How to install and use ComfyUI - Stable Diffusion.
HTML-код
- Опубликовано: 10 июн 2024
- In this ComfyUI Tutorial we'll install ComfyUI and show you how it works.
ComfyUI github.com/comfyanonymous/Com...
Download a model civitai.com
ComfyUI Manager civitai.com/models/71980
ComfyUI Examples github.com/comfyanonymous/Com...
FREE ComfyUI workflow for 1.5 models here:
/ comfyui-1-5-86145057
FREE ComfyUI Workflow for SDXL here:
/ comfyui-workflow-86104919
Prompt styles for Stable diffusion here:
/ sebs-hilis-79649068
Support me on Patreon to get access to unique perks! / sebastiankamph
Chat with me in our community discord: / discord
My Weekly AI Art Challenges • Let's AI Paint - Weekl...
My Stable diffusion workflow to Perfect Images • Revealing my Workflow ...
ControlNet tutorial and install guide • NEW ControlNet for Sta...
Famous Scenes Remade by ControlNet AI • Famous Scenes Remade b...
For anyone stumped about the change in the method of installation for the Manager(v0.22.2):
Move the bat file into comfyUI_windows_portable folder and double click it.
thank you very much, it worked, you helped me a lot.
if you did that and the "manager" button does not appear:
install git and run the bat file again.
KING!
@@trataraja thanks, mate. It helped.
@@trataraja wdym by install git
Thank you so much for the amazing explanatory video! It was exactly what I've been looking for. The way the information was presented was simply fantastic. I could understand everything, and now I finally have clarity on this topic. Once again, thank you for your effort and excellent work!
1000s of thanks from Germany, Murphylanga
Happy to help and very glad that you feel you learned something 😊🌟
Thanks so much for creating this video! Great introduction for ComfyUI & the bonus Manager! Was able to install & started generating images within 1/2 an hour! The ComfyUI manager is a must & without it I would have been lost. 🙏
So pleased you are doing updated install instructions. I'm tidying my various AI setups and these are making it easier for me. Just need an Invoke update now 🙂
Glad to help! I've been meaning to do an Invoke for a long time, but every time I do, they update their version. So I've been waiting for them to drop a new one 😅
FREE ComfyUI workflow for SDXL here (built upon ScythSergal's and RealLucifer's work): www.patreon.com/posts/comfyui-workflow-86104919
FREE ComfyUI workflow for 1.5 models here: www.patreon.com/posts/comfyui-1-5-86145057
Awesome video. Glad to see that you liked my image enough to use it as an example for loading workflows from generated images and fixing missing custom nodes with the comfy-manager extension.
I actually also made that custom node extension (SeargeSDXL) that was used to generate the image because I wanted more flexibility with SDXL and also use Comfy for img2img in SDXL (which is difficult with the default nodes)
👋
Just getting into ComfyUI and it's awesome, but how did you get access to SDXL? I thought it was just a research preview and you need to be a research organization to get the model?
@@eukaryote-prime After the leak, they released access to everyone.
The ability to reverse engineer photos to get the original workflow is very nice, as I have often wondered about how 1111 does this or that. I like the connecting up of nodes, it‘s intuitive and rather like an audio setup, too.
PNG files allows saving metadata inside the image, automatic1111 and comfy just save the prompt and data inside the png file.
This is exactly what I needed to get started. This looks so powerful. It was really well explained. Thanks a lot.
Glad it was helpful! Let me know how it goes once you test it out 🌟
I tried installing ComfyUI manager using the method in this video as well as the github batch file, but the manager does not show? Any help please.
What a perfect tutorial!
Your explanation is soo great you are a good narrator. Thanks a lot
A wonderful video, always so clear and didactic. thank you ..
It's a great UI when you have already a prompt you are happy with, a seed you like and you want to do some heavy postprocessing with. Else it lacks the flexibility of A1111 where you can create 8 images, twist few words, rinse and repeat until you get something you like.
It relies also a lot on external tools and nothing is integrated like with A1111 where you can do a lot in a single interface..
