How to Install and Use Stable Diffusion (June 2023) - automatic1111 Tutorial
Vložit
- čas přidán 19. 06. 2024
- To try everything Brilliant has to offer-free-for a full 30 days, visit brilliant.org/AlbertBozesan/ . The first 200 of you will get 20% off Brilliant’s annual premium subscription.
LINKS
Python: www.python.org/downloads/rele...
Git: git-scm.com/download/win
WebUI .git repository (git clone this): github.com/AUTOMATIC1111/stab...
Cyberrealistic Model: civitai.com/models/15003?mode...
ControlNet Models: huggingface.co/lllyasviel/Con...
ControlNet Extension (for manual installation): github.com/Mikubill/sd-webui-...
CHAPTERS
-------
0:00 - Intro
1:06 - Requirements
1:43 - Installing the UI
3:05 - Finding and Installing Models
4:51 - txt2img Prompting and Settings
9:19 - Brilliant Sponsor Segment
10:32 - ControlNet
14:48 - img2img
15:45 - Inpainting
17:21 - Outro
----------------------------------------------
Did you like this vid? Like & Subscribe to this Channel!
Follow me on Twitter: / albertbozesan
This video was sponsored by Brilliant.
To try everything Brilliant has to offer-free-for a full 30 days, visit brilliant.org/AlbertBozesan. The first 200 of you will get 20% off Brilliant’s annual premium subscription.
👋
@@LouisGedo 👋
Albert I wish you did more content on stable diffusion. This tutorial is great
Your tutorial is an outstanding crash course, thanks so much!!
Awesome breakdown. Thank you.
Got the link from a friend to your CZcams channel. Excellent tutorial how everything works & starts! Working links and top information. Very well done, many thanks!
Thank you! And thanks to your friend :)
This is a fantastic tutorial!! I hope you do some more with Stable diffusion.
More to come!
Dude thank you I had so much trouble figuring it out in other tutorials, this one is up to date clear and clean
Thank you! Glad to hear that 😄
You think you had trouble I can't even follow the tutorial
thanks for this...by far this is the most concise and easy to follow installation for SD and first time trying it.
Glad it helped!
Encounterd a few errors with Xformers, fixed them no problem. Tutorial was great, thanks!
very nice tips thanks
I learned a lot. Thanks
JUST WHAT WE NEED THANKS
WHAT a great guy help me out big time that’s a big thumps up from me!!!
Thank you for your very clear and straight forward tutorial.
You’re very welcome!
Thank you! This will be an adventure♥
Indeed! Hope you enjoy :)
Somehow I downloaded this perfectly with your help. You are awesome. I'm a noob and it took 3 hours but it's working perfectly.
Thank you! Super glad it worked for you 😄
This helped me get started thanks! had to use 3.10 tho - 3.12 had errors
You're a natural education Albert, thanks for a fantastic tutorial. Subscribed!
Awesome, thank you!
very helpful and such a good quality video
Thank you so much!
That was very helpful, Thank you 👍👍
You're welcome!
this video helped me a lot I can't thank you enough .much love from iran
Very glad to hear that! You’re welcome.
Awesome!
Thanks for pointing out the bias bit! =D
It has to be said!
hands down, you're the best!!!
Thank you so much! I hope I can get some new vids up soon, have been very busy at work.
thank you so much for this tutorial!
You are so welcome!
I've been in entertainment and content creation for 27 years and OMG your voice! 12 seconds in you've already got a sub. 🤣 I'll begin appreciating the content as soon as I can get past imagining you doing our next VO. ...okay watched more and you've got spectacular content also... Well done.👊🤙
Haha thanks so much! Hit me up, I do voiceovers ;)
Had few issues of it not downloading something unlike the video, still I simply copied the file link in cmd and manually downloading it worked like a charm. Great video was super simple to fallow. I think if that does not break them I might rename model names to keep track of them easier and what they do. The names can get odd :D
your a blessing thank you so much
What a kind thing to say! Thank you and you’re welcome!
you're a lifesaver thanks
Happy to help
Very helpful tutorial!
