Stable Diffusion-Master AI Art: Installation, Prompts, txt2img-img2img, out/inpaint &Resize Tutorial
Dear friends, come and join me on an incredible journey through Stable Diffusion. Let's dive in deep and learn how to generate beautiful AI Art based on prompt and image input. From download to installation to learning all the important tools and exploring different use cases. It can be used for having a lot of random fun but it can also serve as a powerful tool to create gorgeous art or even textures for your game project. Warning: I can confirm that it's highly addictive! 😎
I enjoyed working on this tutorial a lot and dedicate plenty of time to bring you content that I hope you enjoy watching too. I don't have a Patreon account but if you want you can buy me a coffee by getting it as a tip jar edition in best 4K quality. Your support is much appreciated and motivates me to bring you more tutorials like this one 🙂☕ chamferzone.gumroad.com/l/Sta...
00:00 Intro
2:42 Download & Installation
5:56 Checking for updates
9:42 Text 2 Image (txt2img)
10:52 Prompt inspiration: lexica.art
12:31 Output Folder
14:35 Using Txt2img to create a secret cat civilizaiton
17:31 Resizing an image to a higher resolution
19:48 Prompt Builder tool (promptomania)
23:09 Let's generate a human
25:38 GFPGAN - Face restauration
26:22 Loading a different model file (waifu model)
28:03 Image to image (img2img) Kidsdrawing to masterpiece
29:03 Outpaiting - expanding our image
31:21 Inpainting - fixing or changing parts of our image
33:39 Combining Photoshop with img2img
37:48 Tileable PBR textures with Substance Sampler
38:31 Creating variations of our image
42:44 Creating stylized images for our video game
43:40 Updating to Stable Diffusion 1.5
47:02 Meme creation with CFG Scale and Denoising Strength
49:57 From chili to beautiful woman with img2img
53:23 Inpaint yourself or your cat as a superhero
57:36 ChamferZone Discord and outro
Download & Installation
Open these links and download the first five into a folder. Watch the video to see for details.
- Git git-scm.com/download/win
- Phyton www.python.org/downloads
Important mention for Python: Some people reported problems getting S.D. to run with the "latest version" of Python. Installing the one I used (3.10.8) should make sure it's working: www.python.org/ftp/python/3.1...
- GFPGAN Model: github.com/TencentARC/GFPGAN
- Stable Diffusion 1.4 Model: huggingface.co/CompVis
- Waifu Model: huggingface.co/hakurei
- Stable Diffusion Web UI (Just open the URL. We will install it through git. No need to download) github.com/AUTOMATIC1111/stab...
- Stable Diffusion 1.5 and 1.5 Inpaint model (will be explained later in the tutorial. No need to download for the initial installation)
- huggingface.co/runwayml/stabl...
- huggingface.co/runwayml/stabl...
Useful links:
Prompt and Image library: lexica.art
Another Prompt and Image library: www.krea.ai
Prompt Builder: promptomania.com
List of artists: rentry.org/artists_sd-v1-4
Latest model links: rentry.org/sdmodels
Check if your art is in S.D: laion-aesthetic.datasette.io/...
ChamferZone Tutorials and community links:
gumroad.com/chamferzone
/ discord (Safe for work chat)
chamferzone.com
www.artstation.com/timbergholz
/ chamferzone
/ chamferzone
This is the most complete and easy guide i have found. Really compliments! Very clear in everything!
Hi Tim, i am a long time follower since the 3ds max ak tutorial to your newer blender tutorials. thanks for everything you have taught me over the years. I have been working in Cinema4d as a 3d artist for 6 years but because of your videos I was able to learn the basics of blender and land some freelance clients for the use case of in game engine ready assets for unity. Then I learned substance painter and realized the workflow of video game ready assets is not a daunting task I once thought. There is so much to learn in your tutorials and you always seem to be at the cutting edge of workflows and programs. It’s an amazing world artists are working in right now and from my perspective your tutorials have been game changing and very detailed. Thank you again and I hope to keep learning from you and other artists around the world!
