Automatic1111 vs. Fooocus | 🤘How-to use Stable Diffusion UI

hello so at this moment I'm thinking about what do I want to eat and well I was thinking about pizza so here putting my PR uh using focus a pizza with this very DET tight PR also negative PR and I got this but also say well I also want to toer so I put here also in the pront uh what do I want as Sir this detail one and I got a this very nice uh chocolate okay so in this video I just want to show you the difference between focus and automatic 1111 both of them they do use stable diffusion but what it changed uh it is the graphical user interface so I hope to give you some guidelines and also very important I hope you follow the previous two videos and you install both libraries okay both uh user interface which is very straightforward in Windows Windows 11 and then probably later I will prepare another video to do it in Linux or using uh sisting Linux or Mac so let's go into business so the important thing that after after you install automatic 1111 you have the link here and also you have the link to to the installation in description and you install Focus which is another user interface for a stable diffusion so this is very important concept both libraries or bus applications automatic 111 and focus there are user interfaces to another application another big model which is stable diffusion okay so stable diffusion is just a model to generate images no it's a text to image model that it will recognize your text it will have a language model and it is trained using big models and this NE networks and so on I don't want to go into details but it's up to you to look a little bit more in Internet uh so here you have also these links important stable diffusion is not the only one there are some other applications so you have MC Journey something similar a little bit more a little bit different from a stable diffusion the interface and so but the concept is pretty much the same and also probably you might be aware of Del okay the one that's developed by by the same guys from CH GPT but also used in by by Microsoft image generator so they do the same text to image but they have been trained different ways important also to stress that it's not anymore text to image now they do image to image uh text to video so hopefully later on we're going to move into markb concept and we're going to see how to generate text to image we can also improve images or scale in and much much more so you have the links on other important links is where do you have this model so I would like to address another concept because we talk about models then you have refiners Lota libraries Styles prompts and so on so basically you have your main application we're going to use our model our library will be stable diffusion and then we're going to use models that have been trained okay According to some data know going details but we have here this website a huging pH which probably uh it's here where is the most complete one you have all the community get together and put everything here so we have for instance you look here stable diffusion and you have this specific model or let's call it you will find the the the war checkpoint so they call it a checkpoint and you get that specific checkpoint that has been trained for that day envel you can generate your image so there is a lot of stuff here that you can search so feel free to to register and it's toally recommended to register because you can find you're going to find many libr here and to use other applications you need it some other uh extensions that we we can use with stable diffusion or with automatic 1111 or with falus then very important where do we look for more models you have here that is very complete this one but then also you have Civ AI that I strongly recommend here here you have many many many more model checkpoints and and so other extensions that you can add to your library so for instance you will find very often already mentioned the checkpoint checkpoint is just the the the the state of that model that was trained until this point so you have many of them and it will be up to you to pick up one then there are some other things such probably you're going to find wordss like laa laa is something it is a model but it's not a a a big model such as the checkpoints you already know that those can be as big as 12 gigs lis model that is used to improve your your images and also you have refiners buy and so on so probably you don't need to know all this stuff the important thing your checkpoint maybe a refiner and then probably as you go you will need to the load something else but I strong recommend this s this website to look for that that information but not only you get that those checkpoints you can also get for instance I really like to look in this website for instance you can look for wild wild cars okay and wild cars here will be the prompts and we have talked that prompts are very important when you are generating you have to be very specific or not super specific but whatever you put in your in the prom it cannot be something big you need to convey an idea and also try to avoid asra asra concept so kind of you give an adjective noun and a verb so an instruction from that you can get fantastic images you can you can see here so for instance you want to look something to do for Nat and you can get many uh checkpoints or laa or embedding or or wild card so it's up to you to pick up one I strongly recommend it when you pick up one you have for instance the checkpoint read a little bit the description so look at that this is a two 2.4 Jigs and sometimes you have the description the keywords that you need to use and so on L is the same so read a little description what is happening okay to know how to use it uh after this another website that recommend is Lexica Lexica really like this one so Lexica use another model this is propietary to then it's not of Open Source so just talking about this stable diffusion and everything that you find in in h huging face and cvaa is open source lexic is not open source they have their own model so this is the latest one and you cannot download it you need to use it in the website and you need to subscribe okay so the subscription in this case you need to pay to get access to the all this stuff the