Glad you think so!
stable diffusion is the most fun I've had in awhile.. a steep learning curve but great hobby. As a beginner, the most confusing part is trying to figure out what files go where and why... such as how do you know when a model requires an additional download that goes in another folder.. what is the purpose of a yaml file, etc. Someone could get a lot of views if they could go in depth on demystifying SD 🙂
Just wanted to say that I'm enjoying your videos, particularly the one with the 1 hour challenge/battle with the other youtuber.. is there a word for the german word "mitfiebern" in english? Shared the stress with you ;)
Two tips I found that may be interesting:
- Under settings/default you can set your own standard values for all the tabs in Automatic1111
- RightClick on Generate can generate forever. Rightclick again to cancel.
Have a nice day.
Thank you! Yes, there are loads of hidden goodies I couldn’t cover in this short vid 😄 Mitfiebern is the perfect word 😉
You are golden!
Thank you! Glad to help.
Nice, now I know how to use controlnet
Great! This other vid of mine goes even more in-depth :) czcams.com/video/dLM2Gz7GR44/video.html
Super👑🎸✨
hello, thanks for video! please tell me if it is possible to install the program on MacOS in this way?
Hi - thanks for the video. Can you do a tutorial on how to install on a Mac? Or is it not good enough yet on mac's ? or are there any recommendable videos for Mac installment ?
You can check out DiffusionBee. It’s very slow and has almost no features, but works.
Hi, I have a question. My local host is not working now. It was working fine earlier when installed. How to tweak it?
Awesome. Thanks for the tutorial. Now I can see what all the screaming's about.
Glad to help!
wow!
Albert, this is a great video, thanks. I wanted to mention an issue I have in the controlnet section-- I am on version Controlnet 1.1.227, and where you have 'Unit 0', 'Unit 1', 'Unit 2' tabs, I do not have them, there are no tabs for me. I can only add one control net module. How do I fix?
You can add modules in the Settings tab under ControlNet :) it’s pretty low on the page I think
Nevermind I fixed it. I had to go to the Settings -> ControlNet -> "Max multi controls modules" and change the default value from '3' to '4' and then saved/restarted and they showed up. Must be a defect
@@Derekmeetsworld Thanks, I needed that instruction!
This is insanely powerful... Hahahahahaha, I love it!
Thanks!
I am quite new to height and displacement maps of photos and the different software options for doing so (Blender, crazybump, zbrush etc). Then I saw your video and it seems the software Depth-Anything is more impressive. Apologies for simple question - but does the depth map created by this software need some level of postprocessing to be useful for creating CNC carvings or ornaments in other programs like blender (e.g., do I need to first grey-scale the Depth-Anything output etc). Any information would be appreciated...Thank you.
That question is too in-depth (ha-ha) for me to answer with any confidence. The depth extension in Auto1111 works very well for producing depth maps that can work directly in Blender without any further edits.
its been working so far, but i keep getting a warning along the lines of "CUDA error: the launch timed out and was terminated" when my image is almost done. does anyone know how to mend this?
Nice video was easy to follow. I got a different control net version for some reason and I couldn't figure out to have multiple control net units visible? It just had the one , is there a trick to that?
The UIs change pretty regularly, so if it just looks a little different don’t worry. Just make sure it’s the latest version. If you only see one unit, go to settings > ControlNet > multi ControlNet and turn it up to as many as you want :) you’ll never need more than three, in my experience.
@@albertbozesan that worked thanks
is there any problem on Python 3.11? I done all task get error 128 on run bat and tried many solutions not working
I didn't pass level 3 (running the webUI). I get:
Couldn't launch python
exit code: 9009
stderr:
Python was not found; run without arguments to install from the Microsoft Store, or disable this shortcut from Settings > Manage App Execution Aliases.
When you have it locally, are the images you create private? Or available to everyone?
They are on your local drive only.
You got a sponsorship, congrats!
Thank you! They’ve been a great support 😄
Hello. Thanks for your tutorial. I would like to know how to add new control units.?