Hi Lobot! Awesome to read your comment and I am happy to hear it. Congratulations on your freelance clients! I am glad if my tutorials were able to contribute a bit to your success. :) Yeah every once in a while some new things appears and I think Stable Diffusion was one of those, I had to dive in on this and no regrets. It's truly fascinating. I was also pretty much learning it while I did the recording and it serves as great motivation to learn something coupled with the intention to really understand it to the point where you can make a tutorial about it. This is also how I approached the Blender tutorials which I am glad you enjoyed watching. Thanks for your support!
@@ChamferZone love u Tim
Love your content and the way you teach. Subscribed ❤
Always so helpful, thanks Tim 💪
Fantastic tutorial, thanks Tim
maybe one of the best tut/intro to stable around
This is super useful for game design
very useful info. thanks for tutorial.
This is a great video, thank you so much for the whole process on how to install it, it was very smooth. One question I do have though is, whenever I look at Stable Diffusion generated art online, among the prompts is also the model hash number. I was wondering what that is and if it's similar to a specific seed in an image. Can we input the model hash number into somewhere specific? Because I can't find it anywhere in the SD ui. Thank you :)
Thanks pretty cool
Great tutorial! Followed it easily. How come i don't have a history tab like you do? I need that
Wow dude thank you. Subscribed
Defo interested in 3d shortcuts for assets and character creation
@@TrggrWarning Thanks for subbing and happy to hear it!🙂
As a Blackpink fan i died at the Jisso part :^) Anyway good video and well explained!
Hi there.. what model/diffuser libraries/checkpoints of stable diffusion is good for architecture and interior design visualization? Can you share a brief info about the model that can be used for architecture and interior design?
thank you
the best tutorial on this topic for real
Mind-blowing! Thanks a lot Tim for this new tutotial!
vielen dank
Small tip that'll help everyone navigate your tut: you need to have the chapters in your description start with a 00:00 (call it intro or whatever) and only then KZhead will have the chapter names on top of the playback for easier navigation.
Amazing, thanks so much! I was wondering why it doesn't show the chapters in the timeline. I thought they got rid of this feature but I simply forgot about the 00:00 Cheers! 🤜🤛
damnnnn this was so interesting Tim!!! Very useful as well with the textures!
Thanks mate! Glad you find it interesting! :)
Thanks for this tutorial! I have a problem during the "generate" picture, sometimes he don't work when i put a lot of word in "prompt". If I put only 1 to 5 prompt the generate image work, but if I put more like 15 or 20 prompt the generate image don't go to 100%, do you know what it's the problem?
Thank you for this thorough tutorial! I've been following Midjourney with some interest but I much prefer the control that SD gives users. How much GB space would you recommend I put aside for tinkering around with Stable Diffusion? Or is that answer "as long as a piece of string"?
20 GB
8:14 anyone else getting this Error in the CMD? ERROR: Could not find a version that satisfies the requirement torch==1.13.1+cu117 (from versions: 2.0.0, 2.0.0+cu117) ERROR: No matching distribution found for torch==1.13.1+cu117
Wow. Mind-blowing! 😳🤯
Absolutely bonkers. The rate of improvement from even a few months ago is staggering. This technology is changing everything.
Truly is! It's gonna be amazing to see this develop. This is absolutely bananas already. :)
Great Tutorial Tim....! I am an old subscriber this tutorial could have changed my game up but it says my Python version isn't compatible. What Can I do about it?
Whats the computer spec requirements to run stabel diffusion. .5 I am seeing crazy sepcs like CUDA GPU at 24GB, What are you using to run the demo we are seeing ?
i'm new into sd and dreambooth, been trying 5 models, some relatives and other famous influencers, i'm curious why do some of my models can't seem to get the proper and accurate face (different face / defect / crooked eyes)? eventhough they have better photo quality, i put in 30 pics for each model. any tips?
How to install the History button on SD operation interface? Thank you!