whole idea when we install stable diffusion and when we install automatic 1111 or Focus that we want to do it locally using the upu to stress also you need a good upu you can do this with the CPU but it is very very very slow so I recommend you to have a GPU at least eight gigs can work with lower money uh memory but is let's say that uh a JX probably will give you the best performance so this is Lexica and you this I use use it mainly for prompts so let's say that I was talking about dinner and thinking about pizza so I put the pizza War there and then you click there and you can get know some prompts and ideas what to put as I say everything here is using Lex from time to time probably they use another library but you have an idea of the promps on times you have the negative prompts and so on so we're going to address those Concepts and then also I like this this website night Cafe okay it's a fantastic one I use it a lot it's something similar to Lexica but here you have two type of subscription it's free and then there are they have challenges here and you enter and you you participate and you can get free points and you engage with the community and also the subscription is very low so I to have a subscription here and well talking about my favorite foot here and have some very nice ice creams and let's say here oh a nice cake wow beautiful graphing and here you can get the prompt uh not always all the users uh show The Prompt so let's see for instance this one let's see this user is showing the promp as you go here okay C so as hi in the problem but you get the idea that you're using this library and so on so this is my advice just get this website get familiar with the uh get familiar with this concept you need to be an expert in machine learning and the ne nwor everything has been done for you but you have everything there so now that we have this small introduction now let's use the libraries so this is my computer this is the where I have install everything so as you see I have different versions so I have automatic 1111 I have these two versions Focus I have confy and well these are old versions the one that I'm using to experiment so remember this stuff can be big uh and be careful that you need is a space so first let's work with Focus so I will go here and remember that you're entering your here you have your installation and just to to to launch Focus you click here okay and this one it's going to launch the user interface and also it's going to do an update and if there are new files it will update everything and you have this additional it will download extra Library so let let me launch and let's work with focus and while it's opening okay you have here we have here our user interface and everything since that is working and you have all the messages and look at the information here I mention that works in upu is recommended to to have the upu to get the best performance let's talk about a little bit the directory structure so as you enter into Focus there is not much that you need to know so I recommend you not to mess around but us it if you erase something automatically when you run again uh the script is going to download the the files that are missing so something important talking about the files the models and lo librar so all that stuff they can be big so usually in my case look at that I have many uh many user interface or applications Library so I have a lot of spa uh I use a lot of space which I do have but maybe my it would be better just to to centralize everything so usually my practice is to put everything in automatic 1111 and then I can share if I have extra libraries I can share with automatic 1111 and this is what I'm doing so I only have the base libraries of focus and then I share everything with automatic 111 so for instance uh in focus is you want to to change the user path where you have the libraries look at that you have this file here and you can change the path okay so if you have a different path you can move everything to that path and you will find information there so you have also many options here so feel free to experiment something remember that if you damage something you can erase it automatically will download delated file it will update everything but be careful uh then there is another way to do it that probably will be let's say the the most most advanced way is you go into models and you have many Python scripts and you feel free just to see what is happening here but the interesting one here is this path so as you go in that one pass python there as you can guess maybe we're going to have the path to the libraries and so on so is you right here see that you have the location in this case you have relative path so if you want you can also modify directly here but unly I prefer in this case just to use this user pass confit I feel more comfortable but it's up to you and you need to do it okay if you want to have different installations and if you are not downloading many models just have it separated this probably the the is the saf safest way the other uh directory that I want to show you here I don't go not going into details in every all in every directory is models so as you can guess you have here the models so if you go into checkpoints here you have your big model so recall that for instance you want a new checkpoint you can go here into Civ AA and you can download a new model whatever model you want you click you download and then you can put that model directly here and then when you open your application your user interface it will recognize so let me go here and I click advance and see that you have this Tab models and here you can access different models so in this case I'm using this one the default value model here is jger now XL but using this one and this is specific model use a refiner so a refiner is an extra library that it will improve whatever you have in your base model so refiners be careful refiners they do not generate an image they improve this so if you put a refiner here it's going to give you an error and not every big model will have a refiner so here youer now this one they don't have so depending on what you have you have different actions and the end is here for the lotas so you have different lotas and location and you can choose which one do