In the settings tab under ControlNet is a slider for multi controlnet :)
why does it give me a grey box when i load a prompt? i tried to fix it and it kept giving me the same stuff :/
when I open the batch file, the prompt opens and tells me to creating venv in directory
hi do you know how can i use to replicate a render image increasing the realistic effect?
Yes, I suggest using img2img on the render with a medium denoiser scale, and ControlNet canny and/or depth, again with the rendered image. Try out different settings. That should be a decent workflow for what you’re trying to do!
Hello, I encountered an error while trying to run the webui-user file.
Kindly assist
is the process the same for adding lora files with the .safetensors extension?
Yeah, just place them in models > LORA
I don't have units in my ControlNet v1.1.313 ? Did they change something ? I got RTX 3080. Also my "Allow Preview" button makes ERROR (something like AttributeError: 'NoneType' object has no attribute 'encode') but when I push generate seems to work but It would be cool to have those previews + units ?
I’ve gotten more error comments than usual in the past 24 hours or so. It seems there’s a bug which should be cleared up soonish. I’m not a dev so keep an eye out on GitHub!
If I could, I would give you thousands of likes.
i have an error, it says
AttributeError: module 'os' has no attribute 'statvfs'
and my ai wont start, please help
Great tutorial!
For inpainting its not working, doesnt do anything, but i dont have a specific model for inpainting, what do we do if there is not an inpainting version for that model? It cannot work?
Inpainting should work with a normal model, just not as well as it could. There are many reasons it may not be working for you - have you searched for your issue on Reddit?
@@albertbozesan yeah it only blurrs the selected area or it doesnt change anything
Is there a model or a way to create sprite animation sheets for video games in Stable Diffusion?
I haven’t seen any that work super well yet.
@@albertbozesan thanks. That would be game changing.
Hey Albert, great tutorial. I ran in to a problem after typing the sd_vae and applying the setting it just were reloading a long time. I refreshed the tab and now my UI wont open at all. Do you happen to know what I should do?
I don’t know and can’t offer support via comment, unfortunately. Too many factors involved. Can you check Reddit?
Did you get it fixed?
same shit here
Using chair and sofa images, I have finetuned a stable diffusion model. However, the model does not generate a living room using the chair and sofa.
Please help me !!!!
Ask on Reddit, the community will always be quicker to respond than just me :)
Hi thanks alot for the tutorial. I was following up and installing it but I stupidly presses Apply (10:38) and restarted on Installed before going to available to load. It reloaded and now I cant access the website anymore. I keep getting this error code ERR_CONNECTION_REFUSED. I deleted my cache and DNS and so on but I still cant access the website. What can I do now?
I can’t solve very specific errors like that, can you check on Reddit?
it gives me an error i dont know what to do
NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
Please consider the tips I gave at the beginning of the video. This is not the right place for tech support.
hi i have error "Couldn't launch python
exit code: 9009" i add python.exe to PATH and i have 3.10.6 version
can someone help me?
The folder do you make in folder or is it download a folder for it?
What folder specifically? The stable-diffusion-webui folder and the folders inside all happen automatically.
Do you have to install models before you do txt2img prompting?
Yes, they are essential.
Thank you!@@albertbozesan
Hello, do SDXL models work with this?
Yes they do. They’re slower than in ComfyUI for example, but function :)
I don't have webui.user.bat file in my folder. It has only webui.user file.
What should I do ?
It might niot say .bat, it might just say webui but then check the file type is a Batch File
Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check how can i solve this
after i finish message appear press any key to continue then i press command window dis appear
Look up your error on Google and Reddit. Many people have had it before you.
Very nice but you forgot a little detail in compare SD with MJ. What about Mac ??
Check out DiffusionBee and several other ways of using SD on Mac
Im a concept designer, I predominately use 3D, does anyone know of a workflow to pass 3D renders (really just images I guess) through Ai prompts? in attempts to cutdown matte painting.
I'm interested in the 3D to AI workflow as well.
