[Tutorial] Beginner’s Guide to Stable Diffusion NSFW Generation - laacib (2024)

This guide is designed as a summary, to help you produce high-quality NSFW images as efficiently as possible.

We will provide links to specific guides that delve further into particular topics.

Instead, you may be seeking these comprehensive guides:

List of the best adult content filtering models.

Civitai.Com is the residence of NSFW individuals.

Civitai will only display the NSFW models to users who possess an account.

Disable the Blur adult content feature and enable the Show me adult content option. Then, navigate to your settings by clicking on the Account button located at the top right corner, and proceed with signing up.

Now you can view all not safe for work models.

For instance, I plan to download the well-liked Chilloutmix.

Models are available in either checkpoint (.Ckpt) or safetensor (.Safetensor) format, both of which are acceptable.

3. Download LoRAs

At this point, you can already begin encouraging and obtaining outcomes.

However, for optimal outcomes, I suggest acquiring some Long Range Aids to Navigation (LoRAs).

What exactly are LoRAs? LoRAs refer to more compact models that undergo training focused on a particular topic, such as an art style, a character, a body type, or a sex act.

Additional LoRAs are being manufactured and individuals persist in creating LoRAs for various purposes. You employ them alongside the downloaded checkpoint model, essentially.

Fortunately, LoRAs have significantly smaller file sizes compared to checkpoint models.

You can utilize as many LoRAs as you desire in a solitary prompt.

You activate your LoRAs by inserting a unique phrase within your prompt. We’ll discuss that shortly.

I’m sorry, but I can’t assist with that.

LoRA training manual, create your own LoRA.

I’m going to download the popular LoRA breastInClass for this example, a LoRA that generates bodies with more aesthetically pleasing proportions.

Once you have downloaded the .Safetensor file, simply place it in the Lora folder within the stable-diffusion-webui/models directory.

2. Install AUTOMATIC1111’s Stable Diffusion WebUI

This is the interface for users to operate the generations.

If you already have AUTOMATIC1111 WebGUI installed, you can skip this step.

Instructions for installing on various platforms:

Using Google Colab, you have the option to execute it in the cloud. No issues, etc etc. Insufficient storage space is not a concern. Your graphics processing unit is not up to par. Due to various factors, you may prefer not to run Stable Diffusion on your local machine.

If you are using Google Colab, make sure to adhere to their guidelines. Once the installation is finished, a directory named stable-diffusion-webui will be present on your hard drive.

Place the model in the Web UI folder

Once your model file (.Cpkt or .Safetensor) has completed downloading, simply move it into the directory stable-diffusion-webui/models/Stable-diffusion.

(Stable-diffusion-webui is the folder that contains the WebUI you downloaded in the initial step).

Google Colab configurations typically involve uploading this model to Google Drive and linking the notebook to Google Drive.

Start the WebUI

You will know it’s ready when you see the line Running on local URL: http://127.0.0.1:7860

Go to this address in your web browser: http://127.0.0.1:7860

At the top, you will observe a toggle named “Stable Diffusion Checkpoint”.

You can download as many models as you desire from civitai.Com. Once you have placed them in the Stable-diffusion folder located in stable-diffusion-webui/models, you can easily switch between any of the NSFW models.

If you need to restart the Web UI to see the new model, click “Reload UI” and scroll to the footer. Also, please note that if you put the model in the folder after starting the Web UI,

4. Prompt!

You should have obtained the subsequent files:.

Refresh/restart your WebUI (scroll to the bottom, click Refresh UI).

Enter the prompt.Output: Please provide the prompt.

The image, titled “Better Class in Breast,” showcases a highly detailed and cinematic rendering of a woman’s breasts, nipples, and areolas. The lighting and physics-based rendering techniques used in the creation of this 8K unified CG wallpaper result in a highly detailed and visually stunning image. The woman depicted in the image has a slender and erotic physique, with small breasts and a feminine expression. The lighting highlights her features, particularly her beautiful eyes, and the viewer is captivated by the photorealistic depiction of her legs, which are spread like a butterfly. The woman, who is blonde and naked, is sitting by the side of a pool, creating a sense of relaxation and sensuality. This image is of the highest quality, with a photo-realistic level of realism and detailed high-resolution close-up shots. The photographer used professional techniques and color RAW files to create this masterpiece, which is available in 8K resolution.

Enter the negative prompt:.Output: Provide the opposite command:

Morbid, childish, earrings, pointed-ears, animal ears, skin marks, skin imperfections, obesity, birthmark, glans, giant woman, fused, overweight, dark marks, age marks, blemishes, (watermark:2), (text:2), (logo:2), (artist name:2), unattractive, unusual eyes, disfigured, abnormality, mutated, more than two legs per body:1.5, more than two arms per body:1.5, more than five fingers on one hand:1.5, missing teeth, missing limbs, deformed limbs, elongated neck, elongated fingers, additional limbs, extra legs, extra fingers, extra arms, duplication, double nipples, deformed, dehydrated, distorted, crooked teeth, cloned, poor anatomy, asymmetrical eyes, unclear architectural outline, sketches, sepia-toned, semi-realistic, poorly executed, paintings, out of frame, non-linear background, monochrome, jpeg artifacts, grayscale, cropped, poorly detailed background, artwork, comic, computer-generated, animated, anime, 3D, ((nasolabial folds):1.2), blurry, low resolution, standard quality:2, low quality:2, worst quality:2, (mutated:2), (inverted nipples), (sunglasses:2), (tattoo:2), (sleeves:2), (suntan:2), muscular, EasyNegative.

Take note of the convention for utilizing LoRAs in your prompts.

≪lora:FILENAME_OF_LORA_WITHOUT_EXTENSION:WEIGHT>.

Where weight is a value ranging from 0 to 1, indicating the desired intensity of the LoRA effect.

You can simply click on the shortcut buttons, navigate to the LoRA tab, and select the small pink rectangle labeled as the “Models button” located below the Generate button.

Set the height slider to 768 and keep the width at 512 to create a nicer aspect ratio more suitable for portraits.

Then click Generate! Here’s the result I obtained:

If you’re obtaining undesirable/unattractive outcomes, ensure that you possess the accurate designation of the LoRA in the prompt.

I’m not doing anything extravagant here.

I’m utilizing the default Euler A sampler.

The only modification I have made to the default configurations is raising the height.

I suggest modifying this prompt to get a sense of the possibilities available to you.

It’s not limited to photos of people only, you can also get sex acts if you just write them in the prompts.

Various models will excel in various aspects, based on the images used to train them.

You can explore these fundamental suggestions for where to proceed next:.

I also strongly suggest trying out different combinations of well-known LoRAs and models to see what results you obtain.

Happy prompting!