It makes the process (a lot) faster.
A FASTER CPU/GPU DOES NOT MAKE THE PICTURES BETTER OR HIGHER QUALITY!! But it does speed up the generation of images significantly.
3. Addons
-> ADetailer
-> Controlnet
You don't have permission to view the spoiler content.
Log in or register now.
4. UI Settings
Here are my UI Settings if you want to follow 1:1
"Settings/User interface"
-> In the Quicksettings list add those tags from the screenshot
You don't have permission to view the spoiler content.
Log in or register now.
5. Text2Img UI
This is the place where the magic happens.
You don't have permission to view the spoiler content.
Log in or register now.
Generating an image:
1. So lets start with setting up the generation
I will use these settings:
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.
You don't have permission to view the spoiler content.
Log in or register now.
2. Let the fun beginn
Sidenote for Pony Checkpoints/Models. The prompts need the following tags in the positive text box "score_9, score_8_up, score_7_up, score_6_up, score_5_up, ". It is a Pony Checkpoint thing. You should always include that score tag when using this kind of Checkpoint. For more information look here ->
You must be registered to see the links
-> I am using an addon that is automatically inserting these tags. You should not forget to include those.
I admit I like redheads. So we will be creating a female with red hair, freckles and black lipstick. I want to give her a choker and a banger body. Maybe her stockings are torn as well. It would look something like this:
You don't have permission to view the spoiler content.
Log in or register now.
Very acceptable I might say. I used the tag "BREAK" to break up the prompt in more chunks. I did that because when giving parameters like colors or features Stable Diffusion tends to blend that into other things. Like when you want a redhead you would also get red nail polish or something like that. To make it happen less I tend to use BREAK at some places. We can also emphasize tags we want to stick out more or when a tag did persumably nothing. You can do that by either engulfing your tag with parenthesis or like this: (intricate details:1.2)
Every number above 1 will put more importance to the tag, just do not go overboard with it since it can make things worse. Still sometimes AI image generation does what AI image generation does and makes mistakes. Generate a few images unitl you get something you are satisfied with. Minor flaws can be fixed with inpainting.
Since our character has this kind of witch vibe, let us give her a witch hat and sit her on a log in the forest. It could also be night in the forest. She could also lean here chin on her arm with a mischievous grin:
You don't have permission to view the spoiler content.
Log in or register now.
Since this is a website for horny people we should spice things up and make it a little bit more lewd. Instead of sitting on a log she could be on all fours with a see through black dress, i forgot the chocker so lets add it now. I want to add some tattoos as well:
You don't have permission to view the spoiler content.
Log in or register now.
But I don't think it is lewd enough. The "nsfw" tag should just do the trick:
You don't have permission to view the spoiler content.
Log in or register now.
Okay, I like the pose, but what if she looked like a Disney Princess? I theoretically could wirte prompts and regenerate over and over again until I get what I want or I search for a LORA that has what I want. LORAS are like litte addons for a checkpoint/model. We will be using the
You must be registered to see the links
Lora to achieve exactly that. Since I like witches and redheads....witches are often more mature so lets try the LORA:You don't have permission to view the spoiler content.
Log in or register now.
So, we made our witch but what if I told you we can also get poses from another picture and copy them to our character? The answer is Controlnet. The only issue is, that Controlnet currently only functions with SD 1.5 models. SDXL models are not supported at the moment. So lets change the model and Install the Controlnet addon. Furthermore you need some Controlnet models. You can get them here ->
You must be registered to see the links
We only need the openpose one. Grab it and put it into the "Stable Diffusion\webui\extensions\sd-webui-controlnet\models" folder.
Since we are not using an SDXL checkpoint anymore our MILF checkpoint will not work anymore. Some LORAs have to have either SDXL or SD checkpoints. Oh well, we can just revert back to the promt without it.
So my new settings are like this:
Checkpoint: Aniverse 3.0
Resolution: 768 x 1024 (since this is what the checkpoint creator suggests)
You don't have permission to view the spoiler content.
Log in or register now.
Not bad for a baseline. So lets find a picture from the internet that we can use to copy the pose out of. I chose one random from Google. So choose a picture with a pose you like and klick on "add image" in the Controlnet section. Enable Controlnet and Allow Preview to see the Pose it will create. If you downloaded the Openpose model and put it into the correct folder then Stable Diffusion should automatically set itself up. Klick on "Controlnet is more important" and then on the explosion icon. Then it should look something like this:
You don't have permission to view the spoiler content.
Log in or register now.
Now we can delete the "on all fours" prompt and regenerate the pictures:
You don't have permission to view the spoiler content.
Log in or register now.
As you can see it worked quite well. It would still need some tweaking, correcting and retrying but for a quick and dirty generation it worked quite well. For the little finale lets make her a little bit more NSFW and open up here clothes more....I need to go to sleep seriously haha
You don't have permission to view the spoiler content.
Log in or register now.
So thats it for now. If you have questions or want something specific feel free to ask in this thread. Have a good one!
