ComfyUI PhotoMaker: Installation and Advanced Workflow Tutorial

hello everyone been quite some time since photomaker was released and now there are a lot of updates to the phot to makeer custom boot in today's video I will show you how you can install the latest version I will also show you a workflow that you can use in order to generate images using phot maker let's get started you will require the confy UI manager in order to install photometer click on it then go into two custom nodes here at the top that is an input search keyword type in O2 maker and click on search now be sure when you are installing you install the version that is from Shimizu and not the zoo zo zoo implementation basically the Shimizu version is the native kiy implementation whereas the zoo zo zoo implementation it's just a wrapper around the diffusers so make sure that you have the ID 477 isic and then go all the way to the right side click on install you should see installing confy y for to Plus on the bottom once it's completed you will see a restall button click on the restart button all right so once it's completed and you have your web browser restarted click on the manager once again then go into install models at the top type in photomaker and make sure that you are getting the photomaker dv1 dobin it's an sdxl model so you will need to use sdxl checkpoints together with theaker click on install it will download the model may take some time once it's completed click on the close button right so right click on that empty space click go into add node then you should see photomaker inside you will see two noes but this is a little bit deceiving because we have more noes for photomaker they are just in strange areas so I recommend just double clicking and searching for the nodes once you know the node name now I will show you where to get it so yes in the IP adapter category there's also a prepared images or clip vision from PA this is using the implementation by mat from the IP adapter custom noes and the Photo Maker encode plus custom Noe and this is how it looks if you downloaded the old version you may see additional section but for this version we are going to work with eclip input the photomaker model and then an image then we have the text build for the prop there's a trigger work that's already inserted for you and this Noe will output a conditioning so the idea is we are going to have the model the model will go into the aaker en code plus the conditioning will go into the case sampler and this will just replace your positive however however we have to keep in mind that this is sdxl so you will need an sdxl workbook and you can basically just put this note into the sdxl replacing that positive next just to to take a look we can also check the apply autom make sty and this is going to add some tokens to your policy prompt basically just add additional words to your prompt based on the sty that you selected if you have a nude that's already adding a styles to your PR you can use that as well and the way it works is that we need to pass in a positive text negative text and then out will come a positive and negative textt this note will give out texts or strengths so do not confuse the positive and negative as like a positive conditioning or negative conditioning and uh the last note that we can select from the right click menu is on the IP adap section the prepare images or clip vision from path so it allows to pass a path and you can have a path like this through basically going to input the path to a folder containing images of your character so you can get that from online you can generate it using a stable diffusion basically you just need to have a folder containing the base portrait of the character that you want to use as the input image and the N itself it gives you different ways where you can crop it so we have a crop position right now it's Center but it can be changed to top bottom left fried depending on your image now next I said earlier that ideally you should use an sdxl workflow however just to show you you can actually use the basic workflow although the qualtity of the image is variable so in here in the bottom section I have a very basic text to image workflow and the Photo Maker encode plus here is just going to replace the positive prop here so here we go I just deleted the clipex and good for py prompt now I have the Photo Maker en good plus replacing it so from my low checkpoint the clip the clip will go into the photomaker anle plus already have my load image and my V model this conditioning will go in as positive conditioning the negative one can stays the same goes into the case sampler well now I'm just going to use use the load image just load one image but you can also use the prepare images for clip vision from path know that to load multiple images okay so I have everything connected here and I just want to show you what's going to happen if your L checkpoint you have an SD 1.5 checkpoint but uh the foot maker is an sdxl checkpoint if I click on Q prop here and see Q prop one it's loading the buter loader and see here the Bora goes into the encode and it will basically tell us that we have a mismatch shape and if you've been using sdxl you can kind of guess what's happening and see the shape here 4,096 that's basically telling us that we need the sdxl model so just change it to an as the XEL model and then click on load you probably can see the key sound going and you can see that the result is not great and that is because the latent image is still 512 x 512 so I'm going to change it to 1024 by 1024 you can also use any other resolution that's compatible with sdxl all right it's completed and this is the new image now once again I'm going to say that it's it's okay it's not a great image okay next for sdxl workflow we can simply go to the C GitHub repository will have the link somewhere above me then you scroll down where you see confy examples click on it it will open the page for the confy Y examples then look for sdxl click on it and we'll C up giving you an sdxl example we can drag this image so just drag it and drop it inside of confy UI and confy UI it should load up now if you have the model the checkpoint already down downloaded the base sdxl and the refiner model you can simply change this