00:00Hello guys, welcome back with me, Chow Chow 2025.
00:05Today we're going to learn about how to install and use Comfy UI for stable diffusion.
00:12If you already have Automatic 11.11, we're also going to show you how to integrate Automatic 11.11 model
00:20from Checkpoint, LoRa, ControlNet, and other into Comfy,
00:23so you don't have redundant data that will save you a lot of hardest space.
00:30Comfy UI is the most powerful and modular stable diffusion GUI and backend.
00:37This U, I will let you design and execute advanced stable diffusion pipelines using a flowchart-based interface.
00:46With that, you will have a very flexible but fast generation.
00:51Also, it could do almost all what Automatic 11.11 can do, and more.
00:57So, there are many things that you can only do in Comfy UI that you can't do in Automatic 11
01:03.11.
01:04First, to install Comfy UE, you do just need to access the Comfy UI GitHub at github.com slash Comfy
01:15Anonymous slash Comfy UI.
01:17Then, scroll down and go to Installing Windows.
01:23Just click the direct link to download. It will download a 7s file.
01:29You could open 7z with many software, or you could download the 7z software at 7.zip.org.
01:39Anyway, download the 7z file and extract it.
01:52Then, click the update folder.
01:53Access the Comfy UI Windows Portable.
01:56Enter the Update folder.
02:00Double-click Update Comfy UI.bat.
02:04Or you could also double-click Update Comfy UI and Python dependencies.bat if you want.
02:15After that finished, then go back to your Comfy UI Windows Portable Directory and run vidiagpu.bat.
02:25You technically could run with CPU, but it will be very slow, and I don't recommend it.
02:33The first time you use Comfy, you will given a standard workflow, remain you, of the Vanilla Automatic 11.11.
02:43Since this is the first time you use this, you will not have any checkpoint model.
02:49So the first thing to do is to integrate your automatic 11.11 installation with Comfy.
02:57Or if you have only Comfy, you could download your checkpoint model and put it on Comfy UI Windows Portable
03:06slash Comfy UI slash Comfy UI slash Checkpoints.
03:13Comfy also support LoRa and a bunch of other things.
03:17Check the models directory for the list.
03:19Now I assume you have automatic 11.11 already, so the next thing you do is to access Windows Portable
03:29slash Comfy UI.
03:32Search for extra underscore model underscore path dot y-a-m-l dot example.
03:39Rename the model by removing the dot example, so it became extra underscore model underscore path dot y-a-m
03:49-l.
03:50You could also copy-paste the file to make a backup.
03:54Open the file using any text editor.
03:58Change the base path into your automatic 11.11 path.
04:03For my PC, that mean d-drive slash soft slash stable diffusion 3 slash web UI.
04:13Unless you have a custom location setting for your model, all should already set up now.
04:20If somehow you put your model in other location, set up manually here for Checkpoint, LoRa and Other.
04:28Now save the file and reopen Comfy.
04:31Now you will have a model and load Checkpoint node.
04:41Okay guys, first time you load Comfy, it will look like this.
04:49If you are kind of confused, just remember the regular automatic 11.11 stable diffusion.
04:56First you have load Checkpoint here.
04:59This is where you load the Checkpoint.
05:01From here, you see these three things here.
05:05It's model, clip, and file.
05:08It's the output.
05:09So, it's just the clip.
05:14You are continuing here, here.
05:17See, if you, this thing called nodes.
05:20See here, clip, here.
05:22This is an input.
05:23This is an input.
05:24So, in order to move this, you need to move a clip output to a clip input.
05:36From the load, there are three outputs here.
05:39Model, clip, and file.
05:41First, clip.
05:42This is a prompt.
05:45This is a prompt.
05:45We put it a prompt.
05:46This is a positive prompt.
05:48This is a negative prompt.
05:49It actually doesn't matter if you put negative or positive here.
05:55As long as you connect it in the second output.
05:59That is key sampler here.
06:04Conditioning, conditioning, you put it on the positive and the negative.
06:12And the models is continuing from here to here.
06:16And this guy needs one more input.
06:18That is Latin image.
06:20We don't have any Latin image.
