ฝัง
- เผยแพร่เมื่อ 5 ก.ค. 2024
- #A1111 #stablediffusion #automatic1111 #AI #Lora #koyass #sd #civitai
This is Just a Beginners Guide, Don't watch if you know the basics of using SD and A1111, many people are still experimenting for the first time, and hopefully this video could be slightly simpler than what I have seen in other Channels and still useful for some... SD and A1111 contain hundreds of tools and thousands of techniques, but taking it step by step is necessary.
--------------
00:00:00 Introducing A1111 for stable diffusion
00:00:26 installation of A1111
00:03:12 start up parameters
00:04:10 A1111 GUI features and components
00:04:34 adding models and LoRAs from Civitai.com
00:06:49 installing new extensions in A1111
00:07:52 start testing image generation and txt2img details
sampling steps, CFG, hires, controlnet...etc
00:15:29 xyz plot explanation for comparing different images
00:16:25 img2img features and usage
00:18:31 inpainting in A1111
00:20:41 PNG file info of generated images
00:21:13 extras tab for resizing and extra processing
00:22:24 other tabs
00:23:37 changing GUI elements from settings such as clip skip
or contrlnets VAE and other GUI elements
-----------------------
Computer Specs:
Laptop: Legion 5 Pro
Processor :AMD Ryzen 7 5800H , 3201 Mhz
System RAM: 16.0 GB
Graphics GPU: NVIDIA GeForce RTX 3070 Laptop GPU 8GB
------------------
You can also watch some of my other videos that i have done before about outpainting, LoRA training, using Roop for faceswapping and more... in the future, I might do shorter videos about specific tools, and make more Videos about Comparisons
download Automatic1111 from its git hub page
github.com/AUTOMATIC1111/stab...
download models from civitai
civitai.com/
downlaod Controlnet v1.1 files from
huggingface.co/lllyasviel/Con...
Standard LoRA training
• Training a LoRA Model ...
using roop
• Faceswap using Roop in...
different methods for outpainting
• Outpaint using Control...
• simple method for outp...
• Outpaint in stable dif... - วิทยาศาสตร์และเทคโนโลยี
This is very helpful information! Thank you!
Glad it was helpful!
Thank you for your awesome tutorial! I have a small issue, though...
After installing extensions, my web ui didn't change to look anything like yours. It looks the same as it did before installing controlnet or roop. It also didn't start offering to finish words I typed in the prompt like it did for your at 7:46
Any idea what I did wrong?
You are welcome, I think you need to shutdown A1111 completely then restart then, Apply works sometimes, but not always, safest option is to restart .... the layout might look slightly different sometimes depending on the A1111 version installed which keeps updating every now and then as well.
Great Tutorial which helped me learn a bunch of things. Several questions:
I don't have all the icons under GENERATE that you show, nor do I have anything in the Styles drop down, and finally how do I get the black background on the page ?
Thanks !
You are welcome.
these icons used to show up in A1111 before the recent update, you can see how we add new styles/edit in this video th-cam.com/video/71HTtRSXsAc/w-d-xo.html at its end ... or watch the entire video to check the most important updates in recent version of A1111.
you can get sample styles that may be useful for example huggingface.co/datasets/AIHowto/460RegImages1024x1024/resolve/main/styles.csv you can download them and put them under stable-diffusion-webui folder and replace the existing styles file, or create styles of your own and as per the method instructed in the video earlier.
if you mean black page, this is the dark theme of Edge you can paste this in your browser edge://settings/?search=dark and choose dark option on the right, or search for dark and choose dark theme from browser settings, better for the eyes, and saves energy.
Many thanks for the tutorial. I already knew about setting it up (I did it from command line cmd/GIT), pulled a load of checkpoints and got familiar with text2img and img2img. Controlnet was new to me though, so I followed your guide and got working with Canny no probs. Amazing stuff :-) My only query really is - how did you get SD-VAE, Lora and Clip Skip to appear to the right after Stable Diffusion Checkpoint? I switch VAEs a lot and have to go into settings to change that every time, so it would be handy for me to have it on the top bar like in your video / interface. My A1111 version is 1.6. Thanks!
Great to hear, You are welcome.
Go to settings to the right.
click on User Interface on the left side column
scroll down to [info] Quicksettings list (setting entries that appear at the top of page rather than in settings tab) (requires Reload UI)
in the text edit box, look up (sd_vae) (clip_stop_at_last_layer) ...if you just type the first three letters it gives you the option, click enter and it will be added.
go up and reload the UI (or apply settings then reload UI)
@@AI-HowTo That's really helpful - thank you!