replace this one with the buter include L node so here we have the positive and negative so double click on an m space search for photoma and make sure to select the photoma in code plus note not the normal one that says puta in code now since we have a primitive text box here that is holding the positive text we can take this and pass it over to this text box in order to do that we'll have to change the input box you can right click and change it to an input so select convert text to input and you will see that we have X here we can take the string and pass it over to the text okay so just move it in place and then the string to the text the conditioning will go into the key sampler as positive for the refiner section you can leave the click Text include the same or if you want to give it a try you can also duplicate the photo maker and code node and then do the same for the second key samp for this example I'm going to keep it simple and use only one fo maker and code plus node next just fill in the empty no so ground the image drag out and look for a load image node I'm going to use this one for this example then go drag photomaker get the photomaker loader get the model or the clip you need to get the clip around the sdxl base model on the top Drive the clip connect it and that's about it change the positive negative and then click on qpro and um here's an example of what I've got I had to do three generation in order to get a good image so it's probably going to depend on the seed number as well as the different styles that you're using all right so in order to add the style I would take the text bom grab it move it a little bit then double click and type in style you should have apply for to make a style then take the positive from your primitive node pass it over to the Jamaica style Noe take the negative has it over as well into the negative field now this negative will need to go in to the clip text and code for the negative so drag it into the text the negative one should be read and then the positive will go into the photomaker and plus node yeah now keep in mind we're not using this positive for the first case ample so you can disable it by pressing contrl M that's going to mute it or you can also remove it I'm going to keep it just in case I will use it later okay and then you click on Q okay next I'm going to show you how you can get some more work flows so at the bottom here you can see that I have a search bar in my previous video I show how you can add this magnifying search to your confy U click on it says search for workflows and I can type in Photo Maker press enter it will search for a couple of workflows that uses Photo Maker and I can simply use workflow it will download the workflow and load the work though directly here and if you study it you can see that it's using the same principle we have a checkpoint it's an SD XL checkpoint in this case this particular one it goes into the prompt and the prompt you can see we have a positive we have the trigger word and this particular workfl is using the sdxl prop Styler so like I mentioned at the beginning of the video you can use any Styler promp Styler that you have as long as as it's compatible with sdxl and scrolling all the way we can see the basic sdxl and this is the important part we have the clip text en code sdxl version and you can see the number here 496 if you remember I showed you an error earlier when I was using SD 1.5 as basically it's we have a shape mismatch now this particular worklow also has a face section this one is using the nude retina face and it's going to do a b box and crop it based on uh the face so it can automatically crop the image you can use any any anything else to process your input image even it basic drop node is going to work fine then we have the photomaker section and one thing to not here is that this one is using the old buta and cod new you can swamp this or the new one by adding it right click choose convert text to input and then everything else will be exactly the same so you can just drag all the inputs from this old version into the new version like so and the conditioning it goes into the case outla they can pass that over this node no longer required you can just replace it and save this version for later use and go it has a refine up section as well if you are using a community train model you can go from the B sdx Excel section directly to a v a code and then get your image on the top by the way this section is to input your initial image you can replace this with the prepare images for clip vision from path by giving it a app like so and just point it to where you have your your images if you have the comfy White inspired pack there is one that's called load image batch from diio the basically directory and this one will allows you to again load images in batch can put in the directory here but you can select the St image by using the index here and you can also put a cap but say you have four images of a woman the next four image is of a man and you want the woman's part then you can start with the index zero and you can end out or this will get you image 1 2 3 four and always stop there okay so that was it hopefully it was helpful to some of you to upgrade to the latest to make a custom nde or maybe if you have not installed the custom nde previously you are able to get it and I've showed a couple of workflows that you can use with it as always experiment with it try out and see which workflow will best for you for that image you may have to give multiple input image in that case use any new that allows you to load multiple Imes of once I show you two the one that comes built in with the tomator prepare images for clip vision and then the other one is from The Inspired pack okay so thank you for watching as always thank you for your support I will see you in the next one

Channel: Code Crafters Corner

Views: 1,497

Keywords: comfyui, Code Crafters Corner, CodeCraftersCorner, ComfyUI tutorial, Photomaker, Image Generation, Tutorial, ComfyUI, SDXL Workflow, Advanced Workflow, ComfyUI Manager, Creative Tech, Image Editing, AI, Custom Nodes, Clip Vision, SDXL Model, Workflow Examples, Tech Guide, Photomaker Encode Plus Node

Id: WctKNFH6PzQ

Length: 15min 20sec (920 seconds)

Published: Wed Mar 06 2024