Simple Guide To LoRAs In Stable Diffusion | Fooocus

today we are going to go over luras I'll show you where to get them how to install them locally how to use them properly and combine them with one another and even a small not recommended bit about using SD 1.5 luras and with that let's get started first we will go to civid AI click on the models Tab and to make this easier we will go to filters and make sure we check the Laura tab and go down and check the sdxl 1.0 tab make sure everything else is unchecked now you should only see sdxl compatible Loris there are several you can choose from some will give you a specific character or art style While others can be used to enhance an image in some way such as adding or removing details or even one specifically to improve eyes when you find one you want go to its page and you will see a download button here click that and download it to your Laura folder if your model locations are set to the default it will be Focus models lorus then save if you are like me and have another stable diffusion software and change the config file to point to a different folder then make sure you save it to the correct one before we leave this page we want to take note of a few things that will help using this Laura first we want to make sure that the base model does say sdxl 1.0 and then sometimes you might see this a trigger word this is the word or phrase that we want to include into our text prompt to activate this Laura then check the text description for any other helpful info here we can see that it recommends starting with a 0.8 value sometimes it may recommend a specific model or resolution for best results now let's go to focus if you had it open while downloading the Laura let's go into advanced model Tab and make sure we hit refresh all files as long as the downloads are finished this should let us use our new luras now I will select the ALF Laur I just got and put it to the recommended 0.8 and use the trigger word of Al person and there is our Al from there we can continue adding more text prompts to put them in different scenarios now let me switch it up with with a different Laura the Jean Baptist Laura this is a style I like and even inspired some of the thumbnails I have used on the civid a i page this also has a trigger word to use there isn't a mention of any other settings but looking at the pictures here we can click on the little icon in the bottom right this will give us the settings used to create this image we can see they used a one as the weight so I will start with that back to focus I'll put the weight in and then the trigger word and add a little more to the text prompt now it looks good but not quite what we are seeing on the civit AI website and that is because we still have our other styles checked these are great for realism but can alter stylized characters and images too much so let's uncheck them and try again and now we are looking uh more like the other images these look good to me but just for reference we can try to decrease the weight and see the results and then we can increase them a bit not a big difference but I still like the weight at one now let's try adding another Laura the detail tweaker XL Laura works a little different this one makes use of positive and negative weights typically aura that adds a style character or something to the environment will only use positive weight values unless otherwise stated on their download page so on this we will start with the weight at minus two and we can see that it is removing detail and makes the image a bit more plain and flat setting it to zero will basically deactivate it but let's crank it up to two and see what it does and if we compare these three from min-2 to 0 to+ two we can see that at two it adds more detail in Brush strokes and makes it a bit more on the real side as the minus two makes it a bit more flat and on the cartoon side obviously results are going to vary image to image saying it adds and removes detail is vague because it's not always going to change in the way you think sometimes it will crowd an image with more objects when the setting is high and on a low setting it might just change the clothes and may not seem like less actual detail so again it's always going to be playing with the settings and changing the text prompt if you aren't quite getting what you hoped I will add another Laura on this and this is a styled Laura that specializes in creating this dissolve effect so I will add the trigger word and start with a strength of one and we can see it start taking effect I will turn it up a bit more to 1.5 and we get some crazy results it's one of many luras out there that can really boost your creativity I think lots of people might Overlook some of these options usually assuming allur is simply to get a specific character or art style we also have another detailed type Laura called perfect eyes XL this tends to put more color and detail into the eyes using this myself I've had the best results with 2D and artistic styled characters it can work in realism pictures too but it's not better every time the last bit of information here is going to be something I wouldn't recommend unless you are really bored or you really want to try SD 1.5 Loris and have no interest in using any other SD software I say this because the results will be all over the place and not practical but the option is there so I figured I would show you I'm going to reload my page to reset everything then I'm going to go back to the model page and I will set a 1.5 checkpoint as a refiner for this I will use dream shaper 8 also I want to make sure to remove the default Styles then I will add a Laura this Arcane offset is a 1.5 Laura I will copy the prompt from this image and then paste it in our text prompt and generate now the end result will never be as good as using this the correct way you will get artifacts and especially multiple faces and if you are really really determined to keep using this you can even go into the debug menu on the debug tools tab go down to these two settings with this we can set a temporary override resolution of 512 x 512 this will slightly improve the issues but obviously at the cost of a lower resolution again this isn't something I recommend as it's probably more trouble than it's worth but it's there and that's all I have for you guys today hope there was something useful here and I will catch you in the next one

Channel: Jump Into AI

Views: 2,427

Keywords: loras, stable diffusion, stable diffusion xl, stable diffusion lora, fooocus ai, Fooocus tutorial, ai image generator, stable diffusion lora install, sdxl, sdxl 1.0, sdxl turbo, sdxl lora, sdxl local install, fooocus Lora, Fooocus Lora Install

Id: 3YoSj3g8EHg

Length: 6min 25sec (385 seconds)

Published: Thu Feb 01 2024