The Easiest ComfyUI Workflow With Efficiency Nodes
ฝัง
- เผยแพร่เมื่อ 21 ส.ค. 2023
- The Easiest ComfyUI Workflow With Efficiency Nodes. For those of you who are into using ComfyUI, these efficiency nodes will make it a little bit easier to get your workflow started. Although this is just a simple example I think you will see the benefit of using these nodes as it groups all the necessary ones on neat groups! No more nooodles! hahaha! Well at least more controlled noodles. This also works for both SDXL and SD1.5 in one workflow.
Special shout out to @Pixovert where I found this workflow, go on over to his channel and sub! • TIDY UP the COMFYUI In...
How to install ComfyUI @sebastiankamph • How to install and use...
Luciano Cirino efficiency-nodes-comfyui github.com/LucianoCirino/effi...
Even easier with the node manager github.com/ltdrdata/ComfyUI-M...
Install ComfyUi manager: @sedetweiler • ComfyUI - Getting Star... **Note* In the video he meant "Git Clone" not git pull.
*Disclaimer Affiliate Links Below*
📸 Gear I use
Sony A7C CAN amzn.to/3spWX8C
Sony 35mm 1.8 CAN amzn.to/36Apekr
OBS (FREE) obsproject.com/
Editor Davinci Resolve www.blackmagicdesign.com/ca/p...
Audio: Rode Podmic amzn.to/35sSnxv
🎵 Epidemic Sound
🔦 Find us on:
Discord: / discord
Instagram: / monzonmedia
Facebook: / monzonmedia
This has been quite useful for me personally and I'm redoing my workflows with this. I haven't noticed any major performance differences either, if anything 2-3 seconds longer but barely noticeable. Do you have any workflows that you are proud of? 😁
Whoa! This looks like a far superior setup than what I've been doing with all the nodes and noodles everywhere. This is much cleaner and looks much easier to setup a good workflow. Thank you for the video - I'm excited to try these Efficiency Nodes!
So far I've been loving it! Perhaps I'm being to finnicky but why are people not more neat about their workflows? hahaha! Would love to hear your thoughts when you've tried it out. 👍
Great stuff as usual 👍
Discovered Comfy UI nearly 3 weeks ago, and it is now my go to Generative Environment. Very well supported and and has a bright future.
Thanks Serge, I've been enjoying learning about it and honestly didn't think I'd like the node based system even though I use it in Davinci Resolve. However my curiosity of understanding the various pipelines of stable diffusion got the better of me 😁
@@MonzonMedia Oh yeah! Nodes in the COLOR page of Resolve certainly helped, I've given up learning BLENDER, life is too short ha! 🍷
But to use a 3d app to render stills, now that Ai does it so more quickly, and efficiently, is a not a smart path.
I am sure that a CAMERA module for SDXL is being coded in some underground lab, so that changing POV in 3d space will be a reality.
Happy days 🍺
Just started using this. It's great! I went from frustration with Comfy to it being likable. I hope you found a nice flow with it or just in general. Have a good one!
Yes it definitely makes it easier. I've learned a lot since this video and have so much more to share soon! 👍
Thanks for such an informative video, I have installed ComfyUI but use Easy Diffusion currently. Going to try your layout examples shortly on CU!
Glad it was helpful! Let me know how it goes!
I will!
Thank you, I put off learning ComfyUI too long, but I think a 3 block efficiency node workflow is a good start!
It definitely is a lot simpler. You'll find though that there is a different version of efficiency nodes but still the same concept. Will be updating this video soon!
How can I use more than one lora with this workflow?
anybody know how to use EFF LOADER SDXL? it doesnt have a clip or anything like that so im at a lost. (im not sure i need it but since im using SDXL models i feel like i might?)
Thank you!
You're welcome!
Nice video, I'll give it a try. By the documentation, shouldn't you use the Eff. SDXL Loader instead for SDXL models ?
Other than that, there's only 1 positive and 1 negative. For SDXL, there is 2 positives (CLIP_G and CLIP_L) and thus 2 negatives too. I've heard that you can have different results when you take in consideration theses 2 parameters separatly as intended by Comfyui SDXL.
Yes that’s correct however that wasn’t available at the time of recording this. But yes it works great!
This has been VERY helpfull. How do i connect the SDXL Prompt Styler to it?
It's great isn't it? For the prompt styles I tried a node I found on Civitai a few days ago bit haven't gotten it to work. I've just been using a google doc at the moment, haven't had time to look into it further. Will let you know if I figure it out.
@@MonzonMedia I was looking at the same thing. I got styles in Stable Diffusion which is just a styles.csv file in the root which works fine. But haven't got it to work in ComfyUI yet.
A member of the community made this codepen.io/TheCopernicus/full/mdQZGpW I've just been entering my prompts in there to copy and paste. Yes it A1111 it's so much easier. I'm sure there is a working node for ComfyUI just haven't found it yet.
Thanks for this @@MonzonMedia
@@MonzonMedia Thanks so much! Would really appreciate it, I just havne't been able to figure it out. It should be simple and "Efficient" but it's getting too complicated for me lol Thanks again, love your channel
its for sdxl or sd 1.5 models?
