Efficiency Nodes in ComfyUI

hey everybody welcome to another AI fuzz video and today we are going to be going over a group of nodes called the efficiency nodes now these nodes are a pack of nodes that combine a lot of other nodes into one big node and it really helps make the process smoother the workflow process it also declutters your your your desktop here and really makes things a lot more streamlined so in order to get the efficiency node packed you would need to First have comfy UI manager installed and um once you install that you would go into manager here and do install custom nodes and then from there all you would need to do is search for efficiency nodes so you would just come up here and just do that um and it's right here by Jags efficien efficiency NOS for comi version 2.0 now once you have that installed you hit install and then it'll say you would have to restart your comfy UI and just restart and then it's going to install it and you'll should have the nodes available and when you go to add node and you should have efficiency nodes and they're right there so you have have different level you have different sections sampling loaders stackers miscellaneous scripts XY inputs image and simple eval so we're going to do the loaders first and you have two here you have the efficiency loader which I'm going to load and we're going to load the other one and the other one is called the efficiency loader sdxl just going to just color this so we know the difference okay so if you notice if you zoom in here for the efficiency loader regular um it has two inputs low stack and CNET stack we're not going to use that right now but it does have outputs model conditioning positive and negative latent image vae and clip and if you notice here it has also aura I'm sorry not it has also your checkpoint name um VA name and some settings but you've notice here it also has a positive and negative text prompt and if you remember normally you would need to put in that I'm sorry cuz remember you Norm you would normally have have to put this in two of these to be positive and negative but it's already built into this node so from here you can actually let's write a quick prompt I'll put dog eating grapes even though that's not the best uh I put um ugly bad pause so I have my text my positive and negative text in there already I'm going to choose a um I'm going to choose a SD 1.5 model okay so now I don't need to do the conditioning and um I don't need to do those text and codes cuz they're right here so what I'm going to do the next step would be to go to my K sampler and efficien efficiency nodes efficiency noes does have K Samplers so we go here to Samplers and let's do this one case ampler efficiency and we would select model to model let's do the positive negative latent image let's the vae and from here you notice it has all the same outputs and from here you all you would need to do is do a vaed code right here and then from here you would need to do a save image now in the sampler in the efficiency sampler there are options there's the normal group of characters which is your steps CFG sampler name I'll go with that one and I'll go cross v e decode you can have it true or false I'm having it true preview method Auto so it should have that so this is basically the workflow of a text image process that we had before if I hit Q prompt and oop looks like I forgot something missing input is the Nega and see that so I oh I put the sorry so this is positive I'm sorry this is negative and this one is positive there you go so Q prompt there and this is makes it so much easier it it cleans up the thing cleans it up makes it nice and we will see what image we come up with with this efficiency node loader workflow so while we're waiting for that guys let's go into the next efficiency loader which is the efficiency loader sdxl so we're going to do the same thing it has a positive a negative and I will put in cat playing with yarn outside and uh negative we'll do the same thing ugly bad pause um I that's good enough and it also has the um load checkpoint which we can load any just got to get an sdxl checkpoint and we'll do that one and also has the latent width and height there as well and I didn't show you that but this also has a lat and width and height 512 x 512 and this one is 1024 x 1024 and also has a batch size of one and same with the other one so same thing for this so all we have to do let's go to it's a case sampler for sdxl right here see that hook up these pipes light and image vae and from here we would do a VA decode now it does have let's just see so we just do a VA to code from here okay and then we'll just do a save image node from here and that is your sdxl text image node using effic efficiency loader which is 1 2 3 4 prom one 1 2 3 4 nodes and that's it and you have all your options here as well as your Len options here so those are the efficiency loaders and let's see what these two come up with and there you go we have our dog eating grapes out of a wine glass so that's not so that one's done um so we're going to run the other one okay and there is our image of the cat playing with yarn so if you guys want to try that out you can it's it helps out in your workflow if you guys have any other uh comments or or questions put them in the comments like And subscribe AI fuzz and I'll be back with another video thanks see you bye

Channel: AIFuzz

Views: 1,509

Keywords:

Id: nVtkm1JhYc0

Length: 9min 0sec (540 seconds)

Published: Tue Feb 06 2024