ControlNet or Img2Img with ControlNet should help. Or am I misunderstanding?
Awsome! video I am on a M1 Macbook pro with 10 cores, 16 gig ram, and M1-Pro 16 cores. Do you think that is enough?
While there are some attempts at getting SD to work on Macs, it continues to be incredibly slow. An NVIDIA RTX GPU and Windows is the way to go for now.
@@albertbozesan Thank you for the response. I just kind of need it for creating a proof of concept, but you would suggest me to get an PC with nvidia card? :P
@@andreer89 you can try Diffusionbee or one of the paid online Stable Diffusion services alternatively - I get that you don’t want to buy a whole computer for just a few tests 😄
Is it necessary to have an external GPU for installing Automatic1111?
I would strongly recommend it. CPU will take many minutes to generate one image, vs. seconds for a decent GPU.
says launch unsuccessful when i run the webui-user.bat file
do you have a tutorial video on how to download and use stable diffusion on a Mac?
Unfortunately no
2:55 I can't move forward from this part. When installing the remaining components, it stops and presents a "RuntimeError: Torch is not able to use GPU..." and then finishes. No address appears.
Please search for that error on Reddit, it’s my favorite way to solve issues because someone else has always experienced and figured it out before me 😄
when i run webui-user.bat it says I dont have python, also there wasnt an option to add python to path
try uninstalling python (using the installer and reinstalling it, there should be an option to add it to path.
i downladed vladmatic becouse i understood its updating allways ,what do you think?should i download automatic1111 instad? whats the diffrence?
Vladmantic doesn’t support all extensions, because they are still being made for auto1111 first. But keep an eye on it, it’s a good alternative. Can just be frustrating when things don’t work for beginners.
@@albertbozesan Thank you Albert, I understand now
How much space is the git clone?
Be prepared to download many gigs of models etc. it’s a little addictive. I have around 200 gigs of SD stuff.
any idea how to solve this?
statvfs = wrap(os.statvfs)
AttributeError: module 'os' has no attribute 'statvfs'
No, please refer to the resources I mention at the start of the video for support.
Does xformers automatically install?
Hi Albert, I'm looking for an AI image generator, precisely a text-to-image generator. In this video, you showcased a solution that allows prompts to generate images locally on my computer (exclusively, I presume?) for personal use without sharing them with the AI community, correct?
I plan to use the images in my novellas to ensure they remain unique to my work. Running the text-to-image generator locally on my computer will maintain its proprietary nature (for my exclusive use), correct?
Also, you mentioned, Free software. Does that also mean {No Watermarks}? And I can save, download for retention and use, correct?
Yes. Barring some pending questions regarding copyright (is AI art public domain or not), that’s all correct.
Mmmm it is a complicated question. Yes, they would be free of a rights infringement, unlike the cloud service route, where the service provider is facilitating their creation. The nature of an open source licence tends to cascade downwards, so 'exclusive use' may be hard / impossible to enforce. While there is no visible watermark, there is one embedded within the meta data. You'll have to do more research than ask a youtuber. Short answer is yes, you can use them, but try and bring a legal case against someone who used your images and your legal chances are slim to none
@albertbozesan all correct? The exclusive use part is not correct, considering the open source licence of SD. There is a little more to it. The op needs to do some research and take some extra steps if they want exclusive use.
@@LairdJ56 did you read my comment fully, regarding the public domain question?
@@albertbozesan yes I did. Thank you for your comments and clarification. I think it’s best that I continue to look for a graphics artist to design my descriptive photo pages in my Novelas for consistency and proprietary use.
Great tutorial !! Everything was great until I opened the webui and was hit with "Runtime Error: Couldn't install torch" any possible way to fix this issue?
Check for that specific error on Reddit. I can’t offer tech support here. Glad you enjoyed the tut!
Did you ever figure out the problem? I'm having this issue right now and having a hard time finding a solution 😭
please i have lab 12 ram and gpu nividia 2 giga i need to work on it can u help me please
2gb sounds like too little for local SD, sorry. Check out some online solutions like Rundiffusion.
