I paid attention watching your video more than my wife talking to me. I am worried I might miss your joke. Love it. Sometimes I just come to watch your videos just for the joke, not the topics... Thank you for all you do.
@@micbab-vg2mu There are other reasons I plan to keep using A1111 for now, mainly filenaming when saving, but Roop is another. Thanks for reminding me of it.
Well done. You covered all the important details. I just finished installing on MacOS, and I checked all the Controlnet models. Why not? I thought. Well, they are over 20Gb, and my install is pushing 70Gb! So, plan ahead with a large drive if you like a lot of models. Also, my Huggingface disk cache is 30Gb. (a hidden directory) No one told me about that. 😂
Hah! I wish I'd seen this comment before installing, I came to this video because I had pip version conflicts trying to install it myself, then path issues installing manually to a different drive with more HDD space, so I chose a default C:\users\.. install this time, even though I only have 110GB left on C:\ I checked all the boxes for all the models, and all the control net options... its still downloading now, sounds like I'm gonna run out of HD space before it finished... :/ edit: thank you Sebastian for the video though, this is my 3rd attempt installing Invoke with SDXL and it is finally working.
Well Tony, it seems that you are way ahead of me because I get totally stuck already at installing the Visual C++ runtime on a MAC. How did you do that in heavens name? is that even possible?
I'm not sure yet whether I'm here for the information or the Dad jokes. Either way, I'm happy to be here. I had Invoke go weird on me a while back (many versions ago, with strange results), it is probably time to try again.
Invoke is probably easiest. a1111 has most tutorials on YT and will probably be most powerful to learn. Comfy can be really powerful but is quite complicated.
As an amateur myself, Invoke is closest to A1111 but without the bloat of it. I like Comfy but I don't need the customisation power it offers, so it can get confusing quickly. Comfy is lighter on resources though, so that might be a plus. Neither Comfy or Invoke currently offer features I'd like that A1111 does. If I can get Invoke 3.0.1 to install, I'll be using all three.
Professional point of view on how we use the 3 software in our studio workflow. - Automatic1111: extremely powerful when you need to script something or you need to write your extension for a precise project that need produce a lot of images . gradio is crazy easy to use and extension layer in automatic is done in a great way. More over automatic1111 integrate most of what you need to create your models for the researching phase on a project. The API integration with other software, that let you use automatic1111 as an extension is much better than the other 2. Is an extremely stable software. -invoke 2/3: Wonderful interface that let you engineering a prompt very fast and in a fun way. Moreover invoke as the most easy to use inpaint / outpaint tool in the market when you need to do something very fast and you don't have time to import/export very precise masks. Outpaint tool is very intuitive without complicated workflow. However is the more instable of the 3. In a professional workflow is a nightmare needing to restart the workstation with 7-8 software open because your gpu vram is messed up in a not nice way quite often. - comfyUI: Extremely powerful tool and arguably the fastest computation time of the group. Is a great tool to define a complex workflow when you need reproduce some style consistently for many images. If you have a creative mind with tech skills is a very funny tool to use but very time consuming. Becareful, pushing the game to far with the nodes can easily lead to gpu memory leak and after some time the vram is full without reason and you need to restart your WS. Advice for an amateur usage: Invoke is the most intuitive of the 3 but if you don't need outpaint/inpaint in your workflow you will feel limited in many way very soon especially when you start to see new automatic 1111 extension popping out. So what to do? Buy a new 1TB SSD for 50$ that you use only for AI and install all 3 software (and many models!!! 😅 PS and don't forget to try StableSwarmUI
Hey Sebastian, Please make tutorials how to use multiple card graphics (use rig mining) in stable defusion to generate faster images and render video.😊
Hey Sebastian, Please make a tutorial how to use this model on CPU only, in the video you said that it will run slow, but would it be workable for people like me, just getting into sable defussion, and just want to learn the interface, of this model and others, AND have a look at the results generated, before i invest in the same GPU that you are using, thanks for all your excellent uploads you're the best on UA-cam that i have found for learning all about Stable Diffusion. ALL THE BEST FROM ROBERT LONDON. 👍👍👍
Can I replace the autoimport folder with the Stable diffusion models folder where all the models are inside? Or do i have to load all the models again?
i love how it locks up the computer installing it. EDIT: After the first hard reset, it said it wanted torch 2.0.0 but has decided to use 'cached version' of 2.0.1.
