>>32072
Have a Guide:
1. Ensure your PC has a videocard. With AT LEAST 2gbs of VRAM.
2. Install python 3.10.6. Here is a link:
https://www.python.org/downloads/release/python-3106/
Download “windows installer”, run it. Don’t tick any boxes, BUT DO TICK THE “add Python to PATH” one. Preferably reload your pc after installing.
2. Go to
https://github.com/AUTOMATIC1111/stable-diffusion-webui
Find "releases" on the rightmost column of the screen, click on the latest one, scroll down and download “source code (zip)”.
3. Put it somewhere where you have a lot of storage, preferably - ssd. Unzip.
4. Find the “webui-user.bat” file. Not the “webui-user.sh” file. If you don’t see file extensions - look at the top of your file explorer, locate the “view” tab, and tick the “show extensions” checkbox.
5. Now important: click “edit” on the “webui-user.bat”. It should open as a text file. Depending on what system you are running, you might need to add some arguments to the “set COMMANDLINE_ARGS=” line. If you don’t see the line - you got the wrong file.
2gb-4gb VRAM:
set COMMANDLINE_ARGS= --lowvram --opt-split-attention --xformers
If it fails, leave only the --lowvram. Xformers only works if you have NVIDIGA videocard.
4gb - 8gb VRAM (I use this):
set COMMANDLINE_ARGS= --medvram --opt-split-attention --xformers
8gb+ VRAM (I use this):
set COMMANDLINE_ARGS=--opt-split-attention (--xformers/--opt-sdp-no-mem-attention/--opt-sdp-attention)- ONLY USE ONE OF THESE. The latter two eat more VRAM, but are faster.
6. Run the file you just edited. Give it some time to load. It`ll download a shitton of things when loading for the first time, if your internet is shit - may take up to an hour. DO NOT CLOSE THE COMMAND LINE WINDOW, when it eventually finishes - you will see something like this as the final message:
local URL:
http://127.0.0.1:7860
To create a public link, set `share=True` in `launch()`.
<Some loading metrics, ignore these>
If your internet is not shit, go to the next step as it`s loading.
7. Go to Civitai. Find a model you like. Pay attention to the “Base Model” in characteristics. SD 1.5 works best for diaperfur stuff, usually. feel free to experiment with 2.0. ALWAYS READ DESCRIPTIONS, some models require addons to work or use different install methods. recommend starting with the most basic furry model out there: Fluffyrock:
https://civitai.com/models/46779/fluffyrock-e17-ys-mod
It has diaperfur stuff in training data. Download it, and put the “something.safetensors” file into models -> Stable-diffusion folder. They are pretty big, 3-6gb in size. Always go for “pruned” ones, if you see the option - they are compressed, meaning they are smaller, and, allegedly, faster(?)
8. LoRAs help you generate stuff that model hasn’t been trained on. For example, a very good furry model:
https://civitai.com/models/11888/crosskemonofurrymodelandhumanmodel
Can’t generate diapers, because it was trained on a dataset with no diapers. You can fix it with LoRAs.
You browse and download LoRAs same way you do models, except they are considerably smaller, and go into “Models -> Lora” folder.
Best diaper loras for furry stuff:
Best wet diapers:
https://civitai.com/models/83342/om
DiaperMixLoRA:
https://huggingface.co/toasterroasterp1/ABDL_Pack/tree/main
Diapermix was deleted from civitai - IDK why. They work wonderfully together.
9. Textual Inversions are like LoRAs, but not like Lora’s. They go into “embeddigs” folder (note, it’s in the stable-diffusion-webui model, same level as “models”). Most popular one is easynegative:
[Expand Post]
https://civitai.com/models/7808/easynegative
It just makes images better.
10. When your command line finished loading - go to the URL it provided. Should be something like: http://127.0.0.1:7860. It’s your localhost address - it’s only available to your computer.
11. Pic related.