Dude, your tutorial is terrible. you're just mumbling to yourself and the stuff on your screen is not even visible
I didn't realize how easy it was to install! Also, it might be a highly complex interface, but many preloaded examples help to understand the process. Thanks 🙌
Glad it helped! Let's see some amazing creations now!
IS comfy ui free to use ?
OMG Look at that professional quality with multiple angles, slowly creeping zoom and even color graded.
I try, I try! Glad you liked it. Anything I could've improved for the intro part? 🌟
@@sebastiankamph This might be my lack of understanding but I didn't get what the benefit of comfy UI is based on the intro. Yes there's "custom workflows", but how does that translate to real world results and user experience? I use a node-based software for work so I understand how powerful node setups can be, but maybe a quick example of what could be achieved that is hard to do in a1111 wouldv'e been nice to see/hear.
A tumbs up just for the opening (Because !!!). Great video, thx. It'll be great to have more videos on ComfyUI because is more light than A1111 and need only what you really need to do your image.
Took a look at ComfyUI and ran because it wasn't intuitive. Thanks for creating this tutorial.
It could introduce new users better. The default workflow could be set up in a way to feel easier with explanations. But the drag&drop-feature is fantastic and far superior to png info in a1111 imo.
*7Zip and most compression tools:*
Extract - Decompresses to a temp folder on your C: drive then moves it to the location
Extract Here - Does that. This is by far the better option!
I'll never use this but I watched the video until the end because you have such a soothing voice ❤
Oh, why thank you 😘
This was a very simple, great guide for a newbie to use ComfyUI. Thank you!
Glad it was helpful!
And waiting for your SDXL0.9 videos :) I have setup my own Comfy UI workflow and also use Vladmandic. The results are so great. The CLIP they use now is so impressive, the model is so responsive. It's nearly MJ quality and I can't wait to see your advises about it.
SDXL Comfy video just uploaded! 🌟
All the camera angles lets go!
Now time to try comfyUI I was waiting for your guide on it and in the future SDXL too maybe? :)
That's the plan! I MIGHT even have recorded it. Wink wink. 😊
@@sebastiankamph Cant wait!! :P
you're the first channel i've pressed the bell icon for. thank you
Wow, thanks! I'm really happy to hear that. Hope you'll enjoy! 😊🌟
Thank you Sebastian. Great video for newbies.
Happy to help! Glad you liked it :)
Wow thank you man! Ive been wanting to get into all of this ai image creating scheme but never knew where to even start. You helped me a TON
Glad I could help! If you're completely new to it, I recommend also checking out my guide to automatic1111 which is easier to use than ComfyUI.
@@sebastiankamph I will definetly check that out
Very clear guide, good work
Is there any way to tell if an image has comfy UI drag and drop import data? Other than just dragging I mean. I can drag and drop my example bottle image but nothing else so far. But I can't see any differences in the file. Where is the workflow data being stored?
Fantastic video! Thank you very much!
As always, thank you for your support Jay! Your presence here is a rainbow factory 🌟
The tutorial I really needed. This really is an underated UI
Boom, I got you! What do you feel ComfyUI is best at?
@@sebastiankamph I believe it's best at efficiently utilizing the Gpu I always get a out of memory error in automatic 1111 even with the flags. But never in this UI.
Also since it's a node based work you can use dual model work when upscaling .
I think there is a lot more but I have only just started using this UI.
@@musicandhappinessbyjo795 You're right. I haven't encountered vram issues neither.
Thank you so much for the tutorial Sir!
My pleasure!
Thank you! Starting on this
Thanks. ComfyUI is certainly a powerful tool, especially for creating complex workflow that doesn't require human intervention. One reason why I haven't used ComfyUI is due to the inability to use in-paint / mask feature (meaning, automate as part of the workflow like "target and mask the face then perform in-paint using model X")
For sure. Comfy and a1111/Vlad has benefits both. In the end it's a matter of preference and what fits into your workflow.
Impact pact custom node has face detailer that do just that, similar to after detailer extension for automatic1111. Comfy is just as capable or better than a1111, but because not as many users, less help or documentation is available online, especially regarding custom nodes. I initially got scared of the tangling nodes, but now I'm glad that I did not back down then.
I assume you still really think switching from Automatic1111 to ComfyUI was worth it?