@@sebastiankamph I've heard about Comfybox for Comfyui that can get rid of all the nodes and creates a userfriendly UI but i really struggle do understand how to install it
Thanks for this explicit instruction. I have a stable version of Invoke running for six months that I don't want to endanger. Can I install Invoke 3.0 and SDXL in a way that preserves the existing system intact? I have a near-empty 2TB SSD (not my system drive) Can I install everything there?
Not working - install all from link but when I start download 2.3 Gb -console show error and pres any key - I already lost all my evening but its not working.
very useful videos thank you! But how does one combine 2 existing pictures, like putting an existing foto/generated image character to another already existing image? So this would involve putting a mask on the generated image where one would like to input an already existing one? It's like the other way around. or maybe I'm just dumb? Please tell me it can be done with Automatic1111, since I've already installed it.
Thanks a lot for this invaluable info. Can all this process end up with a portable setup (at least as far as the AI models are concerned)? saves from having to redownload the whole thing. EDIT: I answered my own question, answer is yes : )
I had a standalone 3.0 version installed but it wouldn't update to 3.0.1. So I tried the way in your video but after running the install.bat, it says "No python detected on your system. Please install Python version 3.9.0 or higher" and blah blah. I have Python 3.10.9 installed (with Add to path ticked), plus I uninstalled, reinstalled and restarted etc. My install of Python worked with A1111 before I switched to a standalone version. Anyone got any help or ideas?
@@theprophet2276 I'm trying a solution right now. Uninstall all of Python, then install a version from the MS Store (3.9 and 3.10 are avaialable). The installer works, so far so good. Give it a go or wait until I come back as it's feeding time here so I'll finish it after that. Good luck if you try it. :-)
@@theprophet2276 My post above works for me. It's running and I've used it to generate a couple of images. There were a couple of things different from the video during install but you'll figure them out.
stderr: Error: Command '['C:\\a1111\\stable-diffusion-webui\\venv\\Scripts\\python.exe', '-Im', 'ensurepip', '--upgrade', '--default-pip']' returned non-zero exit status 1. i did everything like the vid.
A woman's manikin breasts count as NSFW? Then why did I get banned from your discord channel. Maybe the admin will wipe his eyes. I've posted my art in other channels and it's been fine where SFW and NSFW are specifically there
Early access to videos as a Patreon supporter www.patreon.com/sebastiankamph
I paid attention watching your video more than my wife talking to me. I am worried I might miss your joke. Love it. Sometimes I just come to watch your videos just for the joke, not the topics... Thank you for all you do.
Haha, I love it! Thank you very much for your support and I'm glad you are enjoying the content :) You're a real superstar 🌟😊
It works and is easy to use - thank you. The one disadvantage of Invoke AI over A1111 is the lack of a roop installation option.
Roop is crazy good, my homie.
@@memoryhero It is wy I will have to install A1111.
@@micbab-vg2mu There are other reasons I plan to keep using A1111 for now, mainly filenaming when saving, but Roop is another. Thanks for reminding me of it.
@@Elwaves2925 I have just installed A1111 - fast and easy - I will keep both UI. Both systems have pros and cons.
Nice! I haven’t tried Invoke yet. Thanks for the vid!
The canvas feature is really lovely (it has been for a long time tbh)😍
Great video Seb! Thanks for sharing!! Jay
You bet! Always happy to see you in the comments section 😊 One of the real mvps
Well done. You covered all the important details. I just finished installing on MacOS, and I checked all the Controlnet models. Why not? I thought. Well, they are over 20Gb, and my install is pushing 70Gb! So, plan ahead with a large drive if you like a lot of models. Also, my Huggingface disk cache is 30Gb. (a hidden directory) No one told me about that. 😂
Maybe I forgot to mention that 😅. Thank you!
Hah! I wish I'd seen this comment before installing, I came to this video because I had pip version conflicts trying to install it myself, then path issues installing manually to a different drive with more HDD space, so I chose a default C:\users\.. install this time, even though I only have 110GB left on C:\ I checked all the boxes for all the models, and all the control net options... its still downloading now, sounds like I'm gonna run out of HD space before it finished... :/ edit: thank you Sebastian for the video though, this is my 3rd attempt installing Invoke with SDXL and it is finally working.
Well Tony, it seems that you are way ahead of me because I get totally stuck already at installing the Visual C++ runtime on a MAC. How did you do that in heavens name? is that even possible?
