Fooocus Updated With SDXL ControlNet & Img2img Support (Installation Tutorial)

today we're diving into the latest update of the fuocus web UI which is now compatible with sdxl and boasts new control net models plus it's got the ability for image to image or as they call it here revision a quick heads up the focus web UI has been optimized to run smoothly with a minimum of four gigabytes of GPU memory for stable diffusion this makes it great for those with Hardware that might not be top of the line but hey if you're rocking an Nvidia 4090 you've got nothing to worry about we're installing it on a Windows PC just scroll down to the download section and click the link to begin the download it's a hefty 1.6 gigabytes so be patient if your Internet isn't Lightning Fast you'll get a 7-Zip file that contains the entire fuocus web UI unzip it using WinRAR or 7-Zip your choice for this tutorial I'm unzipping the files on the C drive the first time around you'll need to click the Run MRE bet this will open the windows command prompt and download the essential models required for image generation first it'll fetch the sdxl base model then the refiner model followed by Laura models like clip vision for Focus revision once everything's downloaded you'll see a local URL copy that and paste it into your web browser hit enter here we are in the focus web UI it's pretty straightforward just a whiteboard with a text box below for your prompt now let's put it to the test I'm going to try creating an image of a samurai standing atop Tokyo using just the sdxl model without any additional checkpoints or Laura it's loading and let's see how it turns out by default the refiner model is activated in the web UI which adds a bit more time to image generation alright that's pretty cool both images have detailed backgrounds which is what sdxl excels at you can spot the Tokyo Tower in the background even though it's blurred the focus remains on the main character certainly the focus web UI seems ideal for those with more modest requirements or stable diffusion beginners who don't need all the bells and whistles of comfy UI and automatic 1111. now let's take the generated output and try some image to image Transformations it's incredibly easy just tick the check box and you're good to go no need for extensive explanations here now let's switch up the background this time we're using New York as our backdrop alright let's delve into the advanced features of the focus web UI click the check box under the text prompt to reveal all the settings you can choose between performance speed and quality allowing you to customize the number of sampling steps for this demo let's go with speed below that you'll find resolution settings with preset options for width and height you can also choose different image Styles such as cartoons Comics or in our case cinematic set a seed number and you can use random or specify one yourself one thing to note when you uncheck the random box the seed number appears in the text box no need to go hunting for it like in automatic 11 11. I'm using the seed number from the samurai image I chose earlier now let's generate it again to confirm we're using that seed number alright so far so good now let's move on to the image to image section click on the tab in the top right corner here they call it revision when you enable the revision checkbox at the top it says prompting with image which means we're generating images based on both the text prompt and the reference image I'm keeping all the revision settings as default and giving it a whirl this time I'm tweaking the prompts a bit we're going for a woman in a ninja suit with the same pose as the samurai after generating I pick this image it's pretty cool showing the main character from another angle gazing at the city from above it's a good example for sdxl control net now click on the CN tab in the top right corner I'm enabling both control net canny and depth these are the new control net models for sdxl remember when you enable control net you need to uncheck image to image the settings look good I'm leaving control net settings at default and now let's try a different prompt this time we're going with a woman in a fancy dress in a party room sporting the same pose as our ninja girl here we go same pose and you can see the depth in action with the party room ceiling lights her fancy dress looks stunning and funny enough her side view resembles Gal Gadot pure coincidence as I haven't used any Laura here that wraps up the new features in the focus web UI it's a simple web interface that supports image to image and control net while also being compatible with sdxl I think it's perfect for beginners and those with lower end gpus who want to unleash their creativity that's it for today's tutorial catch you in the next video and have a great day foreign [Music] [Music] [Music] thank you [Music] foreign

Channel: Future Thinker @Benji

Views: 8,023

Keywords: stable diffusion, ai art, stable diffusion tutorial, stable diffusion ai, stable diffusion xl, ai art thief, beat ai art, ai art generator, SDXL, sdxl controlnet, sdxl img2img, Fooocus Web UI, Image generation, SDXL compatibility, Control Net models, GPU optimization, User-friendly interface, Control Net features, Image generation tutorial, SDXL image generation, Control Net image generation, Creative process, Beginner-friendly image generation, Graphics software

Id: BXqpgDDjea4

Length: 8min 1sec (481 seconds)

Published: Wed Sep 13 2023