@@Najdmie
Thank you so much for the tutorial!
You are so welcome!
Awesome explanation 🎉
Glad you liked it!
this is great! one thing I learned is to NOT run the nvidia bat file as administrator. I usually default to this to not have problems but in this case it created the problem
Hi, I recently started using A1111 and in general using AI for graphics and getting really good results (mainly thanks to your channel). do you think I should now move to this platform or stick with 1111 for some more time? Also my mom got into it and she wants me to install SD on her pc so I’m wondering witch platform is better to start with…
I think the interface of a1111/Vlad is my main recommendation for now. ComfyUI is more for people that know they want a node based system. It's more time consuming and less forgiving. There's also much more content and guides for a1111 on RUclips. The end results will be similar! 😊🌟
Thanks for this tutorial.
I wonder if there is a way to use this node-based system to generate two flows of image generation and combining them using mask node into a third composed version.
Great video, thank you.
very intuitive, thanks
You're very welcome
The ComfyUI Manager link is broken? or no longer supported. It downloads a nearly empty folder that looks nothing like the one you extracted. Help?
Excellent Tut.
Thank you!
Helpful, especially with the MGR install.
Glad to hear!
@@sebastiankamph I was about ready to contact the folks at ComfyUI and ask to speak to the manager. Thanks again!
I love checking out the new uis
I always wondered, you got any good parsley recipes, or is that out of line?
@@sebastiankamph I wish! Maybe I’ll find some and share them with you
OMGGGGoodness. This is wild. O__O thnk you for your service friend. Wow....i have to spend this weekend learning this setup now.
Have fun! 😍
@@sebastiankamph Well finally got to it this weekend, since i decided to spend last weekend learning the webUI one. I have ideas for comics id like to make so this is definetly a step closer. May have to save up to get a new laptop.The Nvidia GTX 980m from my 2017 powerhouse GT-80 Titan is no longer the top dog. But then again if i could get the UI to split the work between my two GPUs in this old school powerhouse then maybe ill hold off.
image drag and drop not working for me, the earlier nodes exist as it is, i see no change in UI. What am i missing ?
gday SEb..... getting this error. Cannot import D:\ComfyUI_windows_portable\ComfyUI\custom_nodes\.git module for custom nodes: [Errno 2] No such file or directory: 'D:\\ComfyUI_windows_portable\\ComfyUI\\custom_nodes\\.git\\__init__.py' it seems to be happening following UI manager install......??
Thank you! does it matter what version of SD i got installed? and if I want to download SDXL, should I delete the previous version?
Im thinking about switching from automatic1111 to ComfyUI mostly because I feel like from the videos ive seen it produces better animations in animatediff but im also learning more and getting a lot better using Defourm.. any thoughts on if I should switch or not?
Did you switch?
Thank you
DaVinci Resolve uses nodes, so its nice to see ComfyUI follow suit
Informative!
Great Video. The one thing I noticed that when installing the "install-manager-for-portable-version" file into custom nodes folder, it would not run, I had to put that file "ComfyUI_windows_portable" folder instead,
Thanks for this video ! it was very informative. Do you happen to know where in comfyUI I can set the "clip skip" ?
Nevermind found it : CLIPSetLastLayer
How does this differ from the available comfyui extension already in the extensions tab list? I noticed the downloaded version per your guide has a hefty 5.18Gb python_embeded folder which doesn't seem to be present in the existing install.
Do I need this extra downloaded content? ComfyUI seems to open just fine already.. haven't tested it yet. Of course with the local install there are no .bat files to specify using an Nvidia GPU. I'd much rather be utilizing my capable Quadro card vs defaulting to a slower experience. Thx for any insight
I don't understand the feature " drag a pictur and you get the workflow" i've trie with many pnj and it never workes for...am i missing something?
Hi, thanks for the tutorial. When I try to drag and drop my image in comfy ui, nothing happens, am I missing something ?
hi is it better to install this manually or is the portable the same because i use the protable but it wont install insight face but the visions of chaos install installs the insightface nodes
I can't drag and drop to recover workflow from a previous image I generated (which I really need right now). I'm also struggling to use it cause the damn thing goes for max zoom in or max zoom out with a single move of my scrollwheel
after 2 months of A1111 I'm back to give comfy another try
also there only seems to be one file in the zip for the manager. not sure if it has changed since you used it.