Both will work, note that this one is outdated but there is one that replaced it. Works pretty much the same way, with some minor differences.
If someone could finally show me how to tile upscale and inpaint, outpaint with Comfy on such a transparent interface, it would be wonderful! I don't like to use Comfy because instead of brainstorming image generation, I'm forced to set up scarily complicated node point systems in it. I don't mind if it's not faster, but it should have a user interface like AUTOMATIC. Thanks for the tutorial!
Working on all of these, just trying to see what set ups work best. ComfyUi isn't for everyone, it's sort of a niche type of system for people that like to tinker and want the obscure advanced benefits. Not to mention it's also quite efficient to run on most systems. The good thing is when you have a workflow you like, there is no need create a new workflow, just open it up and go!
@@MonzonMedia 1)There's already a handy workflow that includes img2img, inpaint and upscale in Comfy (SDXL+Refiner). If the author doesn't mind, here's a link to the review, there's also a resource link: th-cam.com/video/_Qi0Dgrz1TM/w-d-xo.htmlsi=oveGlQD8EGVdUbnF
2) The same author has a video where he connected 2 ControlNet models (Canny and Depth) to SDXL with fairly simple workflow (but without inpaint). th-cam.com/video/reqamcrPYiM/w-d-xo.htmlsi=mGpL0A85mIWnIvR6
But unfortunately I don't have them working together, only both modules separately 🙈 Maybe the respected author can find a common solution and connect these 2 ControlNet models to the first workflow (which has both inpaint and upscale)? 🙏🙏
@@Shisgara77 Thank you! I can always use more resources. I realize there are a lot of places I can get ready made workflows but I rather learn how to put them together myself so I can understand what's happening. For me it's super fun when you get it right! hahaha! I did learn how to do img2img, in-outpaint and xy plotting. Going to share soon! Thanks for the info bud I appreciate it!
@@MonzonMedia Thank you too!)) too feel that urge in me to figure it out from scratch and have a good mess with Comfi 🙈🤯)) So I understand you perfectly! But for now I'm afraid to fall down the rabbit hole with it)) And I'll be looking forward to more informative videos from you 🤗
@Shisgara77 it’s too late for me I need help! Lmao! 🤣☺️
Checkpoint loader in Efficincy node disables xformers,lowvram,medvram commands.
Why do you say that? I have not experienced any performance issues using it.
@@MonzonMedia it overrun the automatic comfyui's optimizations for low or med vram. see the cmd when you generate, and load the default loader and see the difference. for me it ~4s/it with it and under 1s/it without it,
@0A01amir interesting I hadn’t notice but good to know. Is this happening with SDXL or SD1.5? It could be that the clip encoder used for the loader is just the standard one.
I will check later today. Thanks for the heads up! 👍🏼
@@MonzonMedia well plebs like me with lowend machines will notice these :D
Never tried SDXL with it since it's bearly functions on my machine, only using sd1.5.
The sampler works fine, but shame the main reason to use Efficency node is for it's loader and how clean it is.
@@0A01amir 😆😊 ah ok…hhhmmm I’m curious now if nested nodes does the same thing. You can join certain nodes together, that’s basically what these efficiency nodes are. Will let you know once I find out.
please note that Luciano Cirino efficiency-nodes-comfyui came up with errors ! couldnt fix it or use it
Make sure you have the latest update, I was just using them yesterday and it works fine.
do you think efficient nodes will make comfyui faster?
Faster in performance? No..but faster to set up...yes. 👍
@@MonzonMedia thank you🤗
Like 45❤️❤️🌹🌹👍👍
Thank you 👍
i have no manager and so on
im still waiting for someone to explain to me what the different fields does ( like the scheduler, cfg, what is the effect of the steps, the control_after generate etc etc etc) xD
I've been meaning to cover this my friend! Not just for ComfyUI though, more of a general video explaining certain settings for any stable diffusion platform.
@@MonzonMedia would be a great help so hard to figure out what setting I need to tweak when it does something it's not supposed to do, when you don't know what the settings and nodes does 😂😂
Understood, are you only using ComfyUI or have you used other platforms?
Found this on the link...IMPORTANT: Due to shifts in priorities and a decreased interest in this project from my end, this repository will no longer receive updates or maintenance. Use at your own risk..
There is a new one out Efficiency Nodes for ComfyUI Version 2.0+, Author is jags111, pretty much the same thing. I'll have to do an updated video on this.
Please, can you help me, where is the download link ??@@MonzonMedia
wtf opening it via winrar in seconds via 7 zip 3 hrs
So no refiner? Interesting l
There is a KSampler Efficient Advanced that has the denoising option for the refiner. I find that most models on Citivai don't really need it. But then again it's personal preference. I am however experimenting with an SDXL workflow with refiner. 👍
@@MonzonMedia You are right, I haven't looked at the new SDXL custom models seeing that I was building a Comfy Work flow and mostly using the newest Sd1.5 realistic models that is giving me really great generations
All good, I think it's just fun to experiment right now...we have so many options!