I followed the steps and when I go to open the web user file I get "Could not find a version that satisfies the requirement torch==2.0.1 (from versions: none)" any idea what I did wrong?
Torch is my nemesis…could be a few different reasons. Have you searched for your error online?
I figured it out today finally, I installed newest version of python and it screwed something up, make sure you have 3.10.6 installed, thank you Albert @@albertbozesan
When trying to run the batch file, I get a "Couldn't launch python" error, although python is installed (3.11.4)
EDIT: I had to manually change the .bat file to include the path of the python.exe file
how tf do you do that? lol @@spyders03
hi for some reason i don't have the option to fix the faces
I strongly recommend Reddit for any tech support questions, it’s the quickest way to find answers.
noob question, so without a model, does it mean that stable diffusion won't work? how can i understand what a model is? can i take it as a trained algorithm based off of a feed of similar images?
I think of models as the AI 'brain', which is taking your prompts/settings and translating them in to images it knows how to create. Different AI 'brains' may be 'trained' in different styles, photographic techniques or subject matter. For example, some are better are rendering Realistic images of people, others excel at illustrations, anime, etc. Some can be better at clothing/armor/inanimate objects, or backgrounds/landscapes. While some are just general good all-rounders. And of course some users have trained models good at.... anatomically correct people in exacting detail... ahem.
Without the AI 'brain' you simply can't get any images at all. It's the engine that drives the car, so to speak. But the real fun begins when you learn how to Merge different AI Brains together in to entirely new models of your own creation. Take a model that's really good at realistic people, merged with a model that is good at Fantasy/Sci-Fi imagery and you can get some fun, unique and surprising results.
@@mkaleborn OK thank you.👍
Hey thanks for this video. Now my question is... How can I delete totally the SD from my system? is there any tutorial? I want to start from 0
Removing SD itself is straightforward - just delete the stable-diffusion-webui folder. Python can be removed via the “Add or Remove Programs” tool in windows.
@@albertbozesan thanks!!! ✌️✌️✌️
whenever I'm trying to get into the UI this error pops up "raise RuntimeError"
Have you checked the subreddit? I’m sure someone else has solved it - I’m not involved in the dev so I can’t answer tech questions 😅
can you please make a tutorial for mac
Unfortunately not, that’s too far from my core competencies and really not worth it imo…
Can i use with only CPU power? I have Nvidia 1650 4gb but Ryzen 9 with 32gb ddr4 Ram
tested. worked on my low end GPU
That’s great to hear, thanks for the info!
How will you be able to use it in IPad?
You have to run it on a PC or Mac. But you could open up a port on your PC and access the UI via your iPad browser.
I have 16gb ram but 13.8gb is usable.can I run automatic 1111?
It’s barely about RAM, mostly about VRAM. 16 gigs should be good for either.
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check im getting this error when i run the bat file
Ideally, search for the error on the subreddit I recommended in the video. I can’t help with specific tech issues.
Did you manage to resolve this query? I am getting the same issue.
@@palakkothari8257 yeah turns out i ignored the first part of the video and i just don’t have the right computer requirements
how do i fix RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check i cant find any real info on it
Please google your error, you’re not the first one to get it.
@@albertbozesan I fixed it all I had to do was download stable diffusion for AMD
When I open Stable Diffusion, the restore faces button does not appear? What can I do?
They changed a few things in the UI recently, that might be the reason. Have you googled or searched on Reddit?
So you wouldn't suggest installing Stable Diffusion with only 4GB VRAM? Won't it work at all?
There are some low VRAM settings you can try out. It may work but be prepared for extremely slow processing and many extensions not working.
I have this error: OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 4.00 GiB total capacity; 3.42 GiB already allocated; 0 bytes free; 3.46 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
What can i do?
Search your error online - someone has probably had it before. I haven’t.
You need to reduce your batch size, batch count, or resolution. Experiment with that. Your GPU doesn't have enough memory to process what you are requesting it to do.