Same here. It looks like it's been changed. Only one batch file.
hi i have a question can run Stable Diffusion in my local PC i have HP core i3 can run SD on it??? or i need to change and get something specific
I'm struggling, ComfyUI runs the first time, but the second run always gives an error. How do I avoid the double slash in the filepath? : FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\o\\ComfyUI_windows_portable\\ComfyUI\\custom_nodes\\.git\\__init__.py'
Cannot import C:\Users\oakey\ComfyUI_windows_portable\ComfyUI\custom_nodes\.git module for custom nodes: [Errno 2] No such file or directory: 'C:\\Users\\o\\ComfyUI_windows_portable\\ComfyUI\\custom_nodes\\.git\\__init__.py'
Thanks a lot again
Well, I need to apologize..... I was writing a text and because the video change after playing, I lost everything I wrote.... lol :( .... So I will be short..... Since this video.... A1111 is corupted and now, too many people have a lot of problems. Like, xformer, controlnet, tensorflow, unet amd so more. After 4 days, with some AI helper, I resolve some of them.... but not all !!!! Keep going.... I Like your video :P Happy new years !!!! :P
Thanks for the content! Could you help me? I've installed ComfyUI, some models and loaded a worflow. When I try to run Queue Prompt, all I got is (unknown error)?
Can you please explain the custom nodes bit at the 8.50 time mark again with the zip file and extraction as I got a bet lost at that point and assumed it should be opened by z zip .It doesn't .
Hello! when I press Install Missing Custom Nodes (either from local db or not) empty list comes up, nothing to install. do you have any idea what could be the reason? Also I had to copy manager as a separate folder in custom_nodes. coping contents of the manger directly to custom_nodes did not work.
Got to the bit about dragging in an image and it did not work. I realized then its the output images that have the node graph encoded. Without that image (or an image using that example network) I don't have a file to use to identify missing nodes using the manager. So much awesome stuff in the customs nodes but it would be great to know where I can get an example of that specific template. Is it also in the example files for comfyUI? I couldn't spot it.
Same just happened to me lol, thanks for explaining
yeah, thanks. I searched for "ComfyUI" images on civitai and it worked well.
Hi :) Do you have any video on how to speed the process of creating the image?
Hello. Can you tell me if in this UI implemented something similar to Cutoff extension of A 1111? This is primitive color separator, crappy and looks like abandoned by it's creator. Basically my question is - can i get requested color for requested items (e.g. "red shirt, green skirt" only and not green shirt and red skirt, or some crazy mix of both colors everywhere in my image)? If ComfyUI can provide this level of control i would like to try.
thanks a lot !!
ComfyUi looks great...might stick to Automatic1111 because I am such a try and error guy...but for setting batch workflows that are quiet complicated this will rock!! 😊
As I understand it, you can copy/paste the JSON (text file) config that other people create to do all the linking for you. Your only job then, is to tweak the sliders and change the input, just like A1111. I think both will have their place and I plan on using both.
@@erikjohnson9112 its a great concept, and it looks like it is very versatile...just crying over disk space, but might give it a spin! 😊
@@swannschilling474 Thankfully in the video Sebastian shows how to point to your A1111 directory so that they can share models. That certainly helps, but of course it is not everything.
I think both UIs have their strengths. For mass outputs, a1111 will still be key for me. I like how quick to load ComfyUI is, and like you already mentioned, just copying a workflow is so massively powerful.
@@sebastiankamph OK, you guys got me...that'll keep me entertained until SDXL 1.0 will be released next Friday!! 😁
is setting the path to automatic 1111 not loading the models for anyone else?
Thank you, this is fantastic... the problem is if I try to add the Manager in the custom-nodes the esecution stop in "pause" and I have to remove otherway I can't load ConfyUI
Do you have any idea to help me?
I have exactly the same problem.
@@TheAlternativeStudio I've got the same problem have you found out the solution yet??
Is there any way to have Comfy UI use my older NVIDIA Quadro M1000M card? So far I'm restricted to using CPU which is slooow...