Expert in IT industry, someone like you should remove the Windows.old file, which takes 30 GB disk space and does not do anything... ;) Great, professional movie, thanks!
when I do the last part of choosing model.ckpt it keep saying error; connection errored out. Any idea why? Edit: *derp* I figured it out (I'm guessing). I had closed webui-user control prompt.
If you add git pull pause before the @echo off line to the webui-user.bat file it will automatically check for a new version when it starts running.
is that a space after pull
@@JJ-vp3bd It's as I typed it, no space. git pull pause The pause command will pause after "git pull" has finished and stable diffusion will start loading when you hit any key, it will also tell you this on the screen.
What pc specs do you have? Speficically GPU?
is there any minimum required for the gpu version and vram
Can I get help with an error? The error occurred in the txt2img module of the program, specifically in the process_images function of the processing module. The error message suggests that the program is having trouble sampling from the model, and is unable to complete the request. Initially it doesn't load 'xformers', but I added it to the batch file - either way I get a "torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 30.00 MiB (GP U 0; 2.00 GiB total capacity; 1.41 GiB already allocated; 640.00 KiB free; 1.48 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation" on the first instance and "RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0! (when checking argument for argument index in metho d wrapper__index_select)" on a second instance of trying to generate an image.
yow how to install this a new release of pip is available. it said run the python.exe -m pip install -- upgrade pip im confuse the command not work
how did you get the dark background? the white background hurts my eyes
can this create animation with this setup?
How are your generation times so fast? Do you speed up the video or have a powerful GPU?
Hi, I mention in the vid that I fast forward through these parts or else the tutorial would be three times as long 😅
🤣
You are based Tim. I love you
New sub liked and this comment.
when i send images to extra it gives me this error, i don't know what to do. I cannato resize my images urllib.error.URLError:
I am going to try ">5fingers" in the negative prompt field.. You never know..
with the new loader, you don't need to manually install python or other things. It's fully automated now!
An all-in-one installation? How do I find that?
When I try to upscale the image I get this error: URLError:
sir last moment i can,t complete run can you help me
how do you download git hub on mac
This is so fucking cool
Thank you very much, good sir, you are the hero we need, but I don't know if we deserve you, however. I am a game designer/writer who tried to learn concept art but could not have the patience to become good at drawing. Now I don't have to. You saved me so many hours, with this video please continue. Have one of the best days a man can have, my good sir.
This reminds of Nvidia Canvas. Can you imagine what digital art software will look like in 30 years? Maybe we just plug our brain into a smart phone type of device and perfect art just comes out in seconds.
Truly mind boggling to think about that and we already made so much progress just this year. It's moving ahead with lightning speed. 🤯
Be awesome if you made one for MidJourney as well.
Cheers, I want to look into this :)
@@ChamferZone MidJourney is superior in a lot of ways, it also runs through Discord so easier to get.
@@mmckn1885 I certainly want to sign up for it but from what I gathered the differences are not that massive in terms of quality and also let's not forget that MidJourney is a subscription service (and not cheap at that) Here is a good comparison of the two for the art it creates medium.com/codex/midjourney-vs-stable-diffusion-same-prompt-different-result-dd29ca482235#:~:text=In%20terms%20of%20image%20quality,lot%20more%20natural%20than%20MidJourney.
@@ChamferZone It is expensive but a lot of people use it and the system is different than Stable Diffusion. I think it would be good to have a tutorial for a different system but it's not necessary. I pay for it and have yet to regret it. Really really good for making tillable textures.
@@mmckn1885 Awesome sounds good. Yeah I am excited to see for myself :) Having multiple tools at hand is is always a good thing.
No Mac install?
Quite brave to use Waifu Diffusion for demonstrations like this. I guess you could cut any accidents out of the video, but still, brave choice. For those who want anime without the lewd stuff, there are SFW models on that same page with all the nudity removed. But negative prompting should also work, if you know how the Booru tags work.