100 000 subs by the end of the year? 🎉🎉🎉
Subbed, if only for that pan. And the AI. But primarly the pan.
Welcome aboard, sailor!
I'm not sure yet whether I'm here for the information or the Dad jokes. Either way, I'm happy to be here.
I had Invoke go weird on me a while back (many versions ago, with strange results), it is probably time to try again.
Happy to have you here! With the release of Invoke 3, there are huge changes going on under the hood, so might be worth checking out again.
Automatic 1111 vs Invoke vs Comfy? For Amateur user like me :) What do you think is best?
I have found invoke to be the most user friendly
Invoke is probably easiest. a1111 has most tutorials on YT and will probably be most powerful to learn. Comfy can be really powerful but is quite complicated.
As an amateur myself, Invoke is closest to A1111 but without the bloat of it. I like Comfy but I don't need the customisation power it offers, so it can get confusing quickly. Comfy is lighter on resources though, so that might be a plus. Neither Comfy or Invoke currently offer features I'd like that A1111 does. If I can get Invoke 3.0.1 to install, I'll be using all three.
I’ve had a lot of satisfaction with DiffusionBee on MacOS. Limited utility, but runs just like a regular Mac app. No terminal shenanigans.
Professional point of view on how we use the 3 software in our studio workflow.
- Automatic1111: extremely powerful when you need to script something or you need to write your extension for a precise project that need produce a lot of images .
gradio is crazy easy to use and extension layer in automatic is done in a great way. More over automatic1111 integrate most of what you need to create your models for the researching phase on a project. The API integration with other software, that let you use automatic1111 as an extension is much better than the other 2.
Is an extremely stable software.
-invoke 2/3:
Wonderful interface that let you engineering a prompt very fast and in a fun way. Moreover invoke as the most easy to use inpaint / outpaint tool in the market when you need to do something very fast and you don't have time to import/export very precise masks. Outpaint tool is very intuitive without complicated workflow.
However is the more instable of the 3. In a professional workflow is a nightmare needing to restart the workstation with 7-8 software open because your gpu vram is messed up in a not nice way quite often.
- comfyUI:
Extremely powerful tool and arguably the fastest computation time of the group. Is a great tool to define a complex workflow when you need reproduce some style consistently for many images. If you have a creative mind with tech skills is a very funny tool to use but very time consuming.
Becareful, pushing the game to far with the nodes can easily lead to gpu memory leak and after some time the vram is full without reason and you need to restart your WS.
Advice for an amateur usage:
Invoke is the most intuitive of the 3 but if you don't need outpaint/inpaint in your workflow you will feel limited in many way very soon especially when you start to see new automatic 1111 extension popping out.
So what to do? Buy a new 1TB SSD for 50$ that you use only for AI and install all 3 software (and many models!!! 😅
PS and don't forget to try StableSwarmUI
Note to new installers: 3.0.1 is out with fixes for SDXL 1.0!
Convenient with automatic SDXL setup 😊🌟
great tutorial. thanks. 😊
Very kind of you! 😊🌟
Nodes like Comfy too !
Oooh, yes!
Very cool, I didn't think that the ai is getting faster
This is gonna bring my Nvidia 3060 _to its knees_ ! 🔥💻🔥
Will be great for the cold winter nights!
Is there a point in using this if I'm using the A1111? Does it give me anything?
Is the invoke AI gpu usage like comfy UI?. Coz I am only able to use comfy UI well due to it's good gpu usage.
Invoke has a new node based system, however I have yet to see something as efficient as Comfy.
@@sebastiankamph true.
Also just wanted to ask. Can comfyUI do inpainting like A1111?
thanks for make this video
My pleasure! Thanks for watching 😊
invoke final_sigmas_type zero is not supported Please choose sigma_min instead. how do I chage this?
Thank you Sebastian,
Where is the link for how can work with Invoke 3 ? 😊
Coming soon! Not ready yet 😅
Hey Sebastian,
Please make tutorials how to use multiple card graphics (use rig mining) in stable defusion to generate faster images and render video.😊
I made a video on Stable Swarm UI that supports multiple gpus.
When I try to run mine it said there was no "triton" model and in the Invoke UI it shows none of the models I selected for download. Any ideas?
Hey Sebastian,
Please make a tutorial how to use this model on CPU only, in the video you said that it will run slow, but would it be workable for people like me, just getting into sable defussion, and just want to learn the interface, of this model and others, AND have a look at the results generated, before i invest in the same GPU that you are using, thanks for all your excellent uploads you're the best on UA-cam that i have found for learning all about Stable Diffusion. ALL THE BEST FROM ROBERT LONDON. 👍👍👍
How to safety close the tool? Just ctr+c or simply close the window?