As the image downloaded from discord, should it be generated by comfyUI or any other AI generator? Thanks
What is the lowest card i can use for gpu version. I am running it on 940mx and it is pretty slow. Any idea?
After running the comfyui manager installation .bat and restarting comfyui, there is no Manager Button showing up. Anyone knows how to fix this?
For some reason i dont get the manager even if i copy it into the folder you mentioned in the video. Any ideas?
just reinstalled comfy to my external drive thanks for the tip on how to link a1111's models and loras to my comfyui windows portable
You're welcome!
Heya I had a question. I removed the .example part in the Yaml source file, but when I saved it it just switched to extra_model_paths.yaml and the icon changes to a paper with a ! on it. Does anyone know how to resolve this?
Hi. Do you think that I can run Comfy UI with Windows 10, AMD R9 M 380 ?
thank you,can you make more step by step tutorial, for using,controlnet,lora, etc?
Thanks for the rundown Sebastian, and like the new look , keep on truckin' Bro!
You bet! And thank you! Do you feel we've evolved the content? 😁
I'm Sorry but can we get a video detailing how to use comfy on windows with an AMD GPU? ive looked at the docs and it sounds like its possible but i cant figure out how to do it. I have an 8GB 3070 or a 16GB RX7600XT
Can i run comfy ai data on a hardrive because dont have much memory on laptop ? Sadly the HD isnt SSD so will my generation will take time as in slow down ?
Hello there
I used the same image in Comfy UI and WebUI and wanted to get the same style image with the primitive that I downloaded from Civitai, but the result is that I get the same image in WebUI, but not in Comfy UI, I directly dragged the image into Comfy UI to get the same generation data with WebUI, could you help me to figure this out, thanks!
Can this do image 2 images? If so how do I do that? Is this better than A1111? Try running GPU but I only have 2G it was stuck on cmd same on CPU.
At 9:20 when I try to install the missing custom nodes, I get an error message stating that "The custom node DB is currently being updated" and then I can't do anything with the program. Is that normal?
Danke!
A big thank you! 😊😘
Hi. I added the path to my auto1111 webui folder on the same windows drive and it didn't import anything. Any tips?
how to use image to image for the video instead for the outfit? instead of text prompt
lol, you had my subscription at the chicken joke xD
Welcome aboard! 😆😁
Wonderful tutorial - thank you!
I'm going to assume it's quite easy to set up a chain of events, like rendering a first pass, then taking the result as input for hires-fix, then taking that result and performing an ultimate sd upscaler step, etc - any chance you could show how to set up such a pipeline?
Thanks! I've since posting this added a pretty good workflow in the description for both 1.5 and SDXL. I think you can use that as a base and adjust to your needs.
Wow, you actually respond to comments, I love you
@@marcelkuiper5474 How very kind, thank you 😊🌟 Good to have you here!
Is there any plaIs there any place on the internet that teaches how to use ConfyUI on AMD video cards? Or does it only work with Nvidia video cards?
big fan dude
Thank you!
Dragging image that was made by a ComfyUI workflow doesnt work for me. It doesnt do anything, yet I want to retrace the variables on how it was made
Perfect. One small mistake. On my computer the manager didn't load unless it was in its own folder
it says path not found when i run nvidia C:\Windows\System32>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build
Das System kann den angegebenen Pfad nicht finden.
C:\Windows\System32>pause
Drücken Sie eine beliebige Taste . . .
hello! why does your portable build run only the first time and then not!?
Hi Sebastian,
I'm getting this error when I try to run comfyUI;
C:\Windows\System32>.\python_embeded\python.exe -s ComfyUI\main.py --cpu --windows-standalone-build
The system cannot find the path specified.
Where am I going wrong?
I get the same error :/ any luck?
Regarding dropping a image into comfy and it gets wired up, he failed to mention that it only works with images already done in comfy, not just any image
Works on images created by automatic1111 SD too.
@@GrafitTomi do they have to be png or can they be jpgs?
@@Avalon19511 I tried only the pngs that was in the output folder
@@GrafitTomi jpgs don't work
For me just nothing happens, I drag image into comfy and nothing appears ;/