Beginner's Guide to Stable Diffusion and SDXL with COMFYUI

[@Pixovert]

Course Discounts BEGINNER'S COMFYUI and SDXL bit.ly/GENSTART - USE CODE GENSTART ADVANCED COMFYUI and SDXL bit.ly/RESTAD - USE CODE RESTAD 00:00 Introduction 03:18 Downloading and Installing Stable Diffusion 10:20 Installing SDXL important information 15:45 Installing ComfyUI 22:20 ComfyUI - A demo 27:15 Getting started with ComfyUI - Prompting 30:20 ComfyUI - working with checkpoints 36:10 ComfyUI - the User Interface and workflows 49:00 ComfyUI - working with SDXL

[@SebAnt]

Are there newer codes. I wanted to try the beginners course but Udemy says code GENSTART is expired or invalid. 🤷🏽‍♂️

[@lioncrud9096]

for fucks sake how do i add the Lora. been through 12 set up videos no one bloody mentions it

[@carlajimenez1482]

How can I do multiple prompts running at the same time like the prompt matrix in automatic1111?

[@Michael-gf1jn]

comfy really needs tabs to switch between different workflows

[@cosmicstuff44]

Yah, that's a great idea!

[@JXQU3]

fr

[@alexoneprject]

what is the minimum/optimal hardware reqiurement to run COMFYUI+SDXL?

[@sunlightswift]

My 8gb VRAM 3070 is generating with it in comfyUI at 1024x1024. It peaks at about 6.5gb VRAM on Windows 10 with me

[@cosmicstuff44]

For some reason it generates errors with my NVIDIA GeForce GTX 1660 Ti w/ 6GB VRAM... just installed SDXL and ComfyUI so may yet find a solution. Good news is that I am able to run it just fine on my CPU given I have an Intel 3.2GB i7 8700 w/ 12 cores, liquid cooled and 32 GB RAM. I was surprised, but it renders 1024x1024 this way a bit faster than than it does 512x512... not complaining! I really like this ComfyUI gui, seems a lot better than the A1111 which I tried with SDXL and could not get it to work, and also had problems with it using the GPU and no option to fully use the CPU for everything like in ComfyUI.

[@BoldMold]

@@cosmicstuff44 Damn, I also have a GTX 1660 here. Should I even go through the trouble of giving comfy with SDXL a try or should I just stick with Automatic1111 and SD 1.5?

[@cosmicstuff44]

@@BoldMold I resolved the issue and it works great now with my GTX 1660 Ti provided I keep the resolution to one of the SDXL recommended ones. I went the NVIDIA drivers website and get the latest driver package for my GPU. I opted for the Studio version, although the Gaming one may also work. I then also opted for a clean install. This solved it. Windows was reporting I had the latest drives for the 1660, but that was bum info, so now I'm going to be checking for new drivers for a lot of other stuff too! It is so much faster than using the CPU option!

[@BoldMold]

@@cosmicstuff44 Thanks for the info! I just gave it a try with a fresh install of Comfy and it seems to be working (although generating+refining a 1024x1024 SDXL picture takes 10~20 minutes with default settings). I'll go check drivers too like you did!

[@erdbeerbus]

Great, but did you find a way to import a sequence of images in one way, like a batch task?

[@consuelodesautu1225]

Hola! No me funciona el cupon para el curso de udemy, lo porías reactivar?

[@Ca1vema]

Seems like deliberate are no longer available?

[@tomolson6169]

I'm trying to figure out how to build that sdxl workflow but I can't get the step control node to appear...

[@molk]

Primitive Node

[@stefistefi6069]

Thank you, amazing. People just should know that there is folder who save all images that are created, so not worry.

[@yamani3882]

I have no freaking idea what I am watching but I like it. 😅

[@rowelski]

Is there a difference in quality if I use in dreamstudio directly?

[@RedRojo210]

awesome video, could do without the loud music at random times lol.

[@courtneyb6154]

yeah.....kinda takes away from the video for sure. Great info though.

[@eveindy-w3c]

useful info BUT keeps fidgeting with the mouse and screen zoom control while talking, most irritating

[@userrrrrrrrrrrrrrrrrrrrr-tk3tn]

bro is explaining how to rename boxes, that red outlines mean "errors," all before he's generated a non-default image. my god man, just get to the point

[@jcaique]

JFC, GET TO THE POINT ALREADY

[@gurinolegg]

Thank you for this comment, saved me an hour, skipped this vid.

[@dinsy512]

All this detail is good for beginners like me. It is a beginner’s guide, after all. Most vids I’ve watched give you the bare bones and mechanics without improving the deeper understanding. I’m loving it!

[@dontbetea6576]

As a gentleman from Moldova... I can confirm that is their type, yeah, haha.

[@CoolhandLukeSkywalkr]

Can do some celebrities perfectly, but with other celebrities, it struggles, still, with some other celebrities, it can't do them at all. It can do an old Morgan Freeman, but I can't do younger Morgan Freeman. I can do Will Smith or Chris Rock, but I can't do them together. In some ways, it's better than Dall E 3, but in some ways, it's much worse. So, I use both. I can do Donald Trump in comfy, and it works amazingly well, other than the fact that I have to generate a hundred images to get one or two really good ones. Meanwhile, it can do Jennifer Aniston or Jennifer Lawrence perfectly. It can do a pretty darn good Taylor Swift or Miley Cyrus. It even does James Dean.

[@Ca1vema]

Can't believe you're selling courses and getting messed up in a video like that. You could reshoot parts where you got confused!

[@py_man]

❤

[@puja1985]

why stable diffusion if so freaking complicated,, ? i choose midjourney

[@simeondawkins6358]

yes yet mid is very limted also its simpler then it looks

[@LiiwTube]

Complex = Modular, Scalable and Unique for each user. In fact ComfyUI is pretty easy for the users who already use Node Architecture in other Applications. I would recommend you to study about Node Architecture, we have basically in all 3D/ Motion apps

[@cosmicstuff44]

Stable Diffusion can be run on your own computer, that is where the complexity comes in with installing everything and learning how to use the interface. It therefore does not cost you anything but your time as of now, and if you want to do large hi-res renders with SDXL you just rely on your own computers resources, not your wallets resources paying Midjourney for the server time and resources.

[@snark567]

Stable Diffusion isn't complicated, it's just this particular UI is a pain in the ass for people who aren't interested in dragging nodes around. The reason this UI is being shown around right now is because it's the only one that properly supports the newest version of SD. There's also UI like EasyDiffusion which is just a text box with options for you to prompt in, much easier to understand. I'm sure it will eventually support the newest version of SD.

[@burbujin4076]

Bc you can have more control on whatever you want to generate, adding the fact that is open source