you want to to use so I want to mention also that probably you're seeing differences from what you have because my installation already download some other Live Extra models and additions and extension but feel free just to to update it so the L when it comes to to to L uh it's pretty much the same you have here lotas and you put it there okay so look at that likely Pro I think you are not going the original installation you have anything so here already put some models and pretty much it is the same now if you have control net here you have extension for control net that we are going to look about to talk about that later which is it is impressive another interesting uh directory here it is style model so it Styles oh sorry no I was in the wrong models okay nothing else to add here okay so these are these are different concepts just to improve your output so if we go back here another inter and director this one sdl sty Styles and here you have all your styles that is your let's say your digital uh fingerprint so you put a text here but then you give an style like eer realistic and a style of Bango whatever you put it here so this is the tricky part to get those Styles some people use track GPT and whatever but it's up to you to to to try to try but the advantage of focus that I way I like is that Focus have many preconfigured styles so here you go you have all these Styles and you have many styles here so you see that you have a prompt this means whatever when you use that style whatever you put in the promp will be substitute here or you just can copy and paste and have your and negative remember is what you don't want to put there so don't do double negation on negate two times it's not like no wly just wly is no negative and you have many St there and you would see that they do make a difference so be careful about that use Styles but also don't overdo when you put in a style but to stress that this styles that comes with Focus are fantastic and you have many styles there and actually I use them as a reference to move to automatic 1111 that that one that user interface doesn't count with a style so I like to to look at this one so then well we talk about models Styles feel free to play with those Styles and then here you have Advan options so also remember that you have this document here that will be kind of a help it's very useful and actually invite you always to click when you see it to click there then here you're going to see this guidance scale and sometimes you see steps that probably when we move to automatic 11 we are going to have more more options this is one of the other big advantage of focus that it do you need to to deal with all those options it has been already optimized and with the default parameters it works very well but if you want to know what are those parameters uh let me go here I have a couple of figures just to show you these figures you can find this in internet but also you can generate it okay later this is with automatic 111 so this in our case does not apply this is the one that we want so see that CFG this is the scale this is the influence so the larger the value the more details that you're going to to get but maybe you're going to oversaturate your image so try to avoid very large values honestly default value that you're going to find seven is okay sometimes I go to 12 I don't I don't like to go to hi and you can guess the larget also the more time consuming is going to be uh the image generation so this is it and I think we are on time just to do here a first prompt so I would like to do a prompt let's say say related to Nature so I already have here I mentioned that I like to have my prompts and I have some Landscapes here and let me do this one okay I have a few prompts that I generate let me put put it here uh here in advance you have random so you can select this one and then this will be your sit number okay so always now this will be repeat the same model and then you can pay play around with and let me do just one image and negative promp uh uh I have the default by Focus that it is I know enhanc is the default is you are in doubt remember that you can go to your uh installation and here we see Focus enhance it's just negative prims so but if you want you can put here aoid no uh put no there because you're doing a double uh a no double no so let's say that for instance I don't want table okay maybe we'll put a table and let's click generate and this is it off you go let's see what happens also is mentioning that this runs in the GPU so here you see that my U is going on fire not very often I see like this but yeah she going crazy and there you go we have a fantastic image I have to say that the results are it is impressive this fantastic I have no war to describe but this is what we get with this one and feel free at this point just to change your stats and you will see different outcomes so just to get consistency you can put the same set number and then you will see no probably in Sim but you can try different things and to show you that let's go to the default style so I don't want anything here let me use this and I don't okay I want to use this one I know this style you know it's a h realistic style and if you are in doubt you can go here and let me go let's see if I can find it here okay this one this style so this is to do more realistic images so important look at that is way much smaller now that this file size of the checkpoint and the idea is just to improve your input in my image so you have an image and then you have the N networks and everything talking together and you improve that so let's see what happens that now I change to this model and now I also use this L you can combine different L libraries by the way and you can change the weight and so on so let me click here generate okay so we have the image here well this thep usage as you see it's just using the GPU bursting your GPU so important here that probably you realize that this one it was slower okay so depending there are some libraries are much much heavier than the other so there is this different so this joger n in generally speaking is a very good one it's the latest is called sdxl No table diffusion XL extra large I guess will be that XL then you have the standard stable diffusion which is a very good one