Torch is not able to use GPU
After launching the .bat file it worls all fine until he get to the "launching webui" point. I get this error: ImportError: DLL load failed while importing cv2: The specified module could not be found. I tried different versions of python. deleted and reinstalled everything. Nothing worked for me. Wish there could be an intaller like NMKD SD UI for dump people like me ;)
Unfortunately with those errors it's a bit hard to say. I haven't seen this one before. You followed all the steps form the tutorial and you also downloaded the Python 3.10.8 not the "latest" version right? You can try an alternative way with Anaconda (Python distribution) instead of Python itself. Maybe that might fix it? It's not that different from the installation I did. You can see it here: medium.com/geekculture/run-stable-diffusion-in-your-local-computer-heres-a-step-by-step-guide-af128397d424 Worth giving a try.
@@ChamferZone Finally figured out a solution: in the SD folder there is venv\lib\side-loading\cv2 all I had to do was to make a folder named cv2 inside the folder cv2 and copy all files into the new folder.
I installed latest python version it didn't work so uninstalled and installed the same version u used. when i run bat file it shows this venv "C:\StableDiffusion\stable-diffusion-webui\venv\Scripts\Python.exe" No Python at '"C:\Users\Azhar\AppData\Local\Programs\Python\Python311\python.exe' Press any key to continue . . . that directory has python310 now but its not accessing it its checking only for python311. how to fix this
Hey, not sure why it would do that. It sounds like you didn't uninstall everything if it's still checking for 311 version. Try doing the following. Edit the bat file and specify where your exe is located of 310 and save and start again. In my case it would look like this: (x for your username) @echo off set PYTHON="C:\Users\XXXXXXX\AppData\Local\Programs\Python\Python310\python.exe" set GIT= set VENV_DIR= set COMMANDLINE_ARGS=--medvram call webui.bat
@@ChamferZone It worked after deleting the stable diffusion folder and "git pull" again from beginning. The tutorial was great though, thanks for sharing
@@ChamferZone That solved it for me, thanks
@@ChamferZone THANK YOU, i had the same problem and ive been searching so longgggg until i found this and this seemed to have fixed it
secret cat civilization:D)) awesome video man, thanks
29:26
i am gettin "The file may be malicious, so the program is not going to read it. You can skip this check with --disable-safe-unpickle commandline argument." in the cmd bat of the webui
Same to Me.
mine is creating black images. why is that
Hello, thenks for your video, i get this error RuntimeError: Couldn't install torch.
Hi, apparently the latest Python version causes problems. Using the one I used 3.10.8 will make sure it works: www.python.org/ftp/python/3.10.8/python-3.10.8-amd64.exe Also updated the download description for it.
There is a checkbox when installing Python that needs to be checked..like Runs With or something like that.
It keeps having this problem: to COMMANDLINE_ARGS variable to disable this check'") File "D:\SOFTWARES\STABLE DIFFUSION\stable-diffusion-webui\launch.py", line 88, in run_python return run(f'"{python}" -c "{code}"', desc, errdesc) File "D:\SOFTWARES\STABLE DIFFUSION\stable-diffusion-webui\launch.py", line 64, in run raise RuntimeError(message) RuntimeError: Error running command. Command: "D:\SOFTWARES\STABLE DIFFUSION\stable-diffusion-webui\venv\Scripts\python.exe" -c "import torch; assert torch.cuda.is_available(), 'Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check'" Error code: 1 stdout: stderr: Traceback (most recent call last): File "", line 1, in AssertionError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
same here, any update or solutions? thanks @ChamferZone
49:06 Yes, by all means incorporate Hate symbols into your art. Out of all the things you could do...this is what you focus on.
Tim, do you think artists are going to be replaced by AI?
Hey Ricardo, in short no absolutely not but I think that AI will offer more and more solutions to things we had to do the hard way before. For example Adobe is currently improving us artists life by implementing great AI tools like the object picker that perfectly recognizes objects and masks them for you. Substance is also going into this direction with some of their products. Who knows maybe we will finally get an AI that does a perfect unwrap for us but at the end there will always be someone operating in the background that understands the subject matter. That's why I think that excellent artists will not be disappearing but in the contrary can be real gods with AI's now by combining artistic skillsets and merge it with the AI tools. Sure maybe you could make the point that anyone can that invest some time but that simply means that it will open up the world of art to more people which I see as a net positive too. The question will it be bad for artists sometimes reminds me a bit of the first locomotives where people didn't want to enter these hell machines. I feel that if you are not a part of this you are being left behind and I wanted to see how it all works. It's here now and acting as if it doesn't exist or shouldn't be used out of ethical reasons is not valid to me. That's the result of this tutorial and I see only benefits as an Artist myself (for now 😜🦾🤖).