06:21So, we add this thing here.
06:24This is the image setting that you're going to use.
06:27Like, to set up the width and the height.
06:32From here is the regular sampler.
06:34It's seed.
06:35It's randomized.
06:36Step.
06:37GFG.
06:39Set 7.
06:42Sampler name.
06:45It's cars.
06:51pmsd.
06:56Scheduler.
06:57We change it to cars.
06:59The noise.
07:01The noise.
07:03Then from here.
07:05Latents.
07:06The.
07:06output is latent. So you put latent into samples, it's the same.
07:14In FIA you got it FIA from the models or you have loaded your own FIA. Here we're going
07:21to save the image. Now this is regular automatic level of an output.
07:30In here we use this node here. We don't load anything else so like you see when you load
07:39this for the first time it's going to be pretty fast unlike you load samples of vision because
07:44it's going to put the node that only that you use. Automatic 11.11 better force it to load
07:52everything. Let's click prompt. I'm going to click on here. See the green thing? It's currently
08:02in the process.
08:09Yeah. You can change this, do whatever you want, do you call it?
08:16Green face, yeah. Scary face.
08:27This is more fun right? Green face. The process is the same. You know.
08:41I'm going to save a lot of time.
09:08Of course you could query a lot. I'm going to do this eight times.
09:17I'm going to do this eight times.
09:21Get through the query, delete it, clear it, you know, whatever you want.
09:47And yeah, this is the original automatic 11.11. For example, if you want to do something
09:54like at a new node, for example, we want to do high-res.
10:03Here we could have for our obstacle model here.
10:08That's what we have done.
10:13Load, it doesn't have anything.
10:17We could add another node that is.
10:22Load it, log using image, obscuring, image using models.
10:33Okay, from here.
10:35We need the model here, obstacle image.
10:38And we could take the image here.
10:40Since save image doesn't have output.
10:44Not from Latin.
10:46But image from fire decode.
10:48Because there's nobody's image.
10:51And from here, we could actually add a node again.
10:57That is, save image.
11:03Or if you don't want to save, we could just add for view.
11:08To save.
11:16To I upscale, it's going to save here.
11:25I'm going to go here.
11:35Let's do it again.
11:40Okay, sampler.
11:42It's going to save the regular image.
11:46This is the one that's being upscale, I guess.
11:52We could actually set up here.
11:59Up how much it's being upscale.
12:08We don't set up anything.
12:10It's a fault for times.
12:13But, you know.
12:15You could, like, make it if it's work.
12:23Glare.
12:32It's just the upscale.
12:33Now.
12:35Be a different.
12:37See if we zoom more.
12:42See the different, right.
12:44Pass more.
12:46What else.
13:00This is four times the size.
13:07You know.
13:09See.
13:13Anyway, guys.
13:14I think that's it.
13:16There's a lot of things you could do using.
13:20Confi UI.
13:23And more than even what.
13:26The automatic 11 weapon can offer.
13:29And you don't actually need to design this thing.
13:33By yourself.
13:34There's a lot of workflow.
13:36That you could use.
13:37That you could save.
13:38And you could load.
13:40So.
13:42That's it for today's video.
13:43The next time we access Confi UI.
13:45I'm going to.
13:46Inform you about manager.
13:49Cause.
13:49That's the most important thing.
13:51I guess.
13:54For now.
13:55Use this.
13:56You know.
13:56Just.
13:58Tinker it yourself.
14:00Cause.
14:01You know.
14:01When you got the manager.
14:03When you got the workflow.
14:04At least you know how it works.
14:05So you can.
14:07Actually.
14:08Edit it.
14:09Or.
14:10Modify it.
14:11By yourself.
14:13So for now.
14:13Just.
14:14Think about this.
14:15Add node.
14:16Remove node.
14:17And stuff.
14:19That's it for me guys.
14:21Don't forget to like.
14:21Share and comment.
14:22And see you again on the next episode.
14:24Have a nice day.
14:24Bye bye.
14:26получится.
14:30Thanks for tuning in.
14:30We share.
14:30This is one of the work I'm calling.
14:30Is this one of the work I'm calling out?
14:30You
Comments