which is the 1.5 so this one that I used here previously I know it was uh X XL I don't have the 1.5 here so when we move move to automatic 111 we're going to see that one so this is this is it how how how do how you play very important as I mentioned prps is they're very important your negative thr and so and the Styles so this is a big advantage in Focus that you have all those Styles already embeded embedded there and they are really good so feel free just to experiment with all those Styles and there is a link give me a minute okay we have this link it's a fantastic link and you have here all the styles of uh of falus which also will be equivalent to the styles that you put in any user interface using the stable diffusion and the same model so this is fantastic and to stress that this you can do it now using well this is more with automatic 1111 you can create this uh this Matrix of all this stats so fantastic you have all your stats will be up to you to pick out one so it's a really really good reference you will have it in the video the description and to stress that this is not uh only a restricted uh to to to focus can be used with any stable diffusion just copy and paste know the keywords you know where you have it we look at the source code so use those keywords and yeah go crazy uh and now before moving to to automatic 111 I just want to show you another case and this is very particular like it a lot I have it here and I want to do text so it might look something and let me put this one this simple text here an apple a Blackboard with the text hello so look at how specific I want an apple but an apple is kind of an astract because can be one but also the the the the the the language model that you have there my interpret that a two three whatever would put more Apple so let's see what happens but let's focus in the part of the text because getting text is not easy and later we're going to see some techniques just to try to get the the text to improve that or to force to have an specific text so look at what we have here it put the text on the out it didn't put it on the Blackboard but you can try okay you do another another iteration but look that is not properly the war you say hello putting something else and believe but this is not easy to do to get test uh to get tech text using all this mobile text to image it's quite tricky and the best way to do that and here I still have the table keyboard and let me put here and to do a little bit more complex so let me put this one more specific no so now I go put this apple use color lithograph uh Renee mget style and a Blackboard in the background with the text so now it's kind of specifying that I want that Blackboard in the background and try to put that that text there and the test is s pom pom okay a little bit more a little bit little bit no way much much complicated so let's go here random sit and let's see what happens okay here we have the two images and see as I mentioned about pra you know this sty proba you see the similarities with this specific artists and so on but fantastic fantastic I have to say this images but then the problem the big problem is the test the texts and and I'm I'm believe it's extremely difficult to put it and don't get frustrated if you you manage to put Tex text because it's really difficult but then using control net that will prepare the video how to install but this extremities I think things will get much much better so this is it how we use it and at this point I'm done with Focus I give you some guidelines important just get those prompts and get those Styles will be your personal style your fingerprint your digital artistic fingerprint so at this point let's move to automatic 1111 okay now I mve to automatic 1111 I have my installation here and similarly Focus you're going to have the scripts there here you have an extra one update so you click there and it will update your installations there updates so in this case there are a few days there let me press it was updated and now click run back and it's going to launch your your user interface automatic 1111 stable diffusion and while it's opening that and also to stress I recommend you to use uh Google Chrome or in my case I prefer to use Microsoft Ed I found it that is more efficient but it's up to you Firefox is the one that performs the Warth in this case so I don't like to use it but it's up to you they all will they all are going to work uh so similar to focus and well here we have the user interface let me put it in the background there let me minimize here and look at the directory organization uh similar to focus now you have here in system you're going to have all the python also in Focus you have directory where you have all the python remember here you don't need to install any extra Library this is the big advantage of the installation method that we use okay if you follow our videos you will see and as you go in wet UI here is where you have all the DAT and similarities that I want to mention here models similar to the one that we found in in Focus here you have this dories in this case we going to have your stable diffusion checkpoints in this directory so let's look at that already have the checkpoints that I have at this bit one so it's a lot of data that I have there then you have similar lot of models and so on control net and stuff like that later we're going to see so the concept is pretty much the same everything in models uh similar also you go here in modeles you're going to have this file the pass okay the python scre that you can uh here that you can adjust adjust the the the path where you have all your libraries and some other options so feel free to modify but honestly when it comes this one I this is my main installation and I I like to centralize everything here and then all the other user interface I like to point to this one uh then we have another directory extensions and here you're going to install those extension those are additional sourcewareinfo extension here control net and then within this extension you have modeles and you can put more modeles there so this is how it works important folder then you have this folder outputs very important that you also have it in in Focus so this folder be careful that I recommend you to clean a little bit every