Windoze is not the only operating system .... can we have a Linux version please ?
I've used ai a lot enough to know a prompt with that many words is damn near useless
"for doing it I choose this, this, this..." without actually explaining why. Not a tutorial, but rather overview
"It's not meant to ever be a replacement for artists......while also enabling people who have no art background to open this beautiful world up to them." This is literally one sentence after the other. Don't you see the glaring contradiction of your own words?
Does art class do the same thing?
So, not for MAC users?((
This should be OS agnostic. Python is able to run on Mac and Git is available as well. It should only be slightly different setting this up ;)
@@jakobmemborg7963 new to this Python stuff Very impressed by the video Will be the tutorial for "Us"?
Midjourney is more powerful, I'm working with it for 3 months now and it's possibilities for crafting are more advanced and intricate
Very nice, yes I want to try this too if I get a chance.
Human artists are dead as fast AI is improving.
I wouldn't look at it so grim. Thanks to AI Artists now have tools at their hands that even a year ago nobody would have thought are possible. Just look at Photoshop, they also went down the AI route now with their latest "object picker". No more tedious path work required to make selections. It saves me a lot of time. AI is clearly pointing the way to the future but I don't believe in a Skynet scenario where human creativity and skill won't be necessary anymore ;) But yes times are changing due to this and it's necessary to adapt to not be left behind. It's a fascinating new era.
I agree with @ChamferZone here. It is like Houdini can empower 3D artists in entirely new ways using procedural 3D elements, so can AI be used to speed up areas of Concept Art that were previously time consuming. If you look at how artists create mood boards from existing images, now they can generate exactly the elements they need and put it back into the exact idea they want to present for their customers. I can definately see AI generation go hand in hand with some manual tweaks in Affinity Photo or Photoshop
@@jakobmemborg7963 I agree with that, I actually saved a couple of AI generated images that gave me ideas to copy from but then at the same time I'm copying a machine. In the End AI will always come up with the most unique design and we end just copying it for ourself, no pride in that. But maybe I'm looking at it the wrong way.
@@doop00 Think at it like jump starting your creativity. It can be rusty at times (God knows mine is), but when the AI generates something sort of unique, how much more unique can your even more advanced human brain re-imagine it? 😉
tools for people without talent...
AI is not art.
Art theft
quit
There is no AI art. Be careful with that term. Its like green energy or sustainable development. If you stop to actual consider the meaning of it, you will realize I hope that its not a description of what it is, its social engineering term. Art cannot be created by a non sentient being and AI algorithms are neither sentient or a being. There is a reason why the term "artificial" is in there. AI can imitate art, but cannot create it. Depending on where the info to train it came from, there are many ethical and potential legal implications, but that is another topic. Art is surly is not. So I would appreciate if you consider what art is and what art is not and stop using term AI art. Its not art. Its at best an imitation of art. And there is a big problem of normalizing term AI art, I would think its obvious why.
artgerm, wlop, digital art, highly detailed, intricate, fantasy, mystical, sharp focus, Trending on Artstation HQ, deviantart, unreal engine 5, 4K UHD image more brackets u add, more often tag will be focused for example: (((digital art))). U may be also want not only positive but negative prompt: ((((ugly)))), (((duplicate))), ((morbid)), ((mutilated)), [out of frame], extra fingers, mutated hands, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), (((deformed))), ((ugly)), (((blurry))), ((bad anatomy)), (((bad proportions))), ((extra limbs)), cloned face, (((disfigured))). out of frame, ugly, extra limbs, (bad anatomy), gross, (malformed limbs), ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), mutated hands, (fused fingers), (too many fingers), (((long neck)))