Updated Stable Diffusion A1111 2023 Installation Tutorial V1.5, V2.1, and Custom Models

SDXL is out and the only thing you will do differently is put the SDXL Base mode v1.0 and Refiner Model v1.0 into your model's folder the same as you would with any model (i.e. that's where your 1.5 and 2.1 model went) and then generate txt2img using the base model. To use the refiner send your generated image to img2img, choose the refiner model, match the resolution to the txt2img pic, and then set a low denoise (0.1-0.3) and generate.
Here are thew new SDXL models.

This is an updated guide for Installing Stable Diffusion Automatic 1111 (A1111) v1.5, v2.1, and Custom models. Some things have changed recently and installing will slightly easier now we don't have to add yaml config files manually.

Things you will install in order:
Step 1. Install Python 3.10.6 & GIT (do not download updated versions of python)

Step 2. Download A1111 repository & models

Choose pip for package & 11.8 for CUDA

Some useful ARGS
--xformers
--medvram
--no-half
--precision full

==========================
Chapters
00:00 Intro: What has changed & what will be in this video
00:36 PreReqs
00:55 Install Python 3.10.6, GIT, & optional CUDA
02:13 Download Automatic 1111 GitHub Repository to your PC
03:30 Download V1.5, 2.1, & Custom Model & What's EMA vs. Non EMA and Safetensor vs. CKPT
05:40 Bring your Models into Stable diffusion/Models/Stable-Diffusion Folder
09:23 How do you run this thing and generate images?
09:50 Txt2Img Tab the Main tab to create images overview
11:38 Example of Restore Faces + Hires Fix
12:04 CFG Scale
12:28 Seeds
12:49 Img2img
13:05 Denoising
13:43 Inpainting
14:12 What is CLIP?
14:42 Extras Tab (Upscaling and GFPGAN) & Example

Seems like I didn't go over upscaling in the extras tab. It increases the quality and resolution of your base image, just throw it in there and select the number of times you want to increase the resolution. I personally do a latent upscale instead which uses Img2Img w/ denoising under 0.40 and I double the resolution. It will recreate some of the finer details and make the picture look even better.

#Automatic1111 #A1111 #StableDiffusion #StableDiffusiontutorial #stablediffusionaitutorial #aiart #aiartcommunity

[mikesrandomchannel]

The production on your videos is just top-notch. Great video making and great tutorials.

[simonopen]

I can't believe i finally found a tutorial that worked. awesome and thank you.

[wizards-themagicalconcert]

Fantastic video and content mate, keep it up ! Subbed !

[alanfordyce]

Would you use this update installation, to install on the google remote desktop?

[WillowbarkPetPhotography]

Great tutorial! Do you happen to know if Stable Diffusion is private, or if they take our photos (generated or not) to do what they wish with it?

[andrewkarn]

does this have to be on your c drive? trying to install to an external ssd

[cashyappurushottam]

Hey thanks for this! So i was having problem with my previous version of A111...getting error that said..

CUDA out of memory. Tried to allocate 90.00 MiB (GPU 0; 4.00 GiB total capacity; 3.29 GiB already allocated; 0 bytes free; 3.40 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF'. Check your schedules/ init values please. Also make sure you don't have a backwards slash in any of your PATHs - use / instead of . Full error message is in your terminal/ cli....

So i was frustrated looking at Discord. And then found your Video, and so much thankful to you for putting this up Fresh and Neat!
.
Just one question, i have a NVIDIA RTX 3050 with 4gb Vram. will downloading the CUDA driver you said at 1:58 help me ??

[ankurchaudhary]

Stable diffusion model failed to load
Applying optimization: sdp-no-mem... done.

[harryhealy]

stuck on model loaded, isn't going past that on the batch Model loaded in 13.1s (calculate hash: 8.8s, load weights from disk: 0.2s,

[ArtificallyIntelligentAi]

SDXL is out and the only thing you will do differently is put the SDXL Base model v1.0 and Refiner Model v1.0 into your model's folder the same as you would with any model (i.e. that's where your 1.5 and 2.1 model went) and then generate txt2img using the base model with at least 1024 resolution. To use the refiner send your generated image to img2img, choose the refiner model, match the resolution to the txt2img pic, and then set a low denoise (0.1-0.3) and generate.
Here are thew new SDXL models.

By the way SDXL doesn't play nice with some extensions and doesn't work with LoRAs or embeddings right now as those were trained with previous models.

Stable Diffusion no longer requires Yaml files to be installed manually. Here is an updated tutorial for 2023 on how to install Stable Diffusion v1.5, v2.1, and custom models from CIVITAI and also how to edit your Webui-user.bat file with Diff ARGS to utilize xformers or stop v2.1 from running out of VRAM and rendering black/blank images.😇🍒🌶

[blackdubs]

thanks for the video my man, new sub! could you make a video for Dreamboth? i have some issues with model training (Exception training model: 'No executable batch size found, reached zero.')

[Firebird]

I did all of this and mine gets to: Model loaded in 6.8s (load weights from disk: 2.2s, find config: 1.1s, create model: 0.2s, apply weights to model: 0.6s, apply half(): 0.7s, move model to device: 0.9s, load textual inversion embeddings: 1.0s). And then it just does nothing afterwards. No URL or anything. Big sadge 😢

[maitotechlab]

Well now i know why it didnt work yesterday lol, i installed a1111 and at some point it said dowloading something but nothing happened, guess i need to install puthon and the git something lol

[christopherniv]

i just download SD but a dont have the lora model file do you know why '

[vijaykrishna]

Mine always stops at downloading pytorch which is a 2.6 GB and stucks there, what should I do

[chreydylan]

I had a lot trouble, I fixed all of them except this one "ImportError: DLL load failed while importing cv2: The specified module could not be found."
Please help me if someone knows how to fix this. I spends hours looking for answers but none of them work.

[AvdolBeastar]

When i run the git clone and copy paste it says, unable to access howndo i fix it

[EllenVaman]

Hi, I did follow instructions and dowloaded all these files and started using automatic 1111 (my main reason was really to be able to use deforum) so I downloaded deforum into my automatic 1111 also - but not sure why I cant upload images or videos directly from my pc .. I tried clicking on the folder thumbnail but it does not work, I can get image links from online sources and paste into the image section and it works but cant upload my own images - any advice please ? thanks in advance :)

[christopherniv]

Thanks for the video. When i do edit for installing xformer i don't see the same thing. You have like 4 line i have a loot of them. I think i got something wrong

[DeadlyDuel]

ModuleNotFoundError: No module named 'font_roboto' error please fix.