now and then because you can get a lot of data so here I have a few images but I have a lot of data so these are the images that are put there automatically not not big a deal same with focus and another important script let me see that we have here it is Styles okay the Styles here you have it you need to create this file style CSV you don't have it okay this when you have the original installation you don't have it but I put it here and you can put a Styles like this in there so I'm using this Styles we have this YouTuber Sebastian C that he does a lot of videos about uh generative air and very good I strongly recommended to visit his his channel and then using his style you can put it there but by default you don't have it and if you want you can use the same styles that you have in Focus just copy and paste you see the syntax is a little bit different here is a CF CFB file focus is a Json file for those who knows where is a Json file but just copy and paste and you will have it so yeah this is the general organization and now let's move to the graphical user inter so as you see it's way much much different than the one for focus is more Advan so focus it is just one window in the prone and then you have everything there and you're going to get good good images at the first try you I can guarantee that here since are different trick so here you can select your mobile uh this is the default one you have the jogger now and you will realize here you use here the Jugger now is way much slower than in Focus I don't know the risk is a little bit strange but you can choose your models here and then you go down here look at that you have L you can access the L be careful that those uh extra additions like L and eer networks on textual inversion all this stuff they need to be compatible with your checkpoint so if you're using the stable diffusion 1.5 you need to use aot compatible with that if you are using xdx sdxl the lenx to be compatible and so on so similar to one you have to to to focus you have text to image you put it here and then you have negative problem in your original installation you don't have any Styles in this case I have styles because I put that that file there but you need to have those styled remember that is your promp then you have this ENT here probably your user interface will be different because I have many extensions so later we're going to see but I have like aspect rtio the control net fantastic and I hope to do a video in that then you have have image to image you get an image similar concept to text to image extras that probably this is O scaling and B prop processing this is very important PNG and this is specific for automatic 1111 sometime not sometimes by default in the settings it is configured to save the metadata of the image here that you don't have that in Focus but in Focus you have a lock file but let me show you that if I go here and remember that you have in W you have output and you have all your outputs there and let me get one here and I'll put it here you get the meta data instead if I do the same with Focus that just look at you have the output folder and I put the meure here you have the metadata the advantage and focus that Focus automatically is going to create this lock file let me open with firef there and there you have all the data no nice HTML with the data this is the way how you can recover that data so I forgot to mention that in Focus you have the information there so you have more information you can train retrain your model you have general settings here just talking about word the files or the default format and everything you have here images you have the path also default path you can see you can change that uh and more stuff something interesting or important this one live preview so you see when you are generating the image you see that Evolution here you can control that so if you put it in zero you're not going to see or negative nothing so the default value is 10 but feel free to to to play around with this auction just to get an idea you can also go to the website and the website you you have a Wiki key and you all that information you have it there or not all so sometimes you need to dig a little bit more to get that information uh and then extension and this is important T this is where you install those famous or famous extensions so this is what I have installed so probably you're are going only to have this I think and here I installed those extra so we went through this also in the origal installation video so you have here available and as you load you have all the extension that are already safe here in this file that can be updated in gith hot and let me deselect here in install and you can see all the extensions so for instance if you want to install and let me go back to control net and kind of this is the control net installation and you find it that you're going to have it there install your control net and that's all so you have many many there many options there for instance animate this is a nice library that you can do text to animations fantastic C you can install so feel free to explore everything also you can update so as you go here here check updates and you can also get give directly the the path and so on so let me see if there are updates here and automatically will update I have an update I'm not going to install that now after I finish this video and that's all now this is a brief uh the brief explanation of the guy important that now here you see more options now contrary to focus that everything was hidden they are not exposed here you have everything exposed so it will give you more control so this is similar to the one in to the one that you have in Focus that is also default seven but now you have steps and also sampling methods so let me go I have the images here so there are many sampling methods also you have that option you can control that in in focusing the Advan debugging actions but the the the default one work fine so this is what happened different uh different sampl methods and talk about that this kind of Matrix you can create here using uh using automatic 11 l so usually different methods no sample method will give you different outputs there is no nothing written which one is the best one the tendency in the community