You are welcome.
thank you for your effort I follow all your lessons , And I have some questions , Why do I have none? openpose Why do I have no on the interface
in th-cam.com/video/RtjDswbSEEY/w-d-xo.html i display how we can install extensions, then restart A1111 afterwards, so if you installed control net for instance, it might take some time to get the models downloaded automatically, if they were not downloaded automatically and didnt yet see open pose for instance then you can go to the official models page on huggingface.co and download them from huggingface.co/lllyasviel/ControlNet-v1-1/tree/main copycontrol_v11p_sd15_openpose.pth file and they yaml file or any other control net model and save them to \stable-diffusion-webui\extensions\sd-webui-controlnet\models directory and restart A1111
I havent got the Lycoris folder is there a way to update the folder to newest version or someway to get everything up to date without reinstalling please and thanks :)
in recent updates of A1111, Lycon or standard LoRA no longer need separate folders, all can be put inside LoRA folder... and they can all be used with , it automatically detect the type.
@@AI-HowTo Hey I recently download something called Stability Matrix and it sets everything up for you compared to the manual way of setting this up how do people do it these days?
That is good, might be easier... I think people choose what feels ok for them, so there is no standard for what is the best for a specific use case, however thus far ComfyUI have proved to be the most resourceful tool for stable diffusion, but with longer learning curve, its gives more workflows and flexibility, still, what ever works stick with it, the basic concepts remain the same accross all Web UIs.
my webui keeps showing timer on checkpoint and all lora/embeddinsgs/etc and tries to load all again on every start is there a fix for that?
I have not seen this issue, before, but it could be due to low memory in your PC which forces the A1111 to load/unload models on multiple stages, not sure, I suggest to try github.com/lllyasviel/stable-diffusion-webui-forge which has same UI as A1111, but has better memory management and might run faster and automatically detect best memory settings that could run on your PC.
@@AI-HowTo thanks for reply it was working fine for months and broke out of nowhere will wait a week before fresh installing.
i see, you can try using
Git pull
(in the command line after
set COMMANDLINE_ARGS=
line as well, maybe that helps bring up any updates and fix the problem, if auto update is not enable in your installation
I keep getting connection error timeout. Can you help me solve it, please? :((((
Sorry, I have not encoutered this issue before, but some people have, and it seems that export COMMANDLINE_ARGS= --no-gradio-queue or set COMMANDLINE_ARGS= --no-gradio-queue ) worked for them, so in the webui-user.bat file, just right mouse click, edit, and add --no-gradio-queue to its end on the COMMANDLINE_ARGS= line, and hopefully that may help to solve this issue for you.
is this not possible on a windows 7 pc ?
I doubt it does, since Windows 7 support has ended a long time ago by Microsoft, so no one tests anything on it, so I double it works on Windows 7, but it might.... either ways Stable diffusion requires good Graphics cards specs, so if you PC is not of high spec, it is best to use Online Generation tools, especially currently that DALLE 3 is available for free by Microsoft Bing chat and can produce extremely complicated impressive results
@@AI-HowTo thanks for the info i tried to install it but i get stuck with python where stable diffusion will not start up , i had to use an older version of python because the newer ones will not work on windows 7 .
as an alternitive i already use leonardo AI .
Stick with what works best, Stable diffusion with A1111 is very sensitive, even slightest changes in python version/library version could result in errors unfortunately.
hey, do you need to pay money for this?
no all is free, runs on your PC, but your Graphics GPU must have good specs, otherwise, you can test on Google colab which offers the same software and also free, but limited in terms of amount of time you can use per week.
@@AI-HowTo yeah I got the error after Trading to install a1111:
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Does this mean that my laptop does hot have enough GPU/RAM capacity in order for a1111 to be installed?
not necessarily, Torch is often used by certain graphics cards, such as RTX... you should add the recommended args by your console in the webui-user.bat file (right mouse click and edit it) after set COMMANDLINE_ARGS= and try again, hopefully it words...if things didnt work out and your GPU is not powerful enough it is better for you to look up colab alternative.
@@AI-HowTo i did and it somehow worked. but the next issue is that the prompt that im using does not work. When i hit generate in a1111 nothing happens, im using this model: majicmixRealistic_v6
this is a good model, you should check the console for error messages, it should guide you to what is happening and what is wrong with your commands, check if adding --no-half-vae to the args at the web ui may help in some cases where NaNs appear and produce black images.