Can I replace the autoimport folder with the Stable diffusion models folder where all the models are inside? Or do i have to load all the models again?
You can doit in the browser interface. In the settings section on the right you can import models from a custom directory
How much space does the app require?
i love how it locks up the computer installing it.
EDIT: After the first hard reset, it said it wanted torch 2.0.0 but has decided to use 'cached version' of 2.0.1.
any way to use styles in Invoke?
is it fast like comfyui?
I haven't seen anything as fast & optimized as Comfy yet. Invoke however is much more user friendly.
@@sebastiankamph I've heard about Comfybox for Comfyui that can get rid of all the nodes and creates a userfriendly UI but i really struggle do understand how to install it
@@sebastiankamph Is there a way to see the sampling steps realtime in Comfy? like here in Invoke?
No really not, and is quite unstable
Thanks for this explicit instruction. I have a stable version of Invoke running for six months that I don't want to endanger. Can I install Invoke 3.0 and SDXL in a way that preserves the existing system intact? I have a near-empty 2TB SSD (not my system drive) Can I install everything there?
Not working - install all from link but when I start download 2.3 Gb -console show error and pres any key - I already lost all my evening but its not working.
Is there way to install controlnet and other extensions
Don't get me wrong,.. AI is pretty amazing,.. but we're all just here for the jokes.
Inwoke should be named Unstable Diffusion ;)
Unfortunately is true.
@@blender_wiki Not in my experience. It has been rock solid for months on my W64 RTX 3060 12GB system. First time AI user.
very useful videos thank you!
But how does one combine 2 existing pictures, like putting an existing foto/generated image character to another already existing image?
So this would involve putting a mask on the generated image where one would like to input an already existing one? It's like the other way around. or maybe I'm just dumb?
Please tell me it can be done with Automatic1111, since I've already installed it.
Appears this requires you already have SDXL installed?
Nopes, it will sort it automatically for you.
CTRL click on the URL in the comand window opens a new brower windows with that adress, no need to copy paste
I'm just here for the flamingo joke.
Rate today's joke 1-10
Thanks a lot for this invaluable info. Can all this process end up with a portable setup (at least as far as the AI models are concerned)?
saves from having to redownload the whole thing.
EDIT: I answered my own question, answer is yes : )
How to download xformer
Anyone else getting a white and black mess for outputs? (using a sd1.5 model)
Invoke ai seems very good ai ui, but it did not support windows AMD GPU
My computer apparently cant handle XL. 1.5 is okay but as soon as I switch it to XL I lose all my VRAM. I guess I have to build another computer.
For SDXL, try Comfy. It handles low vram very well.
Or just use a online AI service that cost like the Kw in your domestic contract.
I had a standalone 3.0 version installed but it wouldn't update to 3.0.1. So I tried the way in your video but after running the install.bat, it says "No python detected on your system. Please install Python version 3.9.0 or higher" and blah blah. I have Python 3.10.9 installed (with Add to path ticked), plus I uninstalled, reinstalled and restarted etc. My install of Python worked with A1111 before I switched to a standalone version. Anyone got any help or ideas?
same problem here :(
@@theprophet2276 I'm trying a solution right now. Uninstall all of Python, then install a version from the MS Store (3.9 and 3.10 are avaialable). The installer works, so far so good. Give it a go or wait until I come back as it's feeding time here so I'll finish it after that. Good luck if you try it. :-)
@@theprophet2276 My post above works for me. It's running and I've used it to generate a couple of images. There were a couple of things different from the video during install but you'll figure them out.
I've seen other comments mentioning installing from MS Store if they encounter Python errors. Happy it worked for you!
@@sebastiankamph Me too and thanks.
is it possible to run it on AMD Windows?
stderr:
Error: Command '['C:\\a1111\\stable-diffusion-webui\\venv\\Scripts\\python.exe', '-Im', 'ensurepip', '--upgrade', '--default-pip']' returned non-zero exit status 1.
i did everything like the vid.
I have AMD :(
A woman's manikin breasts count as NSFW? Then why did I get banned from your discord channel. Maybe the admin will wipe his eyes. I've posted my art in other channels and it's been fine where SFW and NSFW are specifically there