says that this is the best one but I think it depends what you're trying to do or your image but what is important steps steps is very important the more steps the the more F Fidelity that you get but also the more time consuming so usually 20 is a good value I can go up to 30 but yeah you can change that Val and similar you have access to cfu okay and control that and then also you can control your aspect rate and there are more more options to get know even better images so also I will talk about o scaling which is fantastic that you can do at o scaling so let's do something like in the focus so let's do the nature as well so let's use the same case and I have have it here and let me put it here Focus Majestic and here you have default negative prompts you need to put it so it will be up to you so if you want you can use the one for Focus so I have it here Focus open and let me just to show you let me use the default focus and let me put it here put it there or if you want you can use no in my case that I have it there you can put those Style a little bit about the syntax when you see this parentheses there it means that you put all this words in parentheses and then at the end semicolon and a number so this is the weight Factor so kind of multiply 1.3 times all these keywords to give more weight this is how you put it this is something I think is also coming from from uh from MC Journey they have something similar so sometimes when you see this one that is what is happening so you can add or also reduce white so I let me put this one you can increase here to generate more images so let's use the default value something interes is also to later more advanced stuff so talking about those matrices you can have here okay so a little bit advaned but later we address that so I click generate and off you go so in using this Library you will realize that this is very fast so recall focus focus was slow now this one is way much faster so we get decent results a fantastic result and if I click generate again it's going to generate another different image because we have a random set so let's see that I don't like this one I know why it's putting that there let me go and generate another one and this is idea or you want you can put a batch of four and generate similary images and that's so so when you're happy with one for instance you can click here and this is your sit number and just to have repeti ability okay always you were going to generate that same image just to show you you generate the same image or probably you can now play with these parameters and look at what will happen when we reduce the sample detail you get something this different and if you increase it also likely better quality not necessarily because you might oversaturate those images because it will start to to get too many details okay fantastic I think I like this one uh then you have all these option say you can you can play with this option later we we play with other Advance auction so I'm happy with this uh image here and this is how how you work so now to show you let me change now to the jger now Library the jger now library is the default in Focus it's not the default here the default here is the stable the fusion 1.5 and you click there it's going to load the library and now that you have it you just can click generate and off you go and you will see that this is way much slower so you can look at your terminal here you can see what is happening sometimes is you're missing some features uh it will download everything automatically here I have already everything but uh I might happened that it will be a little bit slow but because it is unloading Sun scene besid know that that it needs to get access to the to the GPU and at this point I think I have to wait oh boy that was time consume so as I was telling you sometimes it might happen that automatic with this specific no Library model will be more time consuming probably also be careful about the the negative prom sometimes they the your proms can can have an strong influence and actually it was also very very very time consuming because I'm using also a large number of sampl steps so it try now the IM generat the mod try it to to stay very uh very close know to the input text that you have here in the positive and negative PRS let's say so be careful with that so now that you get an idea and let me reduce this one I want to do something super time consuming remember that you have the images uh in that output folder and probably you can compare know both Library both user interface Focus the stable diffusion see the outcome but what I want to show you now that you have no the standard Library let's do something a little bit more a little bit different from the previous one so this specific One sheese D is okay let's see what is happening so remember always the Fantastic website that we have here recall also this Styles will also if you use the same keywords it will will apply as well with automatic 1111 so let me go here and let's so this specific one that I want to use is this so you have all your information checkpoint start so and so on so this sh that the it is something specific for landscape nature so as I'm doing some landscapes let's use something that has been specifically trained for that kind of of images so this is the diversity that you have that sometimes you're going to find some checkpoints model specifically know for some some type of images or you can have the the super gener models so let's use this one but also let's use it with a lot of model so remember that those models extensions that you using need to be compatible with this this this model so look at that this is about two Jigs and to use this one I selected here I will erase all this negative prom because sometimes they can they kind of slow down everything or let me leave leave it there I I have my prompt there and let's say that I want to use a lot of model so you have all this stuff here and first let me let let's generate the image okay the standard one and here in L you have there so I'm using this one Let me refresh okay nature prono levels and it's very particular how you use that model so let me go here so you are generating the image and actually we have it was quite fast okay you can see that that jogger now was super super