Fooocus Tutorial - Stable Diffusion Made Easy

Focus was created with the idea of bringing the Simplicity of mid Journey intuitive and easy image prompting to an offline open- Source software in other words all the goodness of mid Journey for free using the power of your own computer while there are some great programs out there that let you create works of art with stable diffusion this way such as automatic 1111 and comfy UI they can be overwhelming for a beginner automatic 1111 is a bit easier to tackle but it also comes at the cost of needing a higher base bam of at least 6 G gab if not 8 gigs to accomplish anything where Focus can work with as low as a 4 GB Nvidia GPU however 6 gigs is still recommended these other programs can go deeper with stable diffusion but focus is very strong in its own right and a perfect place for a brand new user who just want to create awesome AI generated art without the hassle of a complicated UI or trying to figure out this it should be noted that the program was specifically written for Windows and Nvidia GPU users AMD is supported but not as extensively as Nvidia there are steps to install on Mac but it is also not as supported as Windows here is a quick pros and cons to figure out if this is for you Pros easy to install and has a userfriendly UI lower vram requirements of 6 gbes but 4 gig can work with some setups has its own embeds that can help get great results with shorter prompts similar to Mid Journey cons it really comes down to the limitations while it's still quite extensive ultimately Focus can't match the power and customization of something like comfyi with all that let's get started installing focus is very simple coming over to this GitHub page and scrolling down to here click this link and download the software for Windows if you have an AMD GPU there is some changes to the Run bat once you unzip the file pretty simple just rightclick and edit the runat file and add these lines I don't have an AMD GPU so I can only hope it's that easy unzip the file to wherever you like you will then have these three bat files run bat is your go-to for everything but you also have run anime and run realistic for now we will just stick with run bat go ahead and click it the first time this is run it will install the necessary files and models it needs it will be a 6 gig download same goes for the other two bat files if you use them also if you do have automatic 1111 or comfy UI installed you can edit the config file this will let you share the models between these programs so you don't have to download them more than once which is incredibly useful as you can EAS easily clutter up your hard drive with hundreds of Gab worth of models since many of them run around 6 gigs each make sure when putting in the location that you add the double backs slash it is needed for the config file to work if you don't or even miss one backs slash you will get an error in the command window that will tell you it couldn't load the config file okay now that focus is open we can start using it right away type anything you want here let's say a Viking Warrior it will load two images like so they don't look too bad background is so so but still so let's look at the advanced settings here you can change the aspect ratios easily clickable versus having to input the numbers yourself down here is the image number this affects how many pictures it will create at a time let's stick with two for now at the top is performance here is where you can increase quality or increase the speed obviously more speed will be less quality but also less GPU vrm at the bottom is the negative prompt Focus has its own embeds that do a pretty good job by itself so you don't really need to use this but the option is there the random box is the seed which will make for more random images with every generation you can uncheck it if you would rather the seed stay the same next is the style menu these will influence the look of the image and you can get completely different outcomes let's try try digital art H okay well how about pixel art wow that's quite a bit different so you see you can really get a variety of styles and looks from all of these let's try one more game fighting game okay that's cool lots to try out and mix and match for sure then there is the model tab um you can only use sdxl models for this I don't have too many XL models myself but Juggernaut is a good on its own you can pick a refiner but you don't need one for this model you also have luras you can choose this comes with with the pre-selected Laura here for sdxl at the bottom is refresh all files if you download new models or lauras while this is open you can hit this refresh button and they will show up in the list in advanced tab you can change the guidance scale and image sharpness sliders increasing guidance will make the image more vibrant and artistic and increasing image sharpness will do exactly that and sharpen the image clicking the document here will bring up a very in-depth explanation of these two sliders don't worry about debug mode now down here we have this input image box this brings up even more options the first tab is upscale or variation this is where you can upscale the image by 1.5 times or two times you can load a new image here or just drag and drop a generated image from up top and you can see it upscaled the image a little variation but still pretty much the same you can vary the image either subtly or strongly we can see if we can add a brighter background maybe H doesn't look like much of a change let's try the strong setting okay that definitely changed it quite a bit but still true somewhat to the original the other tab is image prompt here you can combine up to four Images together click Advanced to get the extra settings now we can combine multiple images and okay well there's that not quite what you would think let's try again a little more accurate to both eyes kind of ugly but you get the idea we can try face swap and see how this goes the stop at and weight settings don't really have good documentation that I've found so you will have to mess around with them yourself uh I am getting something but not quite what I want I got better results earlier but not so much now but I hope you get the idea here we'll move on to in paint and out paint here you can use MKS to replace an area using a new prompt I will try to get a female face here instead that worked but weird random braid there oops let me see if I can fix that hopefully that worked a bit better yeah that's okay you can also try to improve detail with the in paint prompt and also if you want to uh add something into the scene like a castle just going to mask this area out and type in Castle in the background and hit generate that worked nicely the last thing is the describe which gives a description of the picture pretty simple and that's about it that is focus in a nutshell I hope this information was a bit helpful for anyone here just getting started or wanting to know a little more about Focus if you have any questions or comments you would like to leave please do and I will do my best to answer them thanks everybody

Channel: Jump Into AI

Views: 2,731

Keywords: stable diffusion, stable diffusion tutorial, fooocus, focus tutorial, ai tools, ai art

Id: 828wYIp2HAM

Length: 9min 21sec (561 seconds)

Published: Fri Dec 15 2023