slow probably also was the influence of of the sample step but interesting here that this is what we generated a standard input and now let's use the load extension in that one so it's always text to image here and you click here actually here you have infos and information and then also you you can set up SS in image how you activate the library that you are using remember that they need to be compatible with the model that you are using and so but to use it they do have an activation key that actually you can put it here activation text or you can click here and see that now you app pen this and mean means that you are using that model so if you recall Focus that there was a weight 01 one that's pretty much the same but this is how you act activate it and let's look at this specific one let's see what does it does so you can always come here or hug human phase and this let's see the one that I'm using now n Pro this is the one I'm using so it is a Model A checkpoint in a lot of libr specifically trained to make S more realistic no Landscapes so talking about the size I don't have it here but it's very small uh and then well you can have some information how to activate the trigger war and so on so this is what we're going to do use a additional Library model just to make it let's say if we make something way much better so saying random number set number and generate and let's see what happens so all these images we can always go now to the directory where we have all the output and we can just be sting in spad so and here and the two last images so this is here I can close this one so this was with no Lo model with L model and kind of I think yeah this one probably seems a little bit more realistic Maybe yes maybe not but yeah that's the point okay you can play with this one and now let's say that let me change this when to put 0.5 the weight and let's see what happens and the influence clearly but yeah there is an influence but pretty much the same so this is the idea no this L and there there's much much of stuff that hopefully we're going to to cover uh so just to end this video Let's redo also the casd promp with the with the Apple so I really like that one and this the same we're going to get the same stuff but promps here and but here as I recommend you that to have those keyboards so use it for me now but this is almost a universal negative PR that always you have around and I want to use this one the easy one I'm not going to put negative word I'm not going to use any any extra models um be careful that you need to switch now to the stable diffusion standard Library I'm not going to use the jer jer now is kind of heavy and generate so this is relative fast and there you go and as I say putting text is in incredibly difficult it probably you will get the idea using focus is easier and you can get something closer here is way much much difficult using automatic 1111 but you can get there but there are specific techniques to to do that and now let's go to the one and well uh you can put Also let's randomize here there and let's see what happens with a random so I asked for one apple and look at that it's putting more than one apple the text is way far it's not even a blackb this one one apple kind of a black bar but DET text not even close so this is when things get tricky uh now let me put here and generate so this for sure it will not be not even close you can see here I don't didn't put the text the apple is okay kind of there is something the background here not even close and so on so probably I increase this one might get an improvement so see that sometimes can put some water mark so you can get there and put negative PR that water mark try to avoid the water mark so at this point let's see if we can get something much closer to the previous so let's use Focus style and I want to use Masterpiece okay let's use this style and let's skip this random number let me add here and the Master P the prom I can erase that one and let's see what happens with this one when I add a much refiner style well didn't impress M much but see the difference there and you can also add the negative promp let me go there and let's see what happens with this negative prom okay so look a bit different now add that negative prom DET text is the most difficult thing to to add uh it might be a good idea let's add this L uh model which is not related at all but let's see what happens when I add that so not much different kind of probably is oversaturating so let me go erase all this stuff and go to the original one and just just stay with this one so looking now negative keywords this is that we have low R so I don't want low resolution and so on and let's use another model just to show you so this one that kind of a little bit familiar with that one but don't want to go into details not much difference yeah you can see probably there there kind of the pixels know the little bit more ground so yeah at this point I hope you you have an idea uh remember prompt is important all this is part of this prompt engineering so avoid asra Concepts and convey an idea object action verb nouns subjective not be yeah try to to know just to give that clear idea because you have there a language model that it will give an interpretation to everything and negative words you can put it there so there is a library of a standard negative words and and let play play around with the models there are many of them and go to CIT Ai and look for the best models and get your digital know fingerprint for these Styles so I think uh found this video helpful the next one will be for K control net control net is fantastic it's crazy and that is what I'm using the M right now so some control net animate the diff to do the animations and then we kind of explore more more advanc Concepts so thank you hope you enjoy it I hope you found it useful and if you find it useful do not forget to follow us and subscribe to our channel so I'll see you next time bye

Channel: Wolf Dynamics World

Views: 3,457

Keywords: AI, AI art, stable diffusion, ControlNet, generative art, ai art generator, run ai local, air art local, SDUI, install ai, install stable diffusion, automatic1111, fooocus, civicai, lexica, nightcafe

Id: qYAw18T-eEs

Length: 52min 44sec (3164 seconds